The imprinted Zdbf2 gene finely tunes control of feeding and growth in neonates

  1. Juliane Glaser
  2. Julian Iranzo
  3. Maud Borensztein
  4. Mattia Marinucci
  5. Angelica Gualtieri
  6. Colin Jouhanneau
  7. Aurélie Teissandier
  8. Carles Gaston-Massuet
  9. Deborah Bourc'his  Is a corresponding author
  1. Institut Curie, PSL Research University, INSERM, CNRS, France
  2. Centre for Endocrinology, William Harvey Research Institute, Barts and the London School of Medicine and Dentistry, Queen Mary University of London, United Kingdom
  3. Institut Curie, PSL Research University, Animal Transgenesis Platform, France

Abstract

Genomic imprinting refers to the mono-allelic and parent-specific expression of a subset of genes. While long recognized for their role in embryonic development, imprinted genes have recently emerged as important modulators of postnatal physiology, notably through hypothalamus-driven functions. Here, using mouse models of loss, gain and parental inversion of expression, we report that the paternally expressed Zdbf2 gene controls neonatal growth in mice, in a dose-sensitive but parent-of-origin-independent manner. We further found that Zdbf2-KO neonates failed to fully activate hypothalamic circuits that stimulate appetite, and suffered milk deprivation and diminished circulating Insulin Growth Factor 1 (IGF-1). Consequently, only half of Zdbf2-KO pups survived the first days after birth and those surviving were smaller. This study demonstrates that precise imprinted gene dosage is essential for vital physiological functions at the transition from intra- to extra-uterine life, here the adaptation to oral feeding and optimized body weight gain.

Editor's evaluation

The paper provides an elegant demonstration of the phenotypic consequences of genomic imprinting for postnatal physiology, focusing on the specific case of the Zdbf2 gene in the mouse. Using a series of gain and loss function models they explore how manipulating gene dosage independently of the parent of origin, influences the hypothalamic-pituitary endocrine axis, to regulate feeding behavior and growth during the early post-natal period.

https://doi.org/10.7554/eLife.65641.sa0

Introduction

Genomic imprinting is the process by which a subset of genes is expressed from only one copy in a manner determined by the parental origin. In mammals, genomic imprinting arises from sex-specific patterning of DNA methylation during gametogenesis, which generates thousands of germline differentially methylated regions (gDMRs) between the oocyte and the spermatozoa. After fertilization, the vast majority of gDMRs are lost during the epigenetic reprogramming that the embryonic genome undergoes (Seah and Messerschmidt, 2018). However, some gDMRs are protected through sequence- and DNA methylation-specific recruitment of the KRAB-associated protein 1 (KAP1) complex (Li et al., 2008; Quenneville et al., 2011; Takahashi et al., 2019) and become fixed as imprinting control regions (ICRs). Roughly 20 ICRs maintain parent-specific DNA methylation throughout life and across all tissues in mouse and human genomes, and control the mono-allelic and parent-of-origin expression of approximately 150 imprinted genes (Schulz et al., 2008; Tucci et al., 2019).

DNA methylation-based genome-wide screens have led to the conclusion that all life-long ICRs have probably been discovered (Proudhon et al., 2012; Xie et al., 2012). However, a greater number of regions are subject to less robust forms of imprinted DNA methylation, restricted to early development or persisting in specific cell lineages only. Moreover, imprinted genes are often expressed in a tissue- or stage-specific manner (Proudhon et al., 2012; Andergassen et al., 2017; Monteagudo-Sánchez et al., 2019), adding to the spatio-temporal complexity of genomic imprinting regulation. Finally, while the vast majority of imprinted genes are conserved between mice and humans, a subset of them have acquired imprinting more recently in a species-specific manner (Bogutz et al., 2019). Why is reducing gene dosage important for imprinted genes, why does it occur in a parent-of-origin manner and why is it essential for specific organs in specific species are fundamental questions in mammalian development and physiology.

Imprinted genes have long-recognized roles in development and viability in utero, by balancing growth and resource exchanges between the placenta and the fetus. Moreover, it is increasingly clear that imprinted genes also strongly influence postnatal physiology (Peters, 2014). Neonatal growth, feeding behavior, metabolic rate, and body temperature are affected by improper dosage of imprinted genes in mouse models and human imprinting disorders (Charalambous et al., 2014; Ferrón et al., 2011; Leighton et al., 1995; Li et al., 1999; Plagge et al., 2004; Nicholls et al., 1989; Buiting, 2010). Imprinting-related postnatal effects are recurrently linked to dysfunction of the hypothalamus (Ivanova and Kelsey, 2011), a key organ for orchestrating whole body homeostasis through a complex network of nuclei that produce and deliver neuropeptides to distinct targets, including the pituitary gland that in turn secretes endocrine hormones such as the growth hormone (GH). Accordingly, the hypothalamus appears as a privileged site for imprinted gene expression (Gregg et al., 2010; Higgs et al., 2021). A typical illustration of such association is provided by a cluster of hypothalamic genes whose dosage is altered in Prader-Willi syndrome (PWS). PWS children present neurological and behavioral impairments in particular related to feeding, in the context of hypothalamic neuron anomalies (Swaab, 1997; Cassidy and Driscoll, 2009). In mouse models, single inactivation of the PWS-associated Magel2 gene results in neonatal growth retardation, reduced food intake and altered metabolism (Bischof et al., 2007; Kozlov et al., 2007; Schaller et al., 2010). Fine-tuning hypothalamic inputs is particularly important for adapting to environmental changes, the most dramatic one for mammals being the transition from intra- to extra-uterine life at birth. Early mis-adaptation to postnatal life can have far-reaching consequences on adult health, by increasing the risk of metabolic diseases. It therefore is of the utmost importance to thoroughly document the action of imprinted genes, particularly in hypothalamic functions.

Zdbf2 (DBF-type zinc finger- containing protein 2) is a paternally expressed gene with preferential expression in the brain (Kobayashi et al., 2009; Greenberg et al., 2017). It is one of the last-discovered genes with life-long and tissue-wide imprinted methylation, and conserved imprinting in mice and humans, however its function is not yet resolved (Kobayashi et al., 2009; Duffié et al., 2014). We previously characterized the complex parental regulation of the Zdbf2 locus: it is controlled by a maternally methylated gDMR during the first week of embryogenesis, but for the rest of life, it harbors a somatic DMR (sDMR) that is paternally methylated (Proudhon et al., 2012; Duffié et al., 2014; Figure 1A). The maternal gDMR coincides with a promoter that drives paternal-specific expression of a Long isoform of Zdbf2 (Liz) transcript. In the pluripotent embryo, Liz transcription triggers in cis DNA methylation at the sDMR, allowing de-repression of the canonical promoter of Zdbf2, located ~10 kb downstream (Greenberg et al., 2017; Figure 1A). In fact, although specifically expressed in the embryo where it undergoes stringent multi-layered transcriptional control (Greenberg et al., 2019), Liz is dispensable for embryogenesis itself. Its sole function seems to epigenetically program expression of Zdbf2 later in life: genetic loss-of-function of Liz (Liz-LOF) prevents methylation of the sDMR, giving rise to mice that cannot activate Zdbf2–despite an intact genetic sequence–in the hypothalamus and the pituitary gland and display reduced postnatal growth (Greenberg et al., 2017). The cause of this growth phenotype and whether it is directly linked to the function and dosage regulation of Zdbf2 in the brain cells is unknown.

Figure 1 with 1 supplement see all
Zdbf2 expression localizes preferentially in the neuro-endocrine cells of the hypothalamo-pituitary axis in juvenile animals.

(A) Scheme of the Liz/Zdbf2 locus regulation during mouse development. In the pre implantation embryos, a maternally methylated gDMR allow the paternal-specific expression of the Long isoform of Zdbf2 (Liz). Liz expression triggers, in cis, DNA methylation at the sDMR which is localized 8 kb upstream of Zdbf2 canonical promoter. In the post-implantation embryos and for the rest of the life, the imprint at the locus is controlled by the paternal methylation at the sDMR and this allow de-repression of Zdbf2, leading to its paternal-specific expression in the postnatal brain. (B, C) X-gal staining on brain coronal sections from 2-week-old Zdbf2-lacZ transgenic males. The coronal diagram from the Mouse Brain Atlas (left panel) localizes the sections in zone 40 (B) and zone 47 (C) in stereotaxic coordinates, with the hypothalamus indicated in blue. Whole brain coronal sections (middle panel) show specific staining in the hypothalamus, due to several positive hypothalamic nuclei (right panel, 20 X magnificence). Hi, hippocampus; Co, cortex; Th, thalamus; Am, amygdala; St, striatum; Hy, hypothalamus; 3 V, third ventricule; Pa, paraventricular hypothalamic nucleus; Pe, periventricular hypothalamic nucleus; AH, anterior hypothalamic area; DM, dorsomedial hypothalamic nucleus; VMH, ventromedial hypothalamic nucleus; Arc, arcuate hypothalamic nucleus. (D) X-gal staining on pituitary horizontal sections. The posterior lobe of the pituitary shows no X-gal staining (right panel), while staining is evenly distributed in the anterior lobe (left panel). pl, posterior lobe; il, intermediate lobe; al, anterior lobe. (E, F) Zdbf2 expression measured by RT-qPCR in the hypothalamus (E) and the pituitary (F) from 1 to 70 days after birth. Data are shown as mean ± s.e.m. of n = 3 C57Bl6/J mice.

By generating loss-of-function (LOF) and gain-of-function (GOF) mouse mutants, we show here that ZDBF2 is necessary for optimal growth and survival during the nursing period, by stimulating hypothalamic food circuits immediately at birth. Moreover, our data support that the dose but not the parental origin of Zdbf2 expression is important for its imprinted mode of action. Altogether, our study illustrates the critical function and proper dose regulation of Zdbf2 for adaptation to postnatal life.

Results

Zdbf2 is expressed in the neuro-endocrine cells of the hypothalamo-pituitary axis

Besides a C2H2 zinc finger motif, the ZDBF2 protein does not contain any obvious functional domain that could inform its molecular function. To gain insights into the role of Zdbf2, we first examined the cellular specificity and temporal dynamics of its expression. Comprehensive datasets of adult tissues expression in mice and human suggested Zdbf2 expression is prevalent in brain tissues and pituitary gland (biogps in mouse tissues: http://biogps.org/#goto=genereport&id=73884 and GTex in human tissues: https://gtexportal.org/home/gene/ZDBF2). Our data endorse this brain and pituitary-specific expression of Zdbf2 in adult mice and show the highest level of expression in the hypothalamus (Figure 1—figure supplement 1A; Greenberg et al., 2017). Similarly, Zdbf2 expression was predominantly observed in the brain and the spinal cord of embryos (http://www.emouseatlas.org), a feature we confirmed at embryonic day E12.5 using a previously described LacZ reporter Zdbf2 gene-trap mouse line (Figure 1—figure supplement 1B; Greenberg et al., 2017).

Using this same Zdbf2-LacZ reporter line, we confirmed the expression of Zdbf2 in various brain tissues at 2 weeks of age, and more specifically, the high specificity in hypothalamic cells that belonged to the peri- and paraventricular nuclei in the anterior area (Figure 1B), and the arcuate, the dorsomedial and the ventromedial nuclei in the lateral hypothalamic area (Figure 1C). Analysis of publicly available single-cell RNA-sequencing (scRNA-seq) data further defined Zdbf2 expression to be specific to neuronal cells and absent from non-neuronal cells of the hypothalamus (Figure 1—figure supplement 1C-D; Chen et al., 2017). The three hypothalamic clusters where Zdbf2 was the most highly expressed were both glutamatergic (Glu13 and Glu15) and GABAergic (GABA17) neurons from the arcuate and the periventricular hypothalamic regions (Figure 1—figure supplement 1C). Same results were recently reported using different scRNA-seq datasets (Higgs et al., 2021). Interestingly, these nuclei synthesize peptides that stimulate hormone production from the pituitary gland, or that control energy balance–food intake, energy expenditure, and body temperature–by directly acting on the brain and/or more distal organs (Saper and Lowell, 2014).

Zdbf2 being also expressed in the pituitary gland (Figure 1—figure supplement 1A), we hypothesized it could have a role in the hypothalamo-pituitary axis. When we examined the pituitary gland by X-gal staining, we found Zdbf2 to be expressed in the anterior and intermediate lobes of the gland and almost undetectable signal in the posterior lobe (Figure 1D and Figure 1—figure supplement 1E), a pattern that was confirmed from available scRNA-seq data (Figure 1—figure supplement 1F; Cheung et al., 2018). The anterior and intermediate lobes that form the adenohypophysis are responsible for hormone production (Mollard et al., 2012). The anterior lobe secretes hormones from five different specialized hormone-producing cells under the control of hypothalamic inputs (growth hormone-GH, adrenocorticotropic hormone-ACTH, thyroid stimulating hormone-TSH, luteinizing hormone-LH and prolactin-PRL) and the intermediate lobe contains melanotrope cells (Kelberman et al., 2009). Altogether, the expression specificity of Zdbf2 suggests a role in functions of the endocrine hypothalamo-pituitary axis and/or of the hypothalamus alone. Finally, we further found that steady-state levels of Zdbf2 transcripts progressively rose in the hypothalamus and the pituitary gland after birth, reached their maximum at 2–3 weeks and then stabilized at later ages, in both males and females (Figure 1E and F). The expression of Zdbf2 therefore mostly increases in juvenile pups prior to weaning.

ZDBF2 positively regulates growth postnatally

In Liz-LOF mutants, Zdbf2 failed to be activated and animals displayed postnatal body weight reduction (Greenberg et al., 2017). However, whether this was directly and only linked to Zdbf2 deficiency was not resolved. To directly probe the biological role of ZDBF2, we therefore generated a mouse model of a genetic loss-of function of Zdbf2. More specifically, we engineered a ~ 700 bp deletion of the entirety of exon 6 (Figure 2A) that is common to all annotated Zdbf2 transcripts (Duffié et al., 2014). The Zdbf2-∆exon6 deletion induces a frame-shift predicted to generate a severely truncated protein (wild-type 2494aa versus 19aa mutant protein) that notably lacks the zinc finger motif. As Zdbf2 is an imprinted gene with paternal-specific expression, the deletion should exhibit an effect upon paternal but not maternal transmission. For simplicity, heterozygous mutants with a paternally inherited Zdbf2-∆exon6 deletion are thus referred to as Zdbf2-KO thereafter.

Figure 2 with 2 supplements see all
Zdbf2-KO mice exhibit growth reduction and partial postnatal lethality.

(A) Graphical model of the Zdbf2 deletion generated using two sgRNAs across exon 6. The two differentially methylated (DMR) regions of the locus are indicated (germline-gDMR, and somatic-sDMR), as well as the Long Isoform of Zdbf2 (Liz). The ORF (open-reading frame) of Zdbf2 starts in exon 4. Genomic coordinates of the deletion are indicated. (B) Body weight of Zdbf2-KO mice normalized to WT littermates (100%) followed from embryonic day E18.5–84 days post-partum. Data are shown as means ± s.e.m. from individuals from n = 27 litters. Statistical analyses were performed by a two-tailed, unpaired, non-parametric Mann Whitney t-test. **** p ≤ 0.0001, ***p ≤ 0.001,**p ≤ 0.01, *p ≤ 0.05. (C, D) Growth curve comparing the body weights of WT and Zdbf2-KO mice prior to weaning, from E18.5 to 19dpp (C) and over 3 months after birth (D). n = 15–50 mice were analyzed per genotype, depending on age and sex. Statistical analyses were performed by a two-way ANOVA test. **** p ≤ 0.0001. (E) Half dot plot- half violin plot showing the weight distribution in 2-week-old males (left) and females (right) of WT and Zdbf2-KO genotypes. Statistical analyses were performed by a two-tailed, unpaired, nonparametric Mann Whitney t test. ***p ≤ 0.001, **p ≤ 0.01. (F) Representative photography of a smaller 2 week-old Zdbf2-KO male compared to a WT littermate. (G–I) Dual-energy X-ray absorptiometry (DXA) analysis showing the calculation of body mass (G), fat mass (H) and lean mass (I) in WT and Zdbf2-KO males at 7 weeks. Data are shown as means ± s.e.m. from n = 8 WT and n = 7 Zdbf2-KO males. Statistical analyses were performed by a two-tailed, unpaired, nonparametric Mann Whitney t test. **p ≤ 0.01. (J–K) Kaplan-Meier curves of the survivability from birth to 12 weeks of age comparing WT (plain lines) and Zdbf2-KO (dotted lines) littermates from WT x Zdbf2 KO/WT backcrosses (J). Impaired survivability occurs specifically from the first day to 2 weeks of age. Kaplan-Meier curves of the survivability from 1 to 21 dpp comparing Zdbf2-KO pups generated from WT x Zdbf2 KO/KO intercrosses, with Zdbf2-KO and their WT littermates generated from WT x Zdbf2 KO/WT backcrosses as in J (K). Zdbf2-KO pups are more prone to die only when they are in competition with WT littermates (small dotted lines), while Zdbf2-KO pups have a normal survivability when they are not with WT littermates (large dotted lines). Statistical analyses were performed by a two-tailed, Chi2 test on the last time point for each curves (12 weeks for (J) and 21 days for (K)). ***p ≤ 0.001, *p ≤ 0.05.

Figure 2—source data 1

Survivability counts from different transmission of the Zdbf2-KO allele.

(A) Number of living pups followed every two days from 1 to 21dpp and every week from 21 to 84dpp from 24 litters coming from [female WT x male Zdbf2 KO/WT] crosses (Zdbf2-KO paternal transmission). These 24 litters come from four different crosses including 8 females and 4 males. Day one corresponds to the day of birth (1dpp). We can observe that the number of living Zdbf2-KO (males and females) starts decreasing from 3dpp onwards. (B) Number of living pups followed every two days from 1 to 21dpp and every week from 21 to 84dpp from 21 litters coming from [female WT x male Liz-LOF] crosses (Liz-LOF paternal transmission). These 21 litters come from five different crosses including 10 females and 5 males. Day one corresponds to the day of birth (1dpp). We can observe that the number of living Liz-LOF (males and females) starts decreasing from 3dpp onwards. (C) Number of living pups followed every week from 1 to 84dpp from 11 litters coming from [female Zdbf2 KO/WT x male WT] crosses (Zdbf2-Δexon6, maternal transmission, silent mutation). These 11 litters come from two different crosses including 4 females and 2 males. We did not observe any survivability bias within those litters. (D) Number of living pups followed every two days from 1 to 21dpp and every week from 21 to 84dpp from 11 litters coming from [female LizΔgDMR x male WT] crosses (LizΔgDMR, maternal transmission). These 11 litters come from two different crosses including 4 females and 2 males. Day one corresponds to the day of birth (1dpp). We did not observe any survivability bias within those litters. (E) Number of living pups followed every day from 1 to 21dpp from 19 litters coming from [female WT x male Zdbf2 KO/KO] crosses (Zdbf2-KO, homozygote transmission). These 19 litters come from five different crosses including 10 females and 5 males. All pups from these crosses inherited a mutated allele from their dad and are thus Zdbf2-KO. We did not observe any survivability bias within those litters.

https://cdn.elifesciences.org/articles/65641/elife-65641-fig2-data1-v2.xlsx

At birth, we found that Zdbf2-KO animals were present at expected sex and Mendelian ratios (Figure 2—figure supplement 1A-B). However, while there was no difference in body mass prior to birth (E18.5), Zdbf2-KO neonates of both sexes appeared smaller than WT littermates already at 1 and 5 days of postnatal life (day post-partum, dpp) (Figure 2B–C), indicating slower growth in the first days after birth. At 15dpp, Zdbf2-KO mice were 20% lighter than their WT littermates (Figure 2B and E–F). Lower body mass persisted into adulthood, measured up to 12 weeks of age (Figure 2D). As predicted, when present on the maternal allele, the Zdbf2-∆exon six deletion had no discernable growth effect (Figure 2—figure supplement 1C-D).

The reduced body weight phenotype was highly penetrant (Figure 2—figure supplement 1E) and was not obviously due to a developmental delay: Zdbf2-KO pups and their WT littermates synchronously acquired typical hallmarks of postnatal development (skin pigmentation, hair appearance and eye opening) (Figure 2—figure supplement 1F-I). The growth phenotype appeared to be systemic: it affected both the body length and mass of Zdbf2-KO animals (Figure 2—figure supplement 2A) and a range of organs uniformly (Figure 2—figure supplement 2B). To gain insight into the origin of the body mass restriction, we performed dual-energy X-ray absorptiometry (DEXA) scan at 7 weeks of age to measure in vivo the volume fraction of the three dominant contributors to body composition: adipose, lean, and skeletal tissues (Chen et al., 2012). While we confirmed that adult Zdbf2-KO males exhibit an overall body mass reduction (Figure 2G), we did not observe difference in the composition of fat and bone tissues (Figure 2H and Figure 2—figure supplement 2C-E). However, Zdbf2-KO mice had decreased lean mass (Figure 2I). Collectively, the above analyses demonstrate that the product of Zdbf2 positively regulates body mass gain in juvenile mice.

In conclusion, Zdbf2-KO animals exhibit the same growth defect that we previously reported in Liz-LOF mice, with identical postnatal onset and severity (Greenberg et al., 2017). However, contrary to the Liz-LOF mice, the regulatory landscape of the paternal allele of Zdbf2 was intact in Zdbf2-KO mice: DNA methylation of the sDMR located upstream of Zdbf2 was normal, and accordingly, Zdbf2 transcription was activated (Figure 2—figure supplement 2F-G). This supports that Liz transcription in the early embryo—which is required for sDMR DNA methylation (Greenberg et al., 2017)—is not impacted in the Zdbf2-KO model. We therefore show that a genetic mutation of Zdbf2 (Zdbf2-KO) and a failure to epigenetically program Zdbf2 expression (Liz-LOF) are phenotypically indistinguishable. As Liz-LOF mice show a complete lack of Zdbf2 expression (Greenberg et al., 2017), this incidentally supports that Zdbf2-KO mice carry a null Zdbf2 allele. Unfortunately, we failed to specifically detect the mouse ZDBF2 protein with commercial or custom-made antibodies.

Zdbf2-KO neonates are small or die prematurely within the first week of age

Growth restriction during the nursing period could dramatically impact the viability of Zdbf2-KO pups. Indeed, analysis of Zdbf2-KO cohorts revealed that although there was no Mendelian bias at 1dpp (Figure 2—figure supplement 1B), a strong bias was apparent at 20dpp: 75 WT and 44 Zdbf2-KO animals were weaned among n = 27 litters, while a 50/50 ratio was expected. Postnatal viability was the most strongly impaired within the first days after birth, with only 56% of Zdbf2-KO males and 52% of Zdbf2-KO females still alive at 3dpp, compared to 84% and 88% of WT survival at this age (Figure 2J and Figure 2—source data 1A). Importantly, a partial postnatal lethality phenotype was also present in Liz-LOF mutants who are equally growth-restricted as a result of Zdbf2 deficiency (Figure 2—figure supplement 1H and Figure 2—source data 1B), but did not occur when the mutation was transmitted from the silent maternal allele harboring either the Zdbf2 deletion (Figure 2—figure supplement 1I and Figure 2—source data 1C) or the Liz deletion (Figure 2—figure supplement 1J and Figure 2—source data 1D).

To determine whether partial lethality was due to a vital function of ZDBF2, per se, or to a competition with WT littermates, we monitored postnatal viability in litters of only Zdbf2-KO pups generated from crosses between WT females x homozygous Zdbf2-KO/KO males. Litter sizes were similar than the ones sired by Zdbf2-KO/WT males. However, when placed in an environment without WT littermates, Zdbf2-KO pups gained normal viability (Figure 2K and Figure 2—source data 1E). Perinatal mortality was thus rescued when there was no competition with WT littermates. Incidentally, this shows that Zdbf2-KO pups are mechanically able to ingest milk. Overall, our detailed study of the Zdbf2-KO phenotype reveals that smaller Zdbf2-KO neonates have reduced fitness compared to their WT littermates. In absence of ZDBF2, after a week of life, half of juvenile pups did not survive and the other half failed to thrive.

Postnatal body weight control is highly sensitive to Zdbf2 dosage

The physiological effects of imprinted genes are intrinsically sensitive to both decreases and increases in the expression of imprinted genes. We therefore went on assessing how postnatal body weight gain responded to varying doses of the imprinted Zdbf2 gene. We first analyzed the growth phenotype resulting from partial reduction in Zdbf2 dosage, as compared to a total loss of Zdbf2 in Zdbf2-KO mice. For this, we used the LacZ reporter Zdbf2 gene-trap mouse line (Greenberg et al., 2017), in which the insertion of the LacZ cassette downstream of exon five does not fully abrogate the production of full-length Zdbf2 transcripts (Figure 3—figure supplement 1A). Upon paternal inheritance of this allele, animals still express 50% of WT full-length Zdbf2 mRNA levels in the hypothalamus and the pituitary gland (Figure 3—figure supplement 1B-C). Interestingly, these mice displayed significant weight reduction at 7 and 14dpp compared to their WT littermates (Figure 3—figure supplement 1D), but less severe than Zdbf2-KO mice (Figure 2B–D). At 14dpp, we observed an 8% body mass reduction in males, which is incidentally half the growth reduction observed in Zdbf2-KO mice for the same age range.

We then assessed the consequences of increased Zdbf2 dosage with the hypothesis that this would lead to excessive body mass gain after birth as opposed to reduced Zdbf2 dosage. For this, we took advantage of a Zdbf2 gain-of-function (GOF) line that we serendipitously obtained when generating Liz mutant mice (Greenberg et al., 2017). The Zdbf2-GOF lines carry smaller deletions than what was initially aimed for (924 bp and 768 bp instead of 1.7 kb), resulting from non-homologous end joining (NHEJ) repair of the cut induced by the left sgRNA only (Figure 3—figure supplement 1E-F). Both deletions removed only 5’ portions of the gDMR and Liz exon 1, leaving some part of exon 1 intact, and induces an epigenetic ‘paternalization’ of the maternal allele of the locus. While sDMR methylation exclusively occurs on the paternal allele in WT embryos, animals that maternally inherit this partial deletion also acquired sDMR methylation on the maternal allele (Figure 3A and Figure 3—figure supplement 1F-H). Indeed, important, but not complete, sDMR DNA methylation was observed (65% on the maternal allele compared to 95% on the paternal allele, Figure 3—figure supplement 1H). This was associated with Zdbf2 activation from the maternal allele, although slightly less than from the WT paternal allele (Figure 3B–C). As a consequence, Zdbf2-GOF animals exhibit bi-allelic Zdbf2 expression, with a net 1.7-fold increase of Zdbf2 levels in postnatal hypothalamus and pituitary gland, as compared to WT littermates that express Zdbf2 mono-allelically, from the paternal allele only. In sum, we have generated mutant mice with loss-of-imprinting of the Zdbf2 locus.

Figure 3 with 2 supplements see all
Zdbf2 influences postnatal growth in a dose-dependent manner.

(A) Bisulfite cloning and sequencing showing CpG methylation levels at the sDMR locus of hypothalamus DNA from 6-week-old hybrid WT (left) and Zdbf2-GOF (right) mice (Zdbf2-GOF±x JF1 cross). Red, maternal alleles; blue, paternal alleles. cross, informative JF1 SNP. (B) Allelic expression of Zdbf2 in hypothalamus and pituitary gland from 3-week-old mice, measured by RT-pyrosequencing. Genomic DNA extracted from a C57Bl/6 x JF1 hybrid cross was used as a control for pyrosequencing bias. (C) RT-qPCR measurement reveals a ~ 1.7-fold-increase of Zdbf2 expression in the hypothalamus and pituitary gland of 3-week-old mice with a maternal transmission of the deletion. Expression of Zdbf2 in mice carrying the deletion on the paternal allele is similar to WT. Statistical analyses were performed by a one-way ANOVA test. ***p ≤ 0.001. (D) Normalized body growth of Zdbf2-GOF mice to their WT littermates (100%) followed at different ages (1–84 days) from n = 14 litters. The overgrowth is seen specifically in males, from 5 to 28 days. Statistical analyses were performed by a two-tailed, unpaired, non-parametric Mann Whitney t-test. **p ≤ 0.01, *p ≤ 0.05. The number on top of the data at 5dpp indicate a non-significant but close to be p-value. (E, F) Growth curves of female and male mice, comparing the body weights of WT and Zdbf2-GOF, through the three first weeks of life (D) and through 10 weeks (E). n = 10–30 mice were analyzed per genotype, depending on age and sex. Statistical analyses were performed by a two-way ANOVA test. **** p ≤ 0.0001, **p ≤ 0.01. (G) Half dot- half violin plots showing the weight distribution at 2 weeks of age between WT and Zdbf2-GOF males. Data are shown as means ± s.e.m. from n individuals. Statistical analyses were performed by a two-tailed, unpaired, nonparametric Mann Whitney t test. *** p ≤ 0.005. (H) Representative photography of a bigger Zdbf2-GOF male as compared to a WT littermate at 2 weeks.

Importantly, we found that increased Zdbf2 dosage was impactful for postnatal body mass, specifically in males (Figure 3D–H). From 10 to 21dpp, Zdbf2-GOF juvenile males were 10% heavier when normalized to the average WT siblings from the same litter (Figure 3D) but their viability was similar to that of WT controls (Figure 3—figure supplement 2A-C). Similar to growth restriction observed in Zdbf2-KO mutant mice, males Zdbf2-GOF remained heavier than WT littermates into adult life, as measured until 10 weeks of age (Figure 3F), and a range of organs were uniformly affected (Figure 3—figure supplement 2D). Zdbf2-GOF males were significantly overweighed at 15dpp (Figure 3G–H) and the body mass distribution of males Zdbf2-GOF showed most of them were larger than WT controls (Figure 3—figure supplement 2E). Although Zdbf2 was overexpressed in both Zdbf2-GOF males and females, the male-specific overgrowth phenotype may imply stimulation of the phenotype by sex hormones, directly or indirectly. Comparatively, progenies from a paternal transmission of the deletion showed a growth progression similar to their WT littermates (Figure 3—figure supplement 2F-G). Overall, these data illustrate the importance of Zdbf2 for the regulation of postnatal body weight gain at the onset of postnatal life. We therefore demonstrate that the Zdbf2 imprinted gene encodes a genuine positive regulator of postnatal body weight gain in mice, with highly attuned dose-sensitive effects.

Zdbf2 regulates postnatal growth in a parent-of-origin independent manner

Imprinted genes with a paternal expression generally tend to enhance prenatal and postnatal growth (Haig, 2000). Consistently, we found that the paternally expressed Zdbf2 gene promotes postnatal weight gain. Having determined the dosage effect of Zdbf2, we next wondered what role plays the paternal origin of Zdbf2 expression on postnatal body weight. To invert Zdbf2 parental expression, we intercrossed the Liz-LOF and Zdbf2-GOF lines, which consist in a maternalization of the paternal allele and a paternalization of the maternal allele of Zdbf2, respectively. By crossing an heterozygous Zdbf2-GOF female with an heterozygous Liz-LOF male (Figure 4A, left panel), four genotypes can segregate in the progeny, with littermates displaying various dosage and parent-of-origin expression of Zdbf2 (Figure 4A): (1) wild-type animals with one dose of paternal Zdbf2 expression, (2) animals with a paternal Liz-LOF allele and lack of Zdbf2 expression (equivalent to a Zdbf2-loss-of-function, Zdbf2-LOF), (3) animals with a maternal Zdbf2-GOF allele and biallelic Zdbf2 expression, and finally, (4) animals with combined maternal Zdbf2-GOF and paternal Zdbf2-LOF alleles and potentially, mono-allelic, reverted maternal expression of Zdbf2 (Figure 4A, right panel).

Zdbf2 influences postnatal growth in a parent-of-origin-independent manner.

(A) Scheme of the cross made to obtain embryos with an inversion of the parental origin of Zdbf2 expression. Zdbf2-GOF heterozygote females were crossed with heterozygote males for the Liz-LOF deletion–which we demonstrated as equivalent to a Zdbf2-LOF allele–(left) to obtain one quarter of embryos expressing one dose of Zdbf2 from the maternal allele (right, bottom). (B) Zdbf2 expression in the hypothalamus and the pituitary gland is shown in males for each of the four possible genotypes. The level of Zdbf2 in Zdbf2-GOF /Zdbf2-LOF mice almost completely rescues the defect seen in Zdbf2-LOF and Zdbf2-GOF mutants. Data are shown as means ± s.e.m. from n individuals. Statistical analyses were performed by a two-tailed, unpaired, nonparametric Mann Whitney t test. * p ≤ 0.05. (C) Normalized body growth of Zdbf2-LOF, Zdbf2-GOF and Zdbf2-GOF /Zdbf2-LOF males to their WT littermates (100%) followed at 7, 14, and 21 days after birth. Zdbf2-GOF /Zdbf2-LOF adult mice exhibit a body weight similar to the WT showing a partial rescue of the growth reduction and overgrowth phenotype due to respectively the lack of Zdbf2 and the gain of Zdbf2 expression in the brain. Data are shown as means ± s.e.m. from individuals from n = 17 litters. Statistical analyses were performed by a two-tailed, unpaired, nonparametric Mann Whitney t test. * p ≤ 0.05; ** p ≤ 0.01. (D) Representative photography of four males littermates from a Zdbf2-GOF x Zdbf2-LOF cross (as shown in A) at 2 weeks of age. For each animal, genotype, dose of Zdbf2 expression and weight are indicated.

The uniform strain background of the two lines did not allow us to use strain-specific sequence polymorphisms to distinguish the parental origin of Zdbf2 regulation in the compound Zdbf2-GOF/Zdbf2-LOF animals. However, we found that compared to single Zdbf2-LOF mutants, the presence of the maternal Zdbf2-GOF allele restored DNA methylation levels at the sDMR locus in all tissues of Zdbf2-GOF/Zdbf2-LOF mutants, with an average of 43.5% CpG methylation compared to the expected 50% in WT (Duffié et al., 2014; Figure 3—figure supplement 2H). By RT-qPCR, Zdbf2 mRNA levels were also increased in the hypothalamus and pituitary gland of Zdbf2-GOF/Zdbf2-LOF animals compared to single Zdbf2-LOF animals (Figure 4B), which strongly suggests that expression comes from the maternal Zdbf2-GOF allele. Accordingly, Zdbf2 expression level in Zdbf2-GOF/Zdbf2-LOF animals was on average 0.65-fold the one of WT animals, which is congruent with the partial paternalization of the maternal Zdbf2-GOF allele we reported (Figure 4B and Figure 3—figure supplement 2H). Most importantly, restoration of Zdbf2 expression by the maternal Zdbf2-GOF allele–even though incomplete–was sufficient to rescue the postnatal body weight phenotype in compound Zdbf2-GOF/Zdbf2-LOF males compared to their single Zdbf2-LOF brothers (Figure 4C and D). From a body weight reduction of 20% reported in Liz-LOF or Zdbf2-KO animals at the same age, it was attenuated to only 4% in Zdbf2-GOF/Zdbf2-LOF animals (Figure 4C), showing that maternal Zdbf2 expression is as functional as paternal Zdbf2 expression. Altogether, these results crystallize the importance of Zdbf2 dosage in regulating postnatal body weight and most importantly, demonstrate that the dose but not the parental origin matters for Zdbf2 function.

Zdbf2-KO growth phenotype is correlated with decreased IGF-1 in the context of normal development of the hypothalamo-pituitary axis

Having demonstrated the growth promoting effect of Zdbf2, we next tackle the question of how it influences newborns weight and survival. As Zdbf2 is expressed in the neuroendocrine cells of the hypothalamo-pituitary axis, the phenotype of Zdbf2-KO animals may lay in a defect in producing growth-stimulating pituitary hormones. When we assessed the development and functionality of the hypothalamus and pituitary gland prior to birth, we did not detect any morphological nor histological defects in Zdbf2-KO embryos (Figure 5—figure supplement 1A-C). Normal expression of major transcriptional regulators confirmed proper cell lineage differentiation in the Zdbf2-KO developing pituitary (Figure 5—figure supplement 1B; Raetzman et al., 2002; Rizzoti, 2015; Kelberman et al., 2009). Immunohistochemistry at E18.5 further indicated that Zdbf2-KO pituitary cells acquire normal competency for producing hormones (Figure 5—figure supplement 1D). Similarly, hypothalamic peptides were expressed in comparable levels in Zdbf2-KO and WT embryos, as assessed by in situ hybridization analysis (Figure 5—figure supplement 1E-F; Biran et al., 2015). In sum, the embryonic hypothalamo-pituitary axis develops normally in the absence of Zdbf2. Our data implies that the postnatal growth phenotype does not result from impaired establishment or programming of this axis during embryogenesis.

We then went on to analyze the functionality of the hypothalamo-pituitary axis in producing hormones after birth. Again, GH, ACTH, TSH, LH and PRL all appeared to be normally expressed in the pituitary glands of Zdbf2-KO juvenile animals (immunohistochemistry at 15dpp) (Figure 5A). Although we cannot exclude subtle dysfunctionalities, our results suggest that Zdbf2 deficiency does not drastically compromise pituitary hormone production. Measured plasma GH levels in Zdbf2-KO animals also showed functional hormone release at 5 and 15dpp (Figure 5B).

Figure 5 with 2 supplements see all
Zdbf2-KO phenotype is linked to defective IGF-1 signaling immediately after birth.

(A) Pituitary hormone production is globally normal in Zdbf2-KO mice, as assessed by immunohistochemistry on 15dpp pituitary sections. (B, C) Circulating levels of GH at 5 and 15dpp (B) and of IGF-1 at 1, 5, and 15 dpp (C) in the plasma of WT and Zdbf2-KO mice. Data are shown as means ± s.e.m. of the relative expression to WT values from n replicates. Red and blue dots: females and males data points, respectively. Statistical analyses were performed by a two-tailed, unpaired, nonparametric Mann Whitney t test. *p ≤ 0.05. (D) RT-qPCR from postnatal liver measuring the level of Igf-1 mRNAs between WT and Zdbf2-KO mice at 7 and 21dpp. Data are shown as means ± s.e.m from n WT and Zdbf2-KO animals. (E) Specific growth rate calculated from body weight of WT and Zdbf2-KO males from 1 to 10 weeks of age using the following equation: [(weight t2 - weight t1)/ weight t1]. Pink area: reduced growth rate in Zdbf2-KO pups compared to WT littermates; Grey area: growth rate of Zdbf2-KO mice exceeds the WT growth rate.

However, Zdbf2-KO pups showed reduced plasma circulating levels of insulin growth factor 1 (IGF-1), a main regulator of postnatal growth, which reached only 70%, 45%, and 30% of WT level at 1dpp, 5dpp and 15dpp, respectively (Figure 5C). In adult mice, the liver is the main site of production of IGF-1, following transcriptional activation under the control of circulating GH (Savage, 2013). In contrast, extrahepatic production of IGF-1 during embryonic and early postnatal life is mostly GH-insensitive (Lupu et al., 2001; Kaplan and Cohen, 2007). As mentioned above, decreased IGF-1 levels in juvenile Zdbf2-KO animals seemed to occur in the context of normal GH input. Moreover, we measured normal Igf1 mRNA levels by RT-qPCR in the liver of juvenile Zdbf2-KO animals, further illustrating that decreased IGF-1 levels are not a result of altered GH pathway (Figure 5D).

The association of low levels of circulating IGF-1 with normal GH secretion prompted us to evaluate more thoroughly the Zdbf2-KO growth phenotype. Mouse models of GH deficiency show growth retardation only from 10dpp onwards, while deficiency in IGF-1 affects growth earlier during postnatal development (Lupu et al., 2001). When we calculated the growth rate from day 1 to 8 weeks of age, we revealed two distinct phases: (1) from 1 to 7dpp, the Zdbf2 mutant growth rate was 22% lower compared to WT littermates and (2) from 15 to 35dpp, the mutant exceeded the WT growth rate (Figure 5E) while no differences were observed in the Zdbf2-∆exon six silent mutation (Figure 5—figure supplement 2A). This shows that the growth defect is restricted to the first days of life. More specifically, at the day of birth (1dpp), Zdbf2-KO pups were smaller than their WT littermates by 9% (WT 1.41 ± 0.015 g, n = 99; Zdbf2-KO 1.29 ± 0.013, n = 95). After 1 week of postnatal life, the mutant growth restriction reached 18% (WT 3.4 ± 0.13 g, n = 37; Zdbf2-KO 2.8 ± 0.11, n = 17), indicating a deficit in the ability to gain weight prior to 10dpp. Then, at 6 weeks of age, Zdbf2-KO animals were smaller than WT by only 8%, as a result of enhanced post-pubertal growth spurt (WT 21.7 ± 0.04 g, n = 34; Zdbf2-KO 20.05 ± 0.3, n = 15). We therefore concluded that the Zdbf2-KO phenotype is similar to defective IGF-1 signaling immediately after birth. However, circulating IGF-1 levels were not conversely increased in Zdbf2-GOF males at 5 and 15dpp (Figure 5—figure supplement 2B): IGF-1-independent mechanisms may be responsible for the larger body mass, or IGF-1 may be only transiently upregulated and not detected at these two timepoints.

Overall, we revealed that the body weight restriction of Zdbf2-KO juveniles is associated with a GH-independent decrease of IGF-1 during the first days of postnatal life, in the context of an overall normal development and functionality of the hypothalamo-pituitary axis.

Zdbf2-KO neonates are undernourished and do not properly activate hypothalamic feeding circuits

Having determined that Zdbf2-KO animals are deficient in IGF-1, we attempted to define the molecular events associated with reduced IGF-1 levels. Undernutrition is a well-known cause of IGF-1 level reduction, and also of postnatal lethality (Thissen et al., 1994), which we observed in Zdbf2-KO pups. To investigate the nutritional status of Zdbf2-KO neonates, we weighed stomachs at 3dpp, as a measure of milk intake. Zdbf2-KO pups exhibited a significant reduction in stomach weight relative to body mass as compared to their WT littermates (Figure 6A), suggesting that these pups suffer milk deprivation. Stomach mass of Zdbf2-GOF males was not affected at 3dpp (Figure 6—figure supplement 1A), as expected from the normal body mass at the same age (Figure 3D). Zdbf2-KO pups showed normal relative mass at 3dpp for a range of organs at the exception of the interscapular brown adipose tissue (BAT), which was also reduced in Zdbf2-KO neonates (Figure 6B and Figure 6—figure supplement 1B). BAT-mediated thermogenesis regulates body heat during the first days after birth (Cannon and Nedergaard, 2004) and improper BAT function can lead to early postnatal death (Charalambous et al., 2012). However, despite being smaller, the BAT of Zdbf2-KO neonates appeared otherwise functional, showing normal lipid droplet enrichment on histological sections (data not shown) and proper expression of major markers of BAT thermogenic ability (Figure 6—figure supplement 1C). The BAT size reduction may therefore not reflect altered BAT ontogeny per se, but rather the nutritional deprivation of Zdbf2-KO neonates.

Figure 6 with 1 supplement see all
Zdbf2-KO neonates display a feeding defect.

(A, B) Stomach (A) and brown adipocyte tissue (BAT) (B) mass normalized to the body mass for WT and Zdbf2-KO at 3dpp. Red dots: females; blue dots: males. Data are shown as means ± s.e.m. from n replicates. Statistical analyses were performed by a two-tailed, unpaired, nonparametric Mann Whitney t test.** p ≤ 0.01, ***p ≤ 0.005. (C) Volcano plot representation of RNA-seq of 3dpp hypothalamus of Zdbf2-KO versus WT littermates. n = 3 replicates for each genotype. Red dots: differentially expressed genes with a threshold of FDR < 10%. Npy (FDR 6%) is highlighted in green and Agrp in orange. (D) Representative image of immunofluorescence from brain sections, focused on hypothalamic region in Zdbf2:LacZ animals at 15dpp. Black and white images are shown for DAPI, NPY and Beta-galactosidase and composite images depict them in blue, red and green, respectively. Dotted square (top panel) represent the focused region in the bottom panel. Scale bar: 100 µm. 3 V, third ventricule; Pa, paraventricular hypothalamic nucleus. (E) Heatmap showing the log2 fold change of genes encoding hypothalamic regulators of food intake (RNA-seq data from C). (F) RT-qPCR from hypothalamus of 1 and 3dpp males animals measuring Npy (left panel) and Agrp (right panel) mRNA levels. Data are shown as means ± s.e.m. from n replicates. Statistical analyses were performed by a two-tailed, unpaired, nonparametric Mann Whitney t test. *p ≤ 0.05, **p ≤ 0.01.

Figure 6—source data 1

List of differentially expressed genes in the hypothalamus of Zdbf2-KO versus WT males at 3dpp and 10dpp.

Raw data for the RNA-seq on hypothalamus at 3dpp and 10dpp. Tables are showing the rpkm value, FDR and logFC for the 11 differentially expressed genes between WT and Zdbf2-KO.

https://cdn.elifesciences.org/articles/65641/elife-65641-fig6-data1-v2.xlsx

Because the mothers of Zdbf2-KO pups are of WT background, the undernutrition phenotype is unlikely due to defective maternal milk supply. Restoration of viability when placed in presence of KO-only littermates (Figure 2K) further indicated that Zdbf2-KO pups are competent to feed. Defective nutrition may therefore rather result from altered feeding motivation of the Zdbf2-KO neonates, specifically during the early nursing period. To test this hypothesis, we performed RNA-seq analysis of dissected hypothalami at 3dpp, when the growth phenotype is the most acute. Only 11 genes were significantly misexpressed in Zdbf2-KO hypothalamus relative to WT littermates (FDR 10%) (Figure 6C and Figure 6—source data 1). Only one of these genes is related to feeding, Npy, encoding the Neuropeptide Y, appearing as down-regulated in Zdbf2-KO hypothalamus. In line with our hypothesis, NPY is secreted from neurons of the hypothalamic arcuate nucleus and stimulates food intake and promotes gain weight (Mercer et al., 2011). Using our Zdbf2-LacZ reporter line, we revealed co-localization of beta-galactosidase and NPY staining in the hypothalamus (Figure 6D). This observation reinforces the possible involvement of ZDBF2 in NPY production in the paraventricular nucleus, to control food intake.

Little is known about the determinants of food intake in neonates (Muscatelli and Bouret, 2018); nonetheless, this prompted us to examine the expression levels of other hypothalamic modulators known to influence feeding behavior in adults. Interestingly, genes that positively regulate food intake tended to be down-regulated, while negative regulators of food intake were up-regulated (Figure 6E). One of these genes is the Agouti-related peptide (Agrp)-encoding gene (Figure 6E) that is co-expressed with Npy in a sub-population of hypothalamic neurons and convergently stimulate appetite and food seeking (Gropp et al., 2005). RT-qPCR measurement confirmed lower levels of both Npy and Agrp in Zdbf2-KO hypothalamus compared to WT, at 3dpp but also earlier on, at 1dpp (Figure 6F). It is noteworthy that none of the misregulated genes, including positive regulators of feeding, that we observed at 3dpp were also misregulated in the hypothalamus at 10dpp, as measured by RNA-seq (Figure 6—figure supplement 1D-E).

Together, these results provide key insights into the origin of the Zdbf2 mutant phenotype: in absence of Zdbf2, the hypothalamic circuit of genes that stimulates food intake may not be properly activated at birth. This is associated with reduced milk intake, reduced body weight gain and suboptimal viability of Zdbf2-KO neonates when placed in presence of healthier littermates.

Discussion

The hypothalamus is increasingly regarded as a major site for the action of imprinted genes on postnatal growth, feeding behavior and metabolism (Ivanova and Kelsey, 2011). Here, we found evidence that the imprinted Zdbf2 gene stimulates hypothalamic feeding circuits, right after birth. In its absence, neonates do not ingest enough milk, and suffer from undernutrition. This leads to decreased IGF-1 signaling within the first week of life, reduced body weight gain and more dramatically, lethality of half of the Zdbf2-KO pups when they are in presence of healthier WT littermates. Our work therefore highlights that Zdbf2 is necessary to thrive and survive after birth, by allowing newborns to adapt to postnatal feeding. Interestingly, Zdbf2 expression is persistently high in adult brains, which may point to other functions in later life.

By relying on a unique collection of mouse models of total loss of function (Zdbf2-KO and Liz-LOF), partial loss of function (Zdbf2-lacZ), normal function (Zdbf2-WT), and gain of function (Zdbf2-GOF), we observed that postnatal body growth is exquisitely sensitive to the quantity of Zdbf2 produced in the hypothalamo-pituitary axis. Incidentally, Zdbf2 meets the criteria of a bona fide growth-promoting gene: growth is reduced upon decreased dosage and oppositely enhanced upon increased dosage of Zdbf2 (Efstratiadis, 1998). Consistent with previous results for the imprinted Cdkn1c gene (Andrews et al., 2007), growth reduction was more pronounced than overgrowth upon changes in Zdbf2 expression (18% decrease and 10% increase at 10dpp) and specific to males. This reflects that, with standard diet, overgrowth is a much less frequent response than growth reduction and affects animals with favorable physiology only, potentially explaining male specificity.

With a few exceptions, most studies have only addressed the phenotypic effects of reducing the dose of an imprinted gene, but not what results from overexpressing this same gene (Tucci et al., 2019). Knocking out an imprinted gene provides valuable insights about its physiological function; however, it does not address the evolutionary significance of imprinting of that gene, that is reduction to mono-allelic expression. With the Zdbf2-GOF model, we were able to evaluate the consequences of a loss of Zdbf2 imprinting: bi-allelic and increased Zdbf2 expression exacerbates postnatal body weight gain. However, despite being slightly bigger, viability, fertility or longevity appeared overall normal in Zdbf2-GOF animals. The evolutionary importance of Zdbf2 imprinting is therefore not immediately obvious, at least under unchallenged conditions. Finally, even fewer studies have addressed the importance of parent-of-origin expression for imprinted gene functions (Drake et al., 2009; Leighton et al., 1995). By intercrossing models of Zdbf2 loss and gain of function, we could enforce Zdbf2 expression from the maternal allele while inactivating the normally expressed paternal allele. Restoration of body weight demonstrated the functionality of maternal Zdbf2 expression, reinforcing that Zdbf2 functions on postnatal body homeostasis in a dose-dependent manner. Although not necessarily surprising, this demonstration is conceptually important for understanding the evolution and raison d’être of genomic imprinting. The divergent DNA methylation patterns that are established in the oocyte and the spermatozoon provide opportunities to evolve mono-allelic regulation of expression, but once transmitted to the offspring, the parental origin of expression is not essential per se.

Although Zdbf2 is expressed across the hypothalamo-pituitary axis, we could not find evidence of abnormal development or function of the pituitary gland that could explain the Zdbf2-KO growth phenotype. Notably, GH production and release were normal, at least from 5dpp and on. Additionally, the Zdbf2 growth reduction diverges from GH-related dwarfism: Gh-deficient mice grow normally until 10dpp, after which only they exhibit general growth impairment with reduced levels of circulating IGF-1 (Voss and Rosenfeld, 1992; Lupu et al., 2001). In Zdbf2-KO mutants, the growth defect is apparent as soon as 1dpp, as well as decreased IGF-1. In Igf-1 null mice, birthweight is approximately 60% of normal weight and some mutants die within the first hours after birth (Liu et al., 1993; Efstratiadis, 1998). The Zdbf2-KO phenotype thus resembles an attenuated Igf-1 deficiency, in agreement with half reduction but not total lack of circulating IGF-1. Interestingly, similar IGF-1-related growth defects have been reported upon alteration of other imprinted loci in the mouse: the Rasgrf1 gene, the Dlk1-Dio3 cluster and the Cdkn1c gene (Itier et al., 1998; Andrews et al., 2007; Charalambous et al., 2014). However, the origin of IGF-1 deficiency may be different: in Rasgrf1 mutants, unlike Zdbf2 mutants, this was linked to impaired hypothalamo-pituitary axis and GH misregulation (Drake et al., 2009). Finally, decreased ZDBF2 levels have recently been associated with intra-uterine growth restriction (IUGR) in humans (Monteagudo-Sánchez et al., 2019). Whether this is linked to impaired fetal IGF-1 production or to distinct roles of ZDBF2 related to placental development in humans would be interesting to assess, in regards to conservation or not of imprinted gene function across mammals.

IGF-1 secretion has been shown to drop in response to starvation, leading to disturbed growth physiology (Savage, 2013). Our findings support that limited food intake is probably the primary defect in Zdbf2 deficiency, leading to IGF-I insufficiency in the critical period of postnatal development and consequently, growth restriction. First, we showed that Zdbf2 is expressed in hypothalamic regions that contain neurons with functions in appetite and food intake regulation, such as the arcuate and paraventricular nucleus. Second, Zdbf2-KO neonates show hypothalamic downregulation of the Npy and Agrp genes that encode for orexigenic neuropeptides (Stanley and Leibowitz, 1984; Ollmann et al., 1997). These are likely direct effects: (i) NPY and Zdbf2-LacZ are co-expressed in the same cells and (ii) the rest of the hypothalamic transcriptome is scarcely modified in Zdbf2-KO pups. Quantified changes were not of large magnitude, but AgRP/NPY neurons represent a very small population of cells, present in the arcuate nucleus of the hypothalamus only (Andermann and Lowell, 2017). We are likely at the limit of detection when analyzing these genes in the whole hypothalamus transcriptome. Finally, Npy and Agrp downregulation was observed immediately at birth, along with a phenotype of reduced milk consumption, as measured by stomach weighing. Given our observations, we propose that ZDBF2 activates specialized hypothalamic neurons that motivate neonates to actively demand food (milk) from the mother right at birth, promoting the transition to oral feeding after a period of passive food supply in utero. This function agrees with the co-adaptation theory according to which genomic imprinting evolved to coordinate interactions between the offspring and the mother (Wolf and Hager, 2006). Our results are also in line with the kindship theory of genomic imprinting (Haig, 2000): Zdbf2 is a paternally expressed gene that potentiates resource extraction from the mother. Finally, despite considerable effort, we were unable to specifically detect the ZDBF2 protein with antibodies or using epitope-tagging approaches of the endogenous gene; future studies hopefully will bring clarity to which molecular function ZDBF2 carries in the mouse hypothalamus.

In conclusion, we reveal here that decreasing Zdbf2 compromises resource acquisition and body weight gain right after birth. Restricted postnatal growth can have a strong causal effect on metabolic phenotypes, increasing the risk of developing obesity in later life. This is observed in mouse KO models of the imprinted Magel2 gene that map to the Prader Willi syndrome (PWS) region, recapitulating some features of PWS patients, who after a failure to thrive as young infants exhibit a catch-up phase leading to overweight and hyperphagia (Bischof et al., 2007). Zdbf2-KO mice do present a post-puberal spurt of growth, which attenuates their smaller body phenotype, but we never observed excessive weight gain, even after 18 months (data not shown). It would be interesting to test whether the restricted postnatal growth of Zdbf2-KO mice may nonetheless increase the likelihood of metabolic complications when challenged with high-fat or high-sugar diet.

Materials and methods

Key resources table
Reagent type (species) or resourceDesignationSource or referenceIdentifiersAdditional information
Genetic reagent (Mus. Musculus)Zdbf2-KOThis studyCRISPR/Cas9 generated mutant, sgRNA oligos are listed in Supplementary file 1
Genetic reagent (Mus. Musculus)Zdbf2-GOFBourc’his labGreenberg et al., 2017
Genetic reagent (Mus. Musculus)Zdbf2-LacZ reporter lineBourc’his labEUCOMM Project Number: Zdbf2_82543Greenberg et al., 2017
Genetic reagent (Mus. Musculus)Liz-LOFBourc’his labGreenberg et al., 2017
AntibodyAnti-ACTH, mouse monoclonalFitzgeraldRRID:AB_1282437Ref. 10C-CR1096M1, 1:1,000
AntibodyAnti-GH, rabbit polyclonalNational Hormone and Peptide Program (NHPP)Ref. AFP-5641801, 1:1,000
AntibodyAnti-TSH, rabbit polyclonalNational Hormone and Peptide Program (NHPP)Ref. AFP-1274789, 1:1,000
AntibodyAnti-PRL, rabbit polyclonalNational Hormone and Peptide Program (NHPP)Ref. AFP-425-10-91, 1:1,000
AntibodyAnti-LH, rabbit polyclonalNational Hormone and Peptide Program (NHPP)Ref. AFP-C697071P, 1:500
AntibodyAnti-NPY, rabbit polyclonalCell Signaling TechnologyRRID:AB_2716286Ref. # 11976, 1:1,000
AntibodyAnti-beta-galactosidase, chicken polyclonalAbcamRef. # ab9361, 1:1,000
AntibodyGoat anti-rabbit Alexa-fluorophore 594InvitrogenRRID:AB_2762824Ref. #A32740, 1:1,000
AntibodyGao anti-chicken Alexa-fluorophore 488InvitrogenRRID:AB_2534096Ref. # A11039, 1:1,000
Commercial assay or kitMouse Magnetic Luminex Assay for IGF-1R&D System
Commercial assay or kitMilliplex Mouse Pituitary Magnetic Assay for GHMerck
Software, algorithmSTAR_2.6.1 aDobin et al., 2013

Mice

Mice were hosted on a 12 hr/12 hr light/dark cycle with free access to food and water in the pathogen-free Animal Care Facility of the Institut Curie (agreement number: C 75-05-18). All experimentation was approved by the Institut Curie Animal Care and Use Committee and adhered to European and National Regulation for the Protection of Vertebrate Animals used for Experimental and other Scientific Purposes (Directive 86/609 and 2010/63). For tissue and embryo collection, euthanasia was performed by cervical dislocation. The Zdbf2-KO and Zdbf2-GOF mutant mice lines were derived by CRISPR/Cas9 engineering in one-cell stage embryos as previously described (Greenberg et al., 2017), using two deletion-promoting sgRNAs. Zygote injection of the CRISPR/Cas9 system was performed by the Transgenesis Platform of the Institut Curie. Eight week-old superovulated C57BL/6 J females were mated to stud males of the same background. Cytoplasmic injection of Cas9 mRNA and sgRNAs (100 and 50 ng/µl, respectively) was performed in zygotes collected in M2 medium (Sigma) at E0.5, with well-recognized pronuclei. Injected embryos were cultured in M16 medium (Sigma) at 37 °C under 5% CO2, until transfer at the one-cell stage the same day or at the two-cell stage the following day in the infudibulum of the oviduct of pseudogestant CD1 females at E0.5. The founder mice were then genotyped and two independent founders with the expected deletion were backcrossed to segregate out undesired genetic events, with a systematic breeding scheme of Zdbf2-KO heterozygous females x WT C57Bl6/J males and Zdbf2-GOF heterozygous males x WT C57Bl6/J females to promote silent passing of the deletion. Cohorts of female and male N3 animals were then mated with WT C57Bl6/J to study the maternal and paternal transmission of the mutation.

The LacZ-Zdbf2 reporter line was derived from mouse embryonic stem (ES) cells from the European Conditional Mouse Mutagenesis Program (EUCOMM Project Number: Zdbf2_82543). Proper insertion of the LacZ construct was confirmed by long-range PCR. However, we found that the loxP site in the middle position was mutated (A to G transition at position 16 of the loxP site) in the original ES cells (Figure 3—figure supplement 1A). Chimeric mice were generated through blastocyst injection by the Institut Curie Transgenesis platform. We studied animals with an intact LacZ-KI allele, without FRT- or CRE-induced deletions.

DNA methylation analyses

Request a detailed protocol

Genomic DNA from adult tissues was obtained following overnight lysis at 50 °C (100 mM Tris pH 8, 5 mM EDTA, 200 mM NaCl, 0.2% SDS and Proteinase K). DNA was recovered by a standard phenol/choloroform/isoamyl alcohol extraction and resuspended in water. Bisulfite conversion was performed on 0.5–1 µg of DNA using the EpiTect Bisulfite Kit (Qiagen). Bisulfite-treated DNA was PCR amplified and either cloned and sequenced, or analyzed by pyrosequencing. For the former, 20–30 clones were Sanger sequenced and analyzed with BiQ Analyzer software (Bock et al., 2005). Pyrosequencing was performed on the PyroMark Q24 (Qiagen) according to the manufacturer’s instructions, and results were analyzed with the associated software.

RNA expression analyses

Request a detailed protocol

Total RNA was extracted using Trizol (Life Technologies). To generate cDNA, 1 µg of Trizol- extracted total RNA was DNase-treated (Ambion), then reverse transcribed with SuperscriptIII (Life Technologies) primed with random hexamers. RT-qPCR was performed using the SYBR Green Master Mix on the ViiA7 Real-Time PCR System (Thermo Fisher Scientific). Relative expression levels were normalized to the geometric mean of the Ct for housekeeping genes Rrm2, B-actin and/or Rplp0, with the ΔΔCt method. Primers used are listed in Supplementary file 1.

For RNA-sequencing, hypothalami of three animals at 3dpp and two animals at 10dpp were collected for each genotype (all males) and RNA was extracted. Trizol-extracted total RNA was DNase-treated with the Qiagen RNase-Free DNase set, quantified using Qubit Fluorometric Quantitation (Thermo Fisher Scientific) and checked for integrity using Tapestation (Agilent). RNA-seq libraries were cloned using TruSeq Stranded mRNA LT Sample Kit on total RNA (Illumina) and sequencing was performed on a NovaSeq 6,000 (Illumina) at the NGS platform of the Institut Curie (PE100, approximately 35M to 50M clusters per replicate for 3dpp, and 16–30 M for 10dpp).

LacZ staining

Request a detailed protocol

Whole brains and pituitary glands were fixed in 4% paraformaldehyde (PFA) in PBS (pH 7.2) overnight and washed in PBS. For sections, tissues were then incubated in sucrose gradients and embedded in OCT for conservation at –80 °C before cryosectioning. Sections were first fixed 10 min in solution of Glutaraldehyde (0.02% Glutaraldehyde, 2 mM MgCl2 in PBS). Tissues and sections were washed in washing solution (2 mM MgCl2, 0.02% NP40, 0.01% C24H39NaO4 in PBS) and finally incubated at room temperature overnight in X- gal solution (5 mM K4Fe(CN)63H20, 5 mM K3Fe(CN)6, 25 mg/mL X-gal in wash solution). After several PBS washes, sections were mounted with an aqueous media before imaging. N = 3 biological replicates were tested.

Histological analysis and RNA in situ hybridization

Request a detailed protocol

Embryos at E13.5 and E15.5 were embedded in paraffin and sectioned at a thickness of 5 μm. For histological analysis, paraffin sections were stained with Hematoxylin and Eosin. RNA in situ hybridization analysis on paraffin sections was performed following a standard procedure with digoxigenin-labeled antisense riboprobes as previously described (Gaston-Massuet et al., 2008).The antisense riboprobes used in this study [α-Gsu, Pomc1, Lhx3, Pitx1, Avp, oxytocin and Ghrh] have been previously described (Gaston-Massuet et al., 2008; Gaston-Massuet et al., 2016). N = 3 biological replicates were tested.

Immunohistochemistry on histological sections

Request a detailed protocol

Embryos were fixed in 4% PFA and processed for immuno-detection as previously described (Andoniadou et al., 2013). Hormones were detected using antibodies for α-ACTH (mouse monoclonal, 10C-CR1096M1, RRID:AB_1282437, 1:1000), α-GH (rabbit polyclonal, NHPP AFP-5641801, 1:1000), α-TSH (rabbit polyclonal, NHPP AFP-1274789, 1:1000), α-PRL (rabbit polyclonal, NHPP AFP-425-10-91, 1:1000), and α-LH (rabbit polyclonal, NHPP AFP-C697071P, 1:500). N = 3 biological replicates were tested.

Immunofluorescence on hypothalamus sections

Request a detailed protocol

Cryostat sections were washed with PBS, permeabilized and blocked with Blocking buffer (5% horse serum, 3% BSA, 0.2% Triton X-100 in PBS) prior to primary antibody incubation at 4 °C overnight. Secondary antibody staining was performed for one hour and DAPI for 5 min at room temperature. Finally, slides were mounted with Vectashield mounting media and imaged with a confocal microscope LSM700. The following pairs of primary and secondary antibodies were used: anti-NPY (rabbit polyclonal, Cell Signaling Technology Cat# 11976, RRID:AB_2716286, 1:1000) with goat anti-rabbit Alexa-fluorophore 594 (1:1000; Invitrogen Cat# A32740, RRID: AB_2762824) and anti-β-galactosidase (chicken polyclonal, Abcam Cat# ab9361, RRID:AB_307210, 1:1000) with goat anti-chicken Alexa-fluorophore 488 (1:1000; Invitrogen Cat# A11039, RRID: AB_2534096). N = 3 biological replicates were tested.

LUMINEX ELISA assay

Request a detailed protocol

Plasma from blood was collected in the morning at fixed time on EDTA from 1, 5 and 15 day-old mice after euthanasia and stored at –20°C until use. Samples were run in duplicates using the Mouse Magnetic Luminex Assay for IGF-1 (R&D System) and Milliplex Mouse Pituitary Magnetic Assay for GH (Merck) according to manufacturers’ instructions. Values were read on Bio-Plex 200 (Bio-Rad) and analyzed with the Bio-Plex Manager Software.

Phenotypic analyses of weight

Request a detailed protocol

Postnatal weight measurements were performed every two days from 1dpp to weaning age (21 days). Then, mice were separated according to their genotype, hosted in equal number per cage (n = 5–6) and weighted once per week. As body weight is a continuous variable, we used a formula derived from the formula for the t-test to compute the minimum sample size per genotype: n = 1 + C(s/d)2, where C is dependent on values chosen for significance level (α) and power (1-β), s is the standard deviation and d the expected difference in means. Using α = 5%, β = 90% and an expected difference in means of 1 g at 2 weeks and 2 g later on, we predicted a minimum n of between 10 and 20 and thus decided to increase this number using n size between 20 and 30 per genotype, depending of the age and the sex of the mice. No animals were excluded from the analysis. Animals were blindly weighed until genotyping. E18.5 embryos and postnatal organs were collected and individually, rinsed in PBS and weighted on a 0.001 g scale. All data were generated using three independent mating pairs.

Phenotypic analyses of postnatal lethality

Request a detailed protocol

Material for genotyping was taken the first day of birth and then number of pups for a given litter was assessed every day. To avoid bias, we excluded from the analysis the litters where all the pups died due to neglecting mothers, not taking care of their pups.

DEXA scan analyses

Request a detailed protocol

The DEXA analysis allows the assessment of fat and lean mass, bone area, bone mineral content, and bone mineral density. Practically, mutant and WT littermate males at 2 weeks were sent from the Animal Facility of Institut Curie to the Mouse Clinics along with their mother (Ilkirch, France). The phenotypic DEXA analysis was performed at 7 weeks of age using an Ultrafocus DXA digital radiography system, after the mandatory 5-week-quarantine period in the new animal facility. Mice were anesthetized prior analysis and scarified directly after measurements, without a waking up phase.

RNA- Seq data analysis

Request a detailed protocol

Adapters sequences were trimmed using TrimGalore v0.6.2 (https://github.com/FelixKrueger/TrimGalore; Krueger, 2022). N = 3 biological replicates were sequenced per genotype at 3dpp, and n = 2 at 10dpp. Paired-end reads were mapped using STAR_2.6.1 a (Dobin et al., 2013) allowing 4% of mismatches. Gencode vM13 annotation was used to quantify gene expression using quantification mode from STAR. Normalization and differential gene expression were performed using EdgeR R package (v3.22.3) (Robinson et al., 2010). Genes were called as differentially expression if the fold discovery rate (FDR) is lower than 10%.

Statistical analyses

Request a detailed protocol

Significance of obtained data was determined by performing one-way or two-way ANOVA test or two-tailed unpaired, nonparametric Mann Whitney t- tests using GraphPad Prism6 software. p values were considered as significant when p ≤ 0.05. Data points are denoted by stars based on their significance: ****: p ≤ 0.0001; ***: p ≤ 0.001; **: p ≤ 0.01; *: p ≤ 0.05.

Data availability

All data generated or analysed during this study are included in the manuscript and supporting files. RNA-Seq data have been deposited in GEO under accession code GSE153265.

The following data sets were generated

References

    1. Buiting K
    (2010) Prader-Willi syndrome and Angelman syndrome
    American Journal of Medical Genetics. Part C, Seminars in Medical Genetics 154C:365–376.
    https://doi.org/10.1002/ajmg.c.30273
    1. Efstratiadis A
    (1998)
    Genetics of mouse growth
    The International Journal of Developmental Biology 42:955–976.
    1. Kaplan SA
    2. Cohen P
    (2007) The somatomedin hypothesis 2007: 50 years later
    The Journal of Clinical Endocrinology and Metabolism 92:4529–4535.
    https://doi.org/10.1210/jc.2007-0526
    1. Leighton PA
    2. Ingram RS
    3. Eggenschwiler J
    4. Efstratiadis A
    5. Tilghman SM
    (1995)
    Deletion of the H19 gene region in mice disrupts the imprinting of its neighbouring genes
    Nature 375:34–39.

Decision letter

  1. Joshua M Brickman
    Reviewing Editor; University of Copenhagen, Denmark
  2. Carlos Isales
    Senior Editor; Medical College of Georgia at Augusta University, United States
  3. Andrew Ward
    Reviewer; Bath University, United Kingdom
  4. Stormy Chamberlain
    Reviewer; University of Connecticut, United States
  5. Gavin Kelsey
    Reviewer; The Babraham Institute, United Kingdom

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

Decision letter after peer review:

Thank you for submitting your article "The imprinted Zdbf2 gene finely tunes feeding and growth in neonates" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, and the evaluation has been overseen by a Reviewing Editor and Carlos Isales as the Senior Editor. The following individuals involved in review of your submission have agreed to reveal their identity: Andrew Ward (Reviewer #1); Stormy Chamberlain (Reviewer #2); Gavin Kelsey (Reviewer #3).

The reviewers have discussed their reviews with one another, and the Reviewing Editor has drafted this to help you prepare a revised submission.

Essential revisions:

1) Improved statistics and analysis of the growth phenotype. Reviewer 1 makes some excellent suggestions that could improve the manuscript.

2) An attempt to probe mechanism based on better measurement of IGF and GH levels, including Reviewer 2's point about IGF levels in pups with wild type and mutant littermates. The authors should provide a more careful time course of IGF-1 levels in the various genotypes (both gain and loss of function) and littermate competition situations. Reviewer 3 also points to how immunofluorescence studies could be used to characterize how the different mutants effect the population of AgRP/NPY neurons, to better correlate a cellular phenotypes with the growth ones.

3) A better understanding of the Zdbf2 molecular mechanism. RNAseq demonstrated down-regulation of transcripts for AgRP and NPY in early postnatal hypothalamus. Is this a direct effect on the regulation of AgRP and NPY transcription or impaired development or maturation of AgRP/NPY neurons? Immunofluorescence studies enable an estimate to be made of the AgRP/NPY neuron population to identify any deficiencies in Zbf2 mutants (see above).

4) Better analysis of Zbdf2 expression. Review 3 makes a good suggestion, using β galactosidase staining based on the gene trap allele.

5) Specific revisions to the discussion as suggested by the reviewers.

Reviewer #1:

In this paper both gain- and loss-of-function mouse genetic models are used to characterise a post-natal growth phenotype associated with altered dosage of the imprinted Zdbf2 gene, which is expressed in the hypothalamus and pituitary gland from the paternally-inherited allele. A growth deficit in the immediate post-natal period is identified in Zdbf2-KO animals and mirrored in animals with increased Zdbf2 expression. Some details of the growth phenotypes could be more accurately and fully described but overall there is good evidence that Zdbf2, through expression in specific hypothalamic nuclei and the pituitary gland, promotes suckling in pups during the immediate post-partum period. Intriguingly, the transient growth deficiency appears to affect adult lean body mass and brown adipose tissue (BAT), rather than white adipose (WAT) deposition, though this needs to be more firmly established. If correct, this effect is unlike most early-life programming effects, in which disruptions in growth tend to result in excess WAT deposition in later life.

A genetic experiment, combining both gain- and loss-of function alleles to essentially switch expression of Zdbf2 from the paternal to the maternal allele shows that dosage, and not the parental allele origin of expression per se, is what matters. Though interesting, the outcome is unsurprising, as the authors themselves remark, and I would urge caution in using it to draw conclusions about the evolutionary mechanisms that have driven the evolution of imprinting.

Further experiments link the early changes in body size with a deficiency in Igf1 that may be due to reduced nutrient intake very early post-partum. These observations, together with transcriptomic data indicting altered expression of two genes (among only 11 with significantly different expression levels) known to have roles in hypothalamic regulation of feeding behaviour, comprise a compelling description of the role of imprinted Zdbf2 as a regulator of the hypothalamic feeding circuit. As a paternally-expressed gene that promotes growth and nutrient acquisition from the mother Zdbf2 thereby fits the parental conflict hypothesis for imprinted gene evolution. It may also fit with the mother-offspring co-adaptation theory, as suggested in the Discussion, but evidence for this is incomplete. Continued expression of Zdbf2 in the adult brain begs questions about a potential role in controlling behaviour in later life and were it found to influence maternal nurturing behaviour or milk let-down, for instance, this would better support the idea of co-adaptive evolution.

The authors establish that mouse Zdbf2 expression is restricted to hypothalamus and pituitary using both data generated themselves and that from publically available transcriptomic datasets (Figures1 and S1). However, none of the analysis rules out expression of Zdbf2 elsewhere during fetal or early post-natal life and I wonder if this has been established elsewhere? It is important because expression in, for example, skeletal muscle could provide an alternative explanation for the described phenotype. Also, they show levels of expression in both hypothalamus and pituitary arising during the pre-weaning period, between PN4 and PN21, then remaining high well into adulthood (PN70) (Figure 1E,F). Given that the major effect of Zdbf2 KO on growth is later shown to occur immediately after birth it would strengthen the paper to show expression during this period (PN1-3) and earlier, in the fetal brain. In addition, they might comment in the Discussion on the persistent and relatively high level of expression seen in adults, which points to further functions of Zdbf2 in later life, perhaps relating to behaviour.

Zdbf2 KO pups are shown to have a deficit in their ability to gain weight in the immediate post-natal period, in comparison with WT littermates (Figure 2). My major concern here is the statistical treatment of the data: No test appears to have been applied to data in Figure 2B and for that in Figures2C and 2D the use of a Mann Whitney t-test, presumably applied repeatedly to pairs of samples at each time-point, is inappropriate. Instead, ANOVA or similar should be used. Also, the description of the data is inaccurate: "Growth reduction was the most drastic prior to weaning, from 5 to 21dpp" (Lines 176-177). In fact, KO pups gain less weight than their WT littermates during the period PN1-PN5. It would be better to describe the differences in mass between WT and KO animals that the graphs show, rather than growth which is inferred (and more properly analysed later on in Figure 5E). It is potentially confusing to talk about a "growth reduction" in KO pups because the mass of both WT and KO animals increases over time (i.e. both are growing but the KOs less than normal). It would be more accurate to say, e.g. "while there is essentially no difference in mass at e18.5, KOs appear smaller than WT littermates at PN1 and PN5, indicating slower growth in the first few days after birth. The lower mass of KO animals persists into adulthood, measured up to 12 weeks of age". Here, I am also taking issue with the statement that the body weight reduction had "a tendency to minimize with age (Figure 2B)" (Line 179) because any difference in Figure 2B is subtle and the body mass difference is still evident post weaning (Figure 2C) and is not obviously reduced.

I would be more cautious with the statement, "The reduced body weight phenotype was fully penetrant (Figure S2E)." First, the figure legend says 'highly' penetrant, which seems a better term to use because there are animals with mass equal to or greater than mean WT mass. Also, penetrance is difficult to assess because of the naturally broad range in body mass of individuals even within litters. While I'm not suggesting the conclusions are wrong, I wonder if there is a better way to display the data? Instead of normalising the mass of KO pups to that of WT mean/litter could the distribution of body mass be shown for both KO and WT, which would show the shift to the left for KO body mass and also how the body mass range overlaps.

There is a further anomaly in the description of body mass when describing that all organs were uniformly affected, "such that body proportion was maintained" (Lines 186-187). Crucially, no skeletal muscles (nor BAT which becomes important later) were included in Figure S3B, yet the DEXA data (Figure 2 G-I) clearly shows a reduction in lean mass of Zdbf2-KO animals, but not adipose mass. The DEXA data therefore establishes that body proportions have not been maintained.

Together with the tendency to gain less weight than their WT counterparts immediately after birth the Zdbf2-KO pups also exhibit reduced survival. The data look convincing but the description could be clearer. I suggest avoiding use of the terms "failure to thrive" and "survivability". To firm things up, it would be useful to put p values on the numbers of animals alive at key time-points using Chi-squared tests. The comparison of mixed WT+KO litters versus KO-only litters to assess the impact of competition on KO pup survival is a nice experiment but the conclusion "More likely, the impaired survival of Zdbf2-KO neonates is a consequence of intra-litter competition with healthier WT pups." could perhaps be tweaked. It may be better to say, e.g. "perinatal mortality was rescued when there is no competition with WT littermates." Importantly, it is not made clear whether KO pups in the all-KO litters exhibited a similar growth-deficient phenotype seen in KO pups from mixed litters. If so, then the growth deficient phenotype is not based on competition with WT littermates and removing the competition simply allows for more KO offspring to survive the critical first day or so after birth.

The importance of Zdbf2 dosage is next tested using a hypomorphic knock-down allele, which shows a less severe post-natal growth phenotype. Then a GOF allele is used to show that increasing Zdbf2 expression through (incomplete) activation of the normally silent maternal allele results ins early post-natal overgrowth. These are nice experiments that take the work beyond the all-or-nothing nature of most KO v WT comparisons. The results are not clear-cut since only male Zdbf2-GOF animals exhibit enhanced weight gain and statistical analysis of data could be more robust, as detailed below. As for the previous analysis of Zdbf2-KO animals, organ weight data is interpreted to show that all organs are equally affected (Figure S5D), however it is again a striking omission that no skeletal muscles were included among the organs dissected and weighed. Moreover, in this case there is no DEXA data to show how lean and adipose mass contribute to the increased total body mass. Finally, the importance of Zdbf2 dosage is reinforced through an "allelic switching" experiment.

Further experiments show that at a gross level the growth phenotype of Zdbf2-KO animals is not obviously due to gross changes in hypothalamus or pituitary development or in the ability to express various pituitary hormones. Circulating growth hormone levels are also found to be approximately normal at PN5 and PN15, leading the authors to conclude that GH involvement in the phenotype can be excluded. This is perhaps a little overstated because: (1) A transient deficiency in GH closer to the time of birth has not formally been excluded, which is important as the weight data suggest the growth deficiency occurs prior to PN5; (2) Altered sensitivity to GH has not been excluded. In contrast, evidence is shown for an IGF1 deficiency that provides a plausible molecular basis for the growth phenotype. The authors don't formally test the role of IGF1 (which would likely require a lot of work, e.g. to perform some form of rescue experiment) but the deficiency leads them to investigate growth rate using the weight data presented earlier identify (Figure 5E). The differences between WT and KO animals are subtle but most likely correct, in that they provide a better description of the data than that provided earlier. The conclusions would be more robust if the data were subject to some form of statistical testing. Overall though, a nice link is made between the reduction in Igf1 and poor feeding immediately post-partum which ultimately makes sense in terms of the expression pattern of Zdbf2 in brain and pituitary. Organ weights are revisited again, this time in PN3 pups. BAT was included for the first time among organs weighed and found to be the only organ reduced in mass in KO pups. It is admittedly difficult to measure skeletal muscle mass at this young stage but a histological assessment of muscle size would be possible and potentially informative. A reduction in mass of skeletal muscle and WAT, both derivatives of My5+ mesodermal tissue, as a consequence of the IGF1 deficiency immediately after birth, could explain the weight and body composition data presented in the paper. It would be helpful if more data could be included on these two tissues.

Transcriptomic data are presented consistent with transient disruption to feeding circuit genes in the hypothalamus of Zdbf2-KO pups with Npy and Agrp among only eleven genes mis-regulated at PN3. This provides further support for the mechanism underlying Zdbf2 function. It would be nice to see these data validated and extended. For instance, it could be argued that by PN3 the feeding deficiency has already been established and, therefore, interesting to establish whether expression of Nyp and Agrp is disrupted earlier (e.g. by qRT-PCR at late fetal to PN1 stages).

Specific points:

Lines 60-61. The paper by Andergassen et al., (2017: eLife, 6:e25125. DOI: 10.7554/eLife.25125) would be a good addition here.

Lines 130-143. I note the detailed description of Zdbf2 expression in hypothalamic structures is consistent with analysis in a paper currently on bioRxiv (preprint doi: https://doi.org/10.1101/2020.07.27.222893) that they may wish to cite.

Line 174 (FiguresS2A,B). Data look OK but would be useful to see some Chi-square tests done on the data to back up the statement that at birth mendelian and sex ratios were normal.

Line 175-176. "Failure to thrive" is a loose term here that does not accurately describe the data in Figures2B-D, which show a deficit in the ability KO pups to gain weight, in comparison with WT litermates. Appropriate stats are needed (ANOVA, not t-tests).

Line 177. For "2 weeks of age" refer to "PN15". Use days in this pre-weaning period to match the figure (2C).

Lines 178-179. Change "Body weight reduction persisted through adulthood". Body mass measurements are shown up to 12 weeks of age, which is only a fraction of adult life.

Lines 182-3. The statement, "not due to a developmental delay" seems a little too emphatic given the level of analysis is to assess gross morphological features. Perhaps e.g. "not obviously"?

Line 185 (Figure S2F-I). How many animals scored in each case?

Line 186. Change "all organs" to e.g. "a range of organs" since not all organs of the body have been sampled.

Figure S3B shows organ weight proportional to body mass, which is fine but it would also be useful to show a direct comparison of organ mass for WT and KO animals.

Lines 189-200. A number of figure references are incorrect, perhaps due to a late decision to remove or move one or more panels.

Lines 206-207. The inability to detect ZBDF2 protein is a weakness (we are all at the mercy of good antibodies) but it is nice to see this weakness discussed, here and elsewhere. Overall, the weight of evidence for the Zdbf2-KO being a null allele is strong.

Line 209. Again "fail to thrive" is a loose and unhelpful term that could be replaced with something more descriptive.

Lines 216-220. Several more figure references are incorrect. Also, is there a Table missing for Survival of pups carrying the maternal Liz deletion?

Lines 243-244. The statement "Interestingly, these mice displayed significant weight reduction compared to their WT littermates (Figure S4D)" needs to be backed up through appropriate statistical testing (e.g. ANOVA) of the weight data.

Line 245-246. Describe the data more accurately. Presumably "on average" refers to a comparison of means and consider making a more direct comparison with the Zdbf2-KO data.

Line 248. The end of this sentence needs rewriting to make proper sense.

Line 251. "The Zdbf2-GOF lines carry a ~ 900bp" is inaccurate as overlapping deletions of 924 and 768 bp are shown in S4E. Also, it seems more important to point out that both deletions remove 5' portions of the gDMR region and Liz exon 1.

Line 255. This sentence conflates sDMR methylation assays (Figure 3A and B) with RT-qPCR expression analysis (Figure 3C). All needs to be made clearer.

Figure 3C. Stats should be ANOVA, not repeated t-tests.

Line 262. It is inaccurate to state "animals were consistently overweight" when only males appear to be affected. Also, the difference for males appears marginal (Figure 3D) and there are no stats to support it. It would help to reorder the description of Results, e.g. explain first that the difference is specific to males and peaks at around PN15 during the pre-weaning period (3D and E). Male Zdbf2-GOF animals then remain heavier than WT into adult life, measured until 10 weeks of age (3F). Stats at 2 wks of age (3G) support this.

Line 263, "viability was normal". More accurately, 'viability was similar to that of WT controls'.

Line 266-267, "increased body weight of Zdbf2-GOF males was more pronounced during the nursing period (Figure 3D, G and H). Is this true? The differences are subtle in the pre-weaning period, as shown in Figure 3C and although appearing to diminish after PN21 this effect is not evident in Figures3E and 3F.

Line 271, "while females displayed standard normal distribution". This could be made clearer, e.g. 'female Zdbf2-GOF animal had a weight distribution similar to that of WT'. Also need to describe the body mass distribution of males, where it appears the majority are larger than the WT mean but some are smaller (the distribution looks potentially bimodal).

Line 342. IGF1 is a regulator of post-natal growth, not just a marker (also, use commas not dashes to punctuate).

Lines 355-365. Analysis of growth rates. Care is needed in the interpretation, especially when considering the narrowing of the weight difference between WT and KO animals from 1 week to 6 weeks of age, which is associated with an "enhanced post-pubertal growth spurt". Is the % difference significant given that total body weight is considerably increased over this period?

Line 373, "Having determined that Zdbf2-KO animals exhibit an Igf-1-like phenotype". This could be better expressed, e.g. "Having determined that Zdbf2-KO animals are deficient in Igf-1".

Lines 396-409. Data in the two figures described in this section (Figures6 and S7C) need more careful explanation in order to be clearer about which genes are significantly misregulated in the RNA-seq data.

Line 408. Change "continuously misregulated in the hypothalamus at 10dpp" to "also misregulated in the hypothalamus at 10dpp".

Line 409. I dislike the use of the phrase "catch-up phase" for reasons outlined earlier (see comment on lines 355-365).

Line 489-490. It is not clear that you are able to distinguish support for the parental conflict hypothesis (Zdbf2 acting as a paternally expressed promoter of resource acquisition from the mother) from that for co-adaptation, based on your evidence. Support in favour of the letter would require demonstration of a reciprocal function in the mother, such as an influence on milk let-down or nurturing behaviour.

Line 499-504. In the case of PWS the hyperphagia is prolonged and results in excess adipose deposition for which you have no evidence in your GOF mice. The lack of excess body weight out to 18 months is not shown and it would be interesting to know if the reduction in lean body mass and BAT is maintained, which might impact the glucose handling capability of Zdbf2-KO mice.

Reviewer #2:

Glaser et al., generated loss- and gain-of-function mutations in Zdbf2, where the former reduced protein levels and the latter increased them. They carefully analyzed expression of Zdbf2 throughout the hypothalamus and pituitary. They carefully quantified the growth deficits (or increases) in both male and female mice. Although they sought to identify the underlying mechanism for the growth loss (or gain, in the case of the gain of function mutants), they largely found that most of the known regulators of appetite and growth were not changed. This negative data is important. Plasma IGF-1 was decreased in the Zdbf2 knockout mice, and Npy and Agrp seemed increased in RNAseq data from hypothalami. They conclude that perhaps the lack of nutrition in neonates led to decreased IGF-1.

Strengths:

– The generation of the mutant mice was elegant, careful, and well-informed based on previous knowledge.

– The characterization of the mice was careful and performed with adequate numbers to give confidence in the findings.

– The inclusion of Zdbf2 loss of function and gain of function mutations demonstrates a role for the gene in regulating growth and provides excellent tool to better understand the mechanisms of how it impacts growth.

– The manuscript is well-written and easy to understand.

Weaknesses:

– The impact of this finding is predicated on the importance of imprinted genes in the hypothalamus and the hypothalamus in growth regulation. The authors cite Gregg et al., (2010) to justify the importance of imprinted genes in the hypothalamus, but that same paper also claimed that there were ~2,000 imprinted genes, a finding which has not been replicated.

– The lack of a similar role for ZDBF2 in human disease should be discussed carefully, because it points to the role of the gene/protein in placenta. ZDBF2 is expressed (and likely imprinted, according to publicly available GTex data) in many different human tissues. The relevance to humans should be discussed because it has broader implications for the conserved function of imprinted genes. In this case, the placenta seems to be the primary driver.

– It is helpful to compare this mouse model to the PWS mouse, because the PWS mouse model also shows failure to suckle that can be rescued by removing wild-type litter mates and reduced growth. However, the authors refer to PWS and then discuss the Magel2 deficient mouse, due to the metabolic defects. The PWS IC deletion mouse does not have clear metabolic defects (at least to my knowledge!) and MAGEL2-deficient humans do not phenocopy PWS. The authors should more carefully discuss this literature in the introduction and discussion.

– The poor suckle in the neonates precedes reduced milk and reduced growth. It seems pretty hard to disentangle cause and effect. Do pups with wild-type litter mates end up smaller than pups with only ko litter mates? Are the IGF-1 levels improved in mice with only ko litter mates compared to those with wt litter mates? This would go a long way to supporting the idea that IGF-1 levels follow after reduced milk intake.

– In Figure 6a-the authors should include pups from the ko/ko matings to compare 3 dpp milk weight in pups that only have ko competition. When wt litter mates are included, many of the kos are destined to die, and most of the death appears to occur in the first week. Are pups close to death driving this difference?

– It would be important to learn whether plasma IGF-1 levels catch up after 15 dpp. Perhaps a more careful time course of IGF-1 levels in the various genotypes and littermate competition situations should be followed, if this is the hypothesis as to why the mice are smaller (and larger in GOF Zdbf2 mutants; see next comment).

– Given that the GOF Zdbf2 mice show increased body weight, the authors should measure plasma IGF-1 to see if it is increased. It may help disentangle cause and effect of feeding and IGF-1. Do GOF Zdbf2 male mice have increased milk weight?

– The details of the RNA-seq experiments should be provided. How many replicates of each genotype were used? What sequencing depth? How many significantly DE genes in total and do they reveal anything when looking at the data in total (i.e. gene ontology)? It is not clear how solid the NPY and AGRP data are without this information.

Altogether, the data strongly support a role for Zdbf2 in regulating growth in mice. However, the data do not reveal the mechanism underlying this. The reduced IGF-1 levels are correlated with the reduced growth. The hypothesis that this is driven by reduced neonatal feeding is not convincingly supported by the data. It is also not clear whether Zdbf2 performs a similar function in humans.

These data may be very helpful in understanding the relationship between feeding and growth in mouse models. This mouse looks very much like a PWS-IC deletion mouse model. Perhaps similar pathways are disrupted in these models, and may reveal mechanisms of neonatal failure to thrive-is it hypotonia or lack of drive to eat? On one hand, understanding this in mice is not terribly important, but due to similarity with PWS, it has implications for human disease and perhaps why that mouse model does not model all of the features of the human disorder.

– The IGF-1 levels should be more carefully measured over a developmental time course and in ko mice with and without wild-type litter mates.

– The 3 dpp milk/stomach weight should also be quantified in the ko mice with ko only litter mates.

– Both IGF-1 levels and 3 dpp stomach weights should be measured in GOF mice. This might reveal whether drive to eat is really the effect of Zdbf2 perturbation or whether the IGF-1 levels are affected independent of feeding.

– The details of the RNA-seq experiments should be provided. These should include how many replicates and the sequencing depth. The data should be deposited in a public database, as well.

– As discussed above, the role of imprinted genes in the hypothalamus should be discussed more carefully. This example is seemingly specific to the mouse gene. The human ZDBF2 gene is expressed in many tissues and only IUGR may be impacted by expression defects in humans, and this seems to be maternally-expressed ZDBF2 (?) in placenta. As written, it oversells the impact of the data here. Although ZDBF2 regulates growth in mice, it isn’t clear that this also happens in humans and it’s not clear the mechanism, although it is correlated with low IGF-1 levels.

– Specifically, the authors shouldn’t discount the Gregg et al., finding that there are many more imprinted genes and then cite the paper to justify the importance of imprinted genes in the hypothalamus.

– The Magel2 deficient mouse is not a PWS mouse model and humans with deletions of MAGEL2 do not have overt metabolic defects. The details of the PWS versus Magel2 mouse model should be articulated more accurately.

– The potential for divergence between this role for Zdbf2 in humans versus mouse should be discussed. Even if it is a mouse-specific phenotype, it might help understand how feeding/growth is regulated. It may be that mutations in Zdbf2 are missed in humans, but this possibility should be discussed.

Reviewer #3:

In this manuscript, Glaser et al., report further analysis of the physiological effects of imprinting of the Liz-Zdbf2 locus in the mouse. Liz-Zdbf2 is of particular interest, because its imprinting is conferred by a transient, maternal germline differentially methylated region (DMR) which orchestrates the long-term monoallelic expression of Zdbf2. In previous work, the authors showed that abrogation of Liz transcription prevents expression of Zdbf2, resulting in postnatal growth retardation (Greenberg et al., 2017); they also reported prominent expression of Zdbf2 in the hypothalamus and pituitary dependent on prior Liz transcription. In the current manuscript, the authors explore the role of hypothalamic Zdbf2 expression and dosage, and the resulting phenotypic correlates, in more detail.

First, the authors generate a loss-of-function mutant of Zdbf2 to validate that the growth phenotype found upon deletion of Liz and resultant inability to express Zdbf2 can definitively be attributed to Zdbf2. They then report a very detailed assessment of the growth phenotype, making important observations, such as it is restricted to early post-natal period, there is a fully proportionate effect on organ weights, there is reduced viability in the immediate post-natal period that can be mitigated by removal of competition from wild-type litter mates, etc. The authors then go on to show there is a dosage effect, which they are able to do by use of a hypomorphic allele and a serendipitously produced gain-of-function allele. Demonstrating that there is a phenotypic effect (modest over-growth) of near biallelic expression is an important outcome for an imprinted gene – rather few studies have done this. Further, by combining the various alleles they have in hand, the authors show that providing Zdbf2 expression from the maternal rather than paternal allele (albeit at reduced level) is equally functional (bar a slight reduction in weight), suggesting that which parental allele expression emanates from is immaterial. This finding leads to some limited discussion about the evolutionary basis for genomic imprinting.

Zdbf2 is expressed in hypothalamus structures and the anterior and intermediate pituitary, however, no detectable effect on development of these tissues, presence of the various hormone-producing cells in the pituitary, is apparent in the knock-out, although RNA-seq revealed transient reduction in transcript levels of the oxerogenic neuropeptides AgRP and NPY. Reduced IGF1 circulating levels, which could account for growth restriction, which is apparently not mediated by any effect on GH, lead the authors to conclude that IGF1 reduction may be secondary to reduced food intake in the immediate perinatal period.

The strengths of this manuscript are that the work is done to a very high standard: the analysis is detailed and comprehensive, the data look robust, and the results and their implications discussed in a highly informed and accessible manner.

A weakness is that the molecular function of ZDBF2 is left open.

In addition, in relation to the inferences they draw on the evolution of imprinting, the authors need to be cautious not to make over-interpretations and thereby risk misleading readers.

1. The molecular function of ZDBF2 is left open. By RNA-seq, the authors report down-regulation of transcripts for AgRP and NPY in early postnatal hypothalamus, but does this reflect a direct effect on the regulation of AgRP and NPY transcription or impaired development or maturation of AgRP/NPY neurons? Immunofluorescence studies should enable an estimate to be made of the AgRP/NPY neuron population to identify any deficit. The authors comment on the lack of ZDBF2 antibodies, which would preclude co-IF studies to evaluate whether ZDBF2 is expressed within and has a cell-autonomous effect in these neurons. Perhaps co-IF using a β-galactosidase antibody in the gene-trap line would be productive.

2. In relation to their experiment in which they 'reverse' the imprinting of Zdbf2 by enforcing expression from the maternal allele, the authors conclude that "Zdbf2 functions on postnatal body homeostasis in a dose-dependent but parent-of-origin independent manner" (line 449-450) and "the parental information is not essential per se." (line 454). The authors need to be cautious not to over-interpret this result, and thereby risk misleading readers, as it tests the status quo rather than evolutionary history of the allele (which is not possible to do). The parental origin is probably more a reflection of the evolution of the imprinted state. For example, population genetic based theories, such as the 'conflict hypothesis', predict invasion of growth-promoting paternal alleles in the population in the context of maternal restraint. But at this point in time, the expression has attained an optimum so that artificially switching allelic expression whilst largely retaining dosage is predicted to be inconsequential (all other things being equal). Once the allele is expressed in offspring, its parental origin is not apparent other than in relation to other imprinted genes in cis.

3. Further in the Discussion (line 488-490), the authors conclude that "This function agrees with the co-adaptation theory according to which genomic imprinting evolved to coordinate interactions between the offspring and the mother." The authors should also consider that their results are equally compatible with the kinship/parental conflict theory. Indeed, you see this in the observation that Zdbf2-KO offspring survive better when there is no competition from wild-type littermates, i.e., a stronger Zdbf2 allele (the WT) that is better able to extract resources from mothers would spread more quickly through the population than the weaker allele (the KO). This is completely aligned with a kinship/conflict model.

4. In the Discussion (line 465-467), the authors comment that "similar IGF-1-related growth defects have been reported upon alteration of other imprinted loci: the Dlk1-Dio3 cluster and the Cdkn1c gene." The authors should consider adding Rasgrf1, whose role in IGF-1 regulation of postnatal growth was reported earlier (Itier et al., 1998), and which probably represents the first imprinted gene implicated in post-natal growth control. Rasgrf1 might also be useful as a counterpoint in the discussion as that report and subsequent studies (e.g., Drake et al., 2009) implicated defects in the hypothalamic-pituitary axis and growth hormone regulation.

https://doi.org/10.7554/eLife.65641.sa1

Author response

Reviewer #1:

[…]

The authors establish that mouse Zdbf2 expression is restricted to hypothalamus and pituitary using both data generated themselves and that from publically available transcriptomic datasets (Figures1 and S1). However, none of the analysis rules out expression of Zdbf2 elsewhere during fetal or early post-natal life and I wonder if this has been established elsewhere?

From emouseatlas (http://www.emouseatlas.org/emagewebapp/pages/emage_data_browse.jsf), Zdbf2 appears expressed in most brain structures (including pituitary gland) at E14.5; some expression is also detected in spinal cord, upper leg and upper arm.

We also used our Zdbf2-LacZ reporter line to assess Zdbf2 expression in the developing embryo. Β-galactosidase staining at E12.5 embryonic days supports expression of Zdbf2 in the developing brain and in the spinal cord, as depicted now as Figure 1—figure supplement 1B.

It is important because expression in, for example, skeletal muscle could provide an alternative explanation for the described phenotype. Also, they show levels of expression in both hypothalamus and pituitary arising during the pre-weaning period, between PN4 and PN21, then remaining high well into adulthood (PN70) (Figure 1E,F). Given that the major effect of Zdbf2 KO on growth is later shown to occur immediately after birth it would strengthen the paper to show expression during this period (PN1-3) and earlier, in the fetal brain.

We have now included 1dpp as an earlier timepoint in the Zdbf2 RT-qPCR of hypothalamus and pituitary gland, as reported in Figure 1E-F. Expression levels are similar to 4dpp, confirming that noticeable increase occurs past 7dpp. Due to the difficulty to collect hypothalamic and pituitary glands from fetal brains (specifically and in adequate quantity), we did not investigate prenatal timepoints. Moreover, no difference in weight has been detected before birth (Figure 2B). Finally, regarding the comment about skeletal muscle, we did not see any expression in this tissue in our multi-tissue RT-qPCR assessment (Figure 1—figure supplement 1A, quadriceps and tongue).

In addition, they might comment in the Discussion on the persistent and relatively high level of expression seen in adults, which points to further functions of Zdbf2 in later life, perhaps relating to behaviour.

Thanks for the comment, we have adjusted the discussion (last sentence of the first Discussion paragraph). (Lines 436-437)

Zdbf2 KO pups are shown to have a deficit in their ability to gain weight in the immediate post-natal period, in comparison with WT littermates (Figure 2). My major concern here is the statistical treatment of the data: No test appears to have been applied to data in Figure 2B and for that in Figures2C and 2D the use of a Mann Whitney t-test, presumably applied repeatedly to pairs of samples at each time-point, is inappropriate. Instead, ANOVA or similar should be used.

As recommended, we have improved our statistical tests. Figures 2B, Figure 3—figure supplement 1D and Figure 3D now present a comparison of each data point with a Mann Whitney t-test. Here, we are comparing two groups (WT versus KO) independently at each time point, a Mann Whitney t-test is thus the appropriate test to use. Regarding the growth curves in Figures 2C-D and 3E-F, we implemented a two-way ANOVA test comparing WT and KO for all time points, globally. Results of the test are depicted in the graphs and explained in the legends.

Also, the description of the data is inaccurate: "Growth reduction was the most drastic prior to weaning, from 5 to 21dpp" (Lines 176-177). In fact, KO pups gain less weight than their WT littermates during the period PN1-PN5. It would be better to describe the differences in mass between WT and KO animals that the graphs show, rather than growth which is inferred (and more properly analysed later on in Figure 5E). It is potentially confusing to talk about a "growth reduction" in KO pups because the mass of both WT and KO animals increases over time (i.e. both are growing but the KOs less than normal). It would be more accurate to say, e.g. "while there is essentially no difference in mass at e18.5, KOs appear smaller than WT littermates at PN1 and PN5, indicating slower growth in the first few days after birth. The lower mass of KO animals persists into adulthood, measured up to 12 weeks of age". Here, I am also taking issue with the statement that the body weight reduction had "a tendency to minimize with age (Figure 2B)" (Line 179) because any difference in Figure 2B is subtle and the body mass difference is still evident post weaning (Figure 2C) and is not obviously reduced.

Thanks for the remark. These clarifications have been incorporated in the text.

I would be more cautious with the statement, "The reduced body weight phenotype was fully penetrant (Figure S2E)." First, the figure legend says 'highly' penetrant, which seems a better term to use because there are animals with mass equal to or greater than mean WT mass. Also, penetrance is difficult to assess because of the naturally broad range in body mass of individuals even within litters. While I'm not suggesting the conclusions are wrong, I wonder if there is a better way to display the data? Instead of normalising the mass of KO pups to that of WT mean/litter could the distribution of body mass be shown for both KO and WT, which would show the shift to the left for KO body mass and also how the body mass range overlaps.

We have taken this suggestion into account and replaced the graph in Figure 2—figure supplement 1E by a curve showing body mass distribution for both WT and KO at 15dpp, instead of normalization over WT. We can clearly see the shift of the distribution of KO mass to the left. The same has been applied for the Zdbf2-GOF body mass distribution in males, where a body mass shift towards bigger animals is observed (Figure 3—figure supplement 2E). We have also replaced “fully” by “highly” penetrant in the main text.

There is a further anomaly in the description of body mass when describing that all organs were uniformly affected, "such that body proportion was maintained" (Lines 186-187). Crucially, no skeletal muscles (nor BAT which becomes important later) were included in Figure S3B, yet the DEXA data (Figure 2 G-I) clearly shows a reduction in lean mass of Zdbf2-KO animals, but not adipose mass. The DEXA data therefore establishes that body proportions have not been maintained.

We corrected this statement accordingly.

Together with the tendency to gain less weight than their WT counterparts immediately after birth the Zdbf2-KO pups also exhibit reduced survival. The data look convincing but the description could be clearer. I suggest avoiding use of the terms "failure to thrive" and "survivability". To firm things up, it would be useful to put p values on the numbers of animals alive at key time-points using Chi-squared tests.

The terms “failure to thrive" and "survivability" have been corrected and a Chi2 test has been applied to the last data point of both viability graphs (Figure 2J-K).

The comparison of mixed WT+KO litters versus KO-only litters to assess the impact of competition on KO pup survival is a nice experiment but the conclusion "More likely, the impaired survival of Zdbf2-KO neonates is a consequence of intra-litter competition with healthier WT pups." could perhaps be tweaked. It may be better to say, e.g. "perinatal mortality was rescued when there is no competition with WT littermates."

We corrected this statement accordingly. (Lines 227-228)

Importantly, it is not made clear whether KO pups in the all-KO litters exhibited a similar growth-deficient phenotype seen in KO pups from mixed litters. If so, then the growth deficient phenotype is not based on competition with WT littermates and removing the competition simply allows for more KO offspring to survive the critical first day or so after birth.

We have ourselves thought about this at large. However, as pointed above by this reviewer, “body mass of individuals shows a naturally broad range, even within litters”. As there is no WT littermates in this experimental set up, we would need here to make comparison between litters, on such an intrinsically variable measurable as the weight, and which happens to be affected in a rather subtle manner in the Zdbf2-KO animals. Three types of crosses would be required: KO-only, KO from mixed KO and WT pups and also, WT-only. Inter-litter variability could be modulated by keeping only litters with similar sizes, and by redoing the three types of crosses at the same time, to reduce seasonal impact. This would involve a high number of crosses and mice, and still a significant amount of uncertainty regarding our ability to make conclusions, as inter-litter differences will still persist. We therefore decided to not do these measurements and we hope that the reviewer will understand.

The importance of Zdbf2 dosage is next tested using a hypomorphic knock-down allele, which shows a less severe post-natal growth phenotype. Then a GOF allele is used to show that increasing Zdbf2 expression through (incomplete) activation of the normally silent maternal allele results ins early post-natal overgrowth. These are nice experiments that take the work beyond the all-or-nothing nature of most KO v WT comparisons. The results are not clear-cut since only male Zdbf2-GOF animals exhibit enhanced weight gain and statistical analysis of data could be more robust, as detailed below. As for the previous analysis of Zdbf2-KO animals, organ weight data is interpreted to show that all organs are equally affected (Figure S5D), however it is again a striking omission that no skeletal muscles were included among the organs dissected and weighed. Moreover, in this case there is no DEXA data to show how lean and adipose mass contribute to the increased total body mass. Finally, the importance of Zdbf2 dosage is reinforced through an "allelic switching" experiment.

Thanks for the comments. The main text has been modified regarding the conclusion on the organ weight data and the same statistical tests were implemented as for the Zdbf2-KO data.

Further experiments show that at a gross level the growth phenotype of Zfb2-KO animals is not obviously due to gross changes in hypothalamus or pituitary development or in the ability to express various pituitary hormones. Circulating growth hormone levels are also found to be approximately normal at PN5 and PN15, leading the authors to conclude that GH involvement in the phenotype can be excluded. This is perhaps a little overstated because: (1) A transient deficiency in GH closer to the time of birth has not formally been excluded, which is important as the weight data suggest the growth deficiency occurs prior to PN5; (2) Altered sensitivity to GH has not been excluded.

Thanks for this remark. We removed this strong conclusion from the results and have added some nuance in the discussion (Line 471-472).

In contrast, evidence is shown for an IGF1 deficiency that provides a plausible molecular basis for the growth phenotype. The authors don't formally test the role of IGF1 (which would likely require a lot of work, e.g. to perform some form of rescue experiment) but the deficiency leads them to investigate growth rate using the weight data presented earlier identify (Figure 5E). The differences between WT and KO animals are subtle but most likely correct, in that they provide a better description of the data than that provided earlier. The conclusions would be more robust if the data were subject to some form of statistical testing.

It is unclear to us what kind of statistical test could be performed here. Charalambous 2012 uses the same type of growth rate graph and did not apply any test (DOI 10.1016/j.cmet.2012.01.006). Each data point was generated from this formula: (average weight at day n – average weight at day n-1)/ average weight at day n-1. Statistics have already been done on weight data, when looking at the growth curves (Figures 2C and D). What we have done though is to plot, as a matter of comparison, the growth rate curve from animals from the maternal transmission crosses (now as Figure 5—figure supplement 2A), who are asymptomatic (no weight difference). Maternal KO growth rate appears similar to WT, unlike what we observed in paternal KO animals. Although no statistics were applied here, this reinforces that the difference we see in the Zdbf2-KO growth rate curve is real.

Overall though, a nice link is made between the reduction in Igf1 and poor feeding immediately post-partum which ultimately makes sense in terms of the expression pattern of Zdbf2 in brain and pituitary. Organ weights are revisited again, this time in PN3 pups. BAT was included for the first time among organs weighed and found to be the only organ reduced in mass in KO pups. It is admittedly difficult to measure skeletal muscle mass at this young stage but a histological assessment of muscle size would be possible and potentially informative. A reduction in mass of skeletal muscle and WAT, both derivatives of My5+ mesodermal tissue, as a consequence of the IGF1 deficiency immediately after birth, could explain the weight and body composition data presented in the paper. It would be helpful if more data could be included on these two tissues.

This could be interesting investigating indeed, and could give some future work to be done with laboratories that are specialized in muscle development (which is well beyond our own expertise).

Transcriptomic data are presented consistent with transient disruption to feeding circuit genes in the hypothalamus of Zdbf2-KO pups with Npy and Agrp among only eleven genes mis-regulated at PN3. This provides further support for the mechanism underlying Zdbf2 function. It would be nice to see these data validated and extended. For instance, it could be argued that by PN3 the feeding deficiency has already been established and, therefore, interesting to establish whether expression of Nyp and Agrp is disrupted earlier (e.g. by qRT-PCR at late fetal to PN1 stages).

We have now performed RT-qPCR analysis of Npy and Agrp expression in hypothalamus at 1 and 3dpp. First, at 1dpp, we found that mRNA levels of both genes are already significantly downregulated in Zdbf2-KO (Figure 6F). This result is to be considered with our new measurements of plasmatic IGF-1 levels in 1dpp animals (as requested by Reviewer 2): at birth, IGF-1 is already reduced, although not as importantly as at 5 or 15dpp (Figure 5C). Second, at 3dpp, still by RT-qPCR, we found significant decrease of Agrp expression (downregulated but not significantly in RNA-seq) but a non-significant decrease of Npy expression (significantly downregulated in RNA-seq). Importantly, Npy and Agrp were no longer significantly downregulated later on, from our RNA-seq data at 10dpp (Figure 6—figure supplement 1D-E). Altogether, RNA-seq and RT-qPCR data support downregulation of two major hypothalamic genes controlling feeding behavior immediately at birth and during the first days of post-natal life. Additionally, as suggested by Reviewer 3, we have now performed Immunofluorescence on brain sections of Zdbf2-LacZ animals at 15dpp. We were thus able to validate that Zdbf2 is co-expressed with NPY in hypothalamic cells.

Specific points:

All specific points have been addressed. We sincerely thank Reviewer 1 for dedicating so much effort correcting our manuscript and making some really helpful suggestions to improve its quality and accuracy.

Lines 60-61. The paper by Andergassen et al., (2017: eLife, 6:e25125. DOI: 10.7554/eLife.25125) would be a good addition here.

Added

Lines 130-143. I note the detailed description of Zdbf2 expression in hypothalamic structures is consistent with analysis in a paper currently on bioRxiv (preprint doi: https://doi.org/10.1101/2020.07.27.222893) that they may wish to cite.

Added

Line 174 (FiguresS2A,B). Data look OK but would be useful to see some Chi-square tests done on the data to back up the statement that at birth mendelian and sex ratios were normal.

Done

Line 175-176. "Failure to thrive" is a loose term here that does not accurately describe the data in Figures2B-D, which show a deficit in the ability KO pups to gain weight, in comparison with WT litermates. Appropriate stats are needed (ANOVA, not t-tests).

Done

Line 177. For "2 weeks of age" refer to "PN15". Use days in this pre-weaning period to match the figure (2C).

Done

Lines 178-179. Change "Body weight reduction persisted through adulthood". Body mass measurements are shown up to 12 weeks of age, which is only a fraction of adult life.

Changed

Lines 182-3. The statement, "not due to a developmental delay" seems a little too emphatic given the level of analysis is to assess gross morphological features. Perhaps e.g. "not obviously"?

Corrected

Line 185 (Figure S2F-I). How many animals scored in each case?

Appears now in the legend of the Figures (Figure 2—figure supplement 1F-H) (n=18 WT and n=10 KO)

Line 186. Change "all organs" to e.g. "a range of organs" since not all organs of the body have been sampled.

Done

Figure S3B shows organ weight proportional to body mass, which is fine but it would also be useful to show a direct comparison of organ mass for WT and KO animals.

A direct comparison of organ mass for WT and KO was performed but this graph, to our opinion, does not add information compared to the graph showing organ mass proportional to body mass in Figure 2—figure supplement 2B. You can see this graph with the raw data of organ mass measurement in Author response image 1.

Author response image 1
Comparison of organ mass for wild-type and Zdbf2-KO males at 6 weeks-old.

Lines 189-200. A number of figure references are incorrect, perhaps due to a late decision to remove or move one or more panels.

Corrected

Lines 206-207. The inability to detect ZBDF2 protein is a weakness (we are all at the mercy of good antibodies) but it is nice to see this weakness discussed, here and elsewhere. Overall, the weight of evidence for the Zdbf2-KO being a null allele is strong.

Thanks

Line 209. Again "fail to thrive" is a loose and unhelpful term that could be replaced with something more descriptive.

Done

Lines 216-220. Several more figure references are incorrect. Also, is there a Table missing for Survival of pups carrying the maternal Liz deletion?

Corrected and two tables have been added (Figure 2- source data1B and D).

Lines 243-244. The statement "Interestingly, these mice displayed significant weight reduction compared to their WT littermates (Figure S4D)" needs to be backed up through appropriate statistical testing (e.g. ANOVA) of the weight data.

As for Figure 2B, data represent the percentage of mutant weight, normalized to their WT littermates. Normalization has been done per litter to avoid weight difference resulting from difference in litter size. We applied at Mann-Whitney t-test comparing two groups (WT versus mutant) at each time point and show significant results on the graph (Figure 3—figure supplement 1D).

Line 245-246. Describe the data more accurately. Presumably "on average" refers to a comparison of means and consider making a more direct comparison with the Zdbf2-KO data.

Done

Line 248. The end of this sentence needs rewriting to make proper sense.

Done

Line 251. "The Zdbf2-GOF lines carry a ~ 900bp" is inaccurate as overlapping deletions of 924 and 768 bp are shown in S4E. Also, it seems more important to point out that both deletions remove 5' portions of the gDMR region and Liz exon 1.

Corrected

Line 255. This sentence conflates sDMR methylation assays (Figure 3A and B) with RT-qPCR expression analysis (Figure 3C). All needs to be made clearer.

Done

Figure 3C. Stats should be ANOVA, not repeated t-tests.

Done

Line 262. It is inaccurate to state "animals were consistently overweight" when only males appear to be affected. Also, the difference for males appears marginal (Figure 3D) and there are no stats to support it. It would help to reorder the description of Results, e.g. explain first that the difference is specific to males and peaks at around PN15 during the pre-weaning period (3D and E). Male Zdbf2-GOF animals then remain heavier than WT into adult life, measured until 10 weeks of age (3F). Stats at 2 wks of age (3G) support this.

Done, including new statistics.

Line 263, "viability was normal". More accurately, 'viability was similar to that of WT controls'.

Corrected

Line 266-267, "increased body weight of Zdbf2-GOF males was more pronounced during the nursing period (Figure 3D, G and H). Is this true? The differences are subtle in the pre-weaning period, as shown in Figure 3C and although appearing to diminish after PN21 this effect is not evident in Figures3E and 3F.

We have added statistics on Figure 3D showing that the gain in body mass appears almost significant at 5dpp in Zdbf2-GOF males (p-value 0.055) and significant from 7 to 21dpp, using a Mann-Whitney t-test. On that graph, the increase is still visible at 28dpp but does not appears as significant. The growth curve from Figure 3E measuring weight differences from 1 to 19dpp (nursing period, prior to weaning) shows more significant difference between WT and GOF than the measurements from 3 to 10 weeks on Figure 3F (after weaning). Altogether, those observations indicate that the increased of body weight in this mutant is more pronounced during the nursing period.

Line 271, "while females displayed standard normal distribution". This could be made clearer, e.g. 'female Zdbf2-GOF animal had a weight distribution similar to that of WT'. Also need to describe the body mass distribution of males, where it appears the majority are larger than the WT mean but some are smaller (the distribution looks potentially bimodal).

We have changed the type of graph, as suggested previously for similar type of data in Zdbf2-KO pups, by a curve showing body mass distribution for both WT and KO at 15dpp (Figure 3—figure supplement 2E). With this new graph, the bimodal distribution is less visible as before with the normalization towards WT weights, suggesting that most GOF pups were larger than WT controls, as written in the main text (Line 272-273).

Line 342. IGF1 is a regulator of post-natal growth, not just a marker (also, use commas not dashes to punctuate).

Modified

Lines 355-365. Analysis of growth rates. Care is needed in the interpretation, especially when considering the narrowing of the weight difference between WT and KO animals from 1 week to 6 weeks of age, which is associated with an "enhanced post-pubertal growth spurt". Is the % difference significant given that total body weight is considerably increased over this period?

As mentioned previously, it is unclear to us what kind of statistical test could be performed on this growth rate curve.

Line 373, "Having determined that Zdbf2-KO animals exhibit an Igf-1-like phenotype". This could be better expressed, e.g. "Having determined that Zdbf2-KO animals are deficient in Igf-1".

Corrected

Lines 396-409. Data in the two figures described in this section (Figures6 and S7C) need more careful explanation in order to be clearer about which genes are significantly misregulated in the RNA-seq data.

Better description provided. We also now provide Figure 6- source data 1 with all misregulated genes and their RPKM, logFC and FDR values in the RNA-seq data at both 3 and 10dpp.

Line 408. Change "continuously misregulated in the hypothalamus at 10dpp" to "also misregulated in the hypothalamus at 10dpp".

Done

Line 409. I dislike the use of the phrase "catch-up phase" for reasons outlined earlier (see comment on lines 355-365).

Removed

Line 489-490. It is not clear that you are able to distinguish support for the parental conflict hypothesis (Zdbf2 acting as a paternally expressed promoter of resource acquisition from the mother) from that for co-adaptation, based on your evidence. Support in favour of the letter would require demonstration of a reciprocal function in the mother, such as an influence on milk let-down or nurturing behaviour.

The discussion has been modulated.

Line 499-504. In the case of PWS the hyperphagia is prolonged and results in excess adipose deposition for which you have no evidence in your GOF mice. The lack of excess body weight out to 18 months is not shown and it would be interesting to know if the reduction in lean body mass and BAT is maintained, which might impact the glucose handling capability of Zdbf2-KO mice.

Thanks for the comment. We did not keep Zdbf2-GOF animals for ageing experiments. However, measurements were made for Zdbf2-KO animals, which kept on being smaller/lighter than WT littermates at 18 months. Although our cohort is of small number, results were consistent. We have added “data not shown” in this sentence.

Reviewer #2:

[…]

– The IGF-1 levels should be more carefully measured over a developmental time course and in ko mice with and without wild-type litter mates.

We have added measurements of circulating IGF-1 at 1dpp, thereby showing now a time course from 1 to 15dpp (see new panel Figure 5C). Interestingly, we observed decreased IGF1 in Zdbf2-KO pups at 1dpp already, which then becomes more and more pronounced over time, reaching 70% decrease at 15dpp. Importantly, we also now report that Npy and Agrp downregulation (which could relate to poor suckling in mutant neonates, and hence reduced IGF-1 levels) is also noticeable by RT-qPCR as early as 1dpp (see new panel Figure 6F). Including this 1dpp point was therefore crucial, and we thank this reviewer for the suggestion. However, we did not explore IGF-1 circulating levels at timepoint later than 15dpp: body mass reduction in the mutants does not become more important with age (Figure 2B-D), and Npy and Agrp resume to normal levels at 10dpp (Figure 6—figure supplement 1E).

Regarding measurements in Zdbf2-KO-only pups, because of the lack of WT littermates, we would need here to make comparison between litters, on such an intrinsically variable as plasma IGF-1 levels. We would actually need three types of crosses: KO-only, KO from mixed KO and WT pups and also, WT-only. Inter-litter variability could be modulated by keeping only litters with similar sizes, and by redoing the three types of crosses at the same time, to reduce seasonal impact. This would involve a high number of crosses and mice, and still a significant amount of uncertainty regarding our ability to make conclusions, as inter-litter differences will still persist. We therefore decided to not do these measurements and hope that the reviewer will understand.

– The 3 dpp milk/stomach weight should also be quantified in the ko mice with ko only litter mates.

We did not perform this experiment for the reason stated above. The subtle phenotype and lack of littermate’s comparison will render the conclusions of those experiments difficult to interpret.

– Both IGF-1 levels and 3 dpp stomach weights should be measured in GOF mice. This might reveal whether drive to eat is really the effect of Zdbf2 perturbation or whether the IGF-1 levels are affected independent of feeding.

We have now measured stomach weights at 3dpp and IGF-1 plasma levels at 5 and 15dpp in Zdbf2-GOF males and WT littermates. Regarding stomach weight (Figure 6—figure supplement 1A), no significant difference was observed at 3dpp between GOF and WT. This goes along with the fact that we did not observe a significant body mass increase in GOF males at this same age (Figures 3D). On Author response image 2 provided hereafter to the reviewer only, we can appreciate that Zdbf2-KO males at 3dpp are smaller than WT littermates and for most of them, this is correlated with smaller stomach mass. However, there is no such correlation for the Zdbf2-GOF males. This implies that increased body mass might not result from enhanced feeding after birth, but rather from later signaling leading to increased body mass visible at 15 dpp (Figure 3G-H and Figure 3—figure supplement 2E).

Author response image 2
Correlation between body mass and stomach mass at 3dpp for Zdbf2-KO males (left panel) and Zdbf2-GOF males (right panel).

Regarding IGF-1 (Figure 5—figure supplement 2B), plasma levels were also not different in Zdbf2-GOF compared to WT males, at both 5dpp and 15dpp. IGF-1 may not be the main driver of the male-specific increased body weight upon ZDBF2 overexpression. Other hypothalamic circuits might be affected. Alternatively, IGF-1 might be (only transiently) affected in these GOF mutants but we were not able to detect it with our experimental set up. We mention these possibilities in the Results.

– The details of the RNA-seq experiments should be provided. These should include how many replicates and the sequencing depth. The data should be deposited in a public database, as well.

Sorry for the omission, n=3 replicates were used per genotype at 3dpp, and n=2 at 10dpp. This information has been added in the legend and in the method section, where we also inform the reader about the sequencing depth. Data have been publicly deposited under the GEO number GSE153265, as it was originally reported in the Material and Method section (under “Data Availability”). Regarding the comment about the number of significant DE genes and their GO analysis: as reported in the results, there are 11 differentially expressed (DE) genes in Zdbf2-KO hypothalami at 3dpp compared to WT littermates. The list of these DE genes was provided in Figure 6—figure supplement 1E, as well as the list of DE genes at 10dpp (n=16). We have now worked on being clearer in the text regarding the RNAseq results. Also, raw values of the RNA-seq are now provided as Figure 6- source data 1. Regarding GO analysis, this could not be performed on such small gene set.

– As discussed above, the role of imprinted genes in the hypothalamus should be discussed more carefully. This example is seemingly specific to the mouse gene. The human ZDBF2 gene is expressed in many tissues and only IUGR may be impacted by expression defects in humans, and this seems to be maternally-expressed ZDBF2 (?) in placenta. As written, it oversells the impact of the data here. Although ZDBF2 regulates growth in mice, it isn't clear that this also happens in humans and it's not clear the mechanism, although it is correlated with low IGF-1 levels.

Thanks for opening this discussion. As in the mouse, ZDBF2 is highly expressed in brain tissues in humans, including the pituitary gland (GTEXdata). It has also been shown to be expressed in human placenta (where it is also paternally expressed, as shown by our own work in Duffié et al., Genes and Dev 2014 and by others, such as Monteagudo-Sanchez et al., Clinical Epigenetics 2019). It is difficult to know whether placental expression is relevant, as there is no direct comparison with brain tissues in terms of levels. In the mouse, we can score modest expression in the placenta (Figure 1—figure supplement 1A)—which we previously showed to emanate from the labyrinth layer in Greenberg, Glaser et al., Nature Genetics 2017—but the level is by 10-fold lower than in the hypothalamus, and we did not find any placental phenotype upon Zdbf2 depletion in this previous study. Concerning ZDBF2 role in human trait and physiology, there has indeed been an association between IUGR and reduced ZDBF2 expression. However, difference was only seen in male conceptuses with IUGR, not females, and notably, ZDBF2 levels in the placenta of IUGR males were similar to the levels seen in the placenta of control female conceptuses. It is also unclear as to whether the reduction in expression is cell autonomous, there could be differences in the cellular composition of IUGR placentae compared to controls. Finally, it could also be that the IUGR condition triggers ZDBF2 downregulation, rather than the reverse. As a whole, it is true that there is not so much literature about ZDBF2 in humans and that is mostly linked to placenta, but we feel that the link between ZDBF2 and placental function in humans is far from being proved, and would require independent cohorts to be analyzed, single-cell analysis to avoid confounding effects etc….

Nonetheless, we have now nuanced the discussion, as we cannot indeed exclude that ZDBF2 may perform different functions in mice and humans : “Finally, decreased ZDBF2 levels have recently been associated with intra-uterine growth restriction (IUGR) in humans (Monteagudo-Sánchez et al., 2019). Whether this is linked to impaired fetal IGF-1 production or to distinct roles of ZDBF2 related to placental development in humans would be interesting to assess, in regards to conservation or not of imprinted gene function across mammals” (Lines 484-488).

– Specifically, the authors shouldn't discount the Gregg et al., finding that there are many more imprinted genes and then cite the paper to justify the importance of imprinted genes in the hypothalamus.

We originally cited Gregg et al., (2010) for their analysis of expression of known (and well established) imprinted genes, showing their prevalent expression in the hypothalamus compared to other brain structures. To further justify that the hypothalamus is a privileged site for imprinted gene expression, we now refer to a more recent study (preprint) that explored brain scRNA-seq datasets and showed overrepresentation of (known) imprinted gene expression in hypothalamic neurons [Higgs et al., bioRxiv 2021 (https://doi.org/10.1101/2020.07.27.222893)].

– The Magel2 deficient mouse is not a PWS mouse model and humans with deletions of MAGEL2 do not have overt metabolic defects. The details of the PWS versus Magel2 mouse model should be articulated more accurately.

We are sorry if we were unclear. Our point is not to compare Magel2-KO mice with PWS patients (which has been done by others, including the refence we cite) but to emphasize that there are cases where initial poor feeding and altered capacity to gain weight after birth can be associated with obesity and metabolic defects later in life. We have modified our sentence: “This is observed in mouse KO models of the imprinted Magel2 gene that map to the Prader Willi syndrome (PWS) region, recapitulating some features of PWS patients, who after a failure to thrive as young infants exhibit a catch-up phase leading to overweight and hyperphagia (Bischof et al., 2007).” (Line 516-520).

– The potential for divergence between this role for Zdbf2 in humans versus mouse should be discussed. Even if it is a mouse-specific phenotype, it might help understand how feeding/growth is regulated. It may be that mutations in Zdbf2 are missed in humans, but this possibility should be discussed.

As mentioned in #5, this nuance has been added in the Discussion.

Reviewer #3:

[…]

1. The molecular function of ZDBF2 is left open. By RNA-seq, the authors report down-regulation of transcripts for AgRP and NPY in early postnatal hypothalamus, but does this reflect a direct effect on the regulation of AgRP and NPY transcription or impaired development or maturation of AgRP/NPY neurons? Immunofluorescence studies should enable an estimate to be made of the AgRP/NPY neuron population to identify any deficit. The authors comment on the lack of ZDBF2 antibodies, which would preclude co-IF studies to evaluate whether ZDBF2 is expressed within and has a cell-autonomous effect in these neurons. Perhaps co-IF using a β-galactosidase antibody in the gene-trap line would be productive.

Thanks for this suggestion. We have now performed co-localization analysis of NPY and ZDBF2 using anti-NPY and anti-β-galactosidase antibodies on 15dpp brain sections of Zdbf2-lacZ animals. We observed both signals in the paraventricular nucleus of the hypothalamus, the known site of NPY expression (see the new Figure 6D). This suggests that Zdbf2 is expressed in NPY+ neurons.

We tried to confirm these results with an anti-AgRP antibody (rabbit anti-AgRP, Phoenix Pharmaceuticals cat. No. H-003-57), quite popular in the literature, at least on adult brains. Unfortunately, we failed to obtain specific staining in pre-puberal brains after several round of troubleshooting.

We also tried to quantify NPY intensity and NPY-expressing cells on brain sections from WT and Zdbf2-KO at 3dpp but the results were not interpretable. First, NPY staining only allows detection of the fibers (local and projections) but not the neuron’s soma. It is extremely complicated to quantify the intensity of the signal in the fibers and impossible to count NPYexpressing cells. Second, we observed important variations in the intensity and distribution of the signal across different sections from the same animal, implying that comparison of WT and KO could only be done on sections from the exact same coordinate in the hypothalamus. Finally, the decrease of Npy mRNA from the RNA-data is rather mild, and its quantification by IF questionable. For these reasons, our attempt to visualize NPY downregulation and/or to count NPY+ cells using IF was, unfortunately, unsuccessful.

2. In relation to their experiment in which they ‘reverse’ the imprinting of Zdbf2 by enforcing expression from the maternal allele, the authors conclude that “Zdbf2 functions on postnatal body homeostasis in a dose-dependent but parent-of-origin independent manner” (line 449-450) and “the parental information is not essential per se.” (line 454). The authors need to be cautious not to over-interpret this result, and thereby risk misleading readers, as it tests the status quo rather than evolutionary history of the allele (which is not possible to do). The parental origin is probably more a reflection of the evolution of the imprinted state. For example, population genetic based theories, such as the ‘conflict hypothesis’, predict invasion of growth-promoting paternal alleles in the population in the context of maternal restraint. But at this point in time, the expression has attained an optimum so that artificially switching allelic expression whilst largely retaining dosage is predicted to be inconsequential (all other things being equal). Once the allele is expressed in offspring, its parental origin is not apparent other than in relation to other imprinted genes in cis.

We have tried to temper this sentence:

“The divergent DNA methylation patterns that are established in the oocyte and the spermatozoon provide opportunities to evolve mono-allelic regulation of expression, but once transmitted to the offspring, the parental origin of expression is not essential per se.” (Line 467-468).

3. Further in the Discussion (line 488-490), the authors conclude that “This function agrees with the co-adaptation theory according to which genomic imprinting evolved to coordinate interactions between the offspring and the mother.” The authors should also consider that their results are equally compatible with the kinship/parental conflict theory. Indeed, you see this in the observation that Zdbf2-KO offspring survive better when there is no competition from wild-type littermates, i.e., a stronger Zdbf2 allele (the WT) that is better able to extract resources from mothers would spread more quickly through the population than the weaker allele (the KO). This is completely aligned with a kinship/conflict model.

We have added this to the discussion, now considering both theories (Lines 508-510).

4. In the Discussion (line 465-467), the authors comment that “similar IGF-1-related growth defects have been reported upon alteration of other imprinted loci: the Dlk1-Dio3 cluster and the Cdkn1c gene.” The authors should consider adding Rasgrf1, whose role in IGF-1 regulation of postnatal growth was reported earlier (Itier et al., 1998), and which probably represents the first imprinted gene implicated in post-natal growth control. Rasgrf1 might also be useful as a counterpoint in the discussion as that report and subsequent studies (e.g., Drake et al., 2009) implicated defects in the hypothalamic-pituitary axis and growth hormone regulation.

Thanks for the remark. We have now added Rasgrf1 in the discussion (Lines 482-485).

https://doi.org/10.7554/eLife.65641.sa2

Article and author information

Author details

  1. Juliane Glaser

    Institut Curie, PSL Research University, INSERM, CNRS, Paris, France
    Present address
    RG Development & Disease, Max Planck Institute for Molecular Genetics, Berlin, Germany
    Contribution
    Conceptualization, Formal analysis, Investigation, Methodology, Resources, Validation, Visualization, Writing – original draft, Writing – review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-6745-6924
  2. Julian Iranzo

    Institut Curie, PSL Research University, INSERM, CNRS, Paris, France
    Contribution
    Investigation, Validation
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-9369-2530
  3. Maud Borensztein

    Institut Curie, PSL Research University, INSERM, CNRS, Paris, France
    Contribution
    Investigation
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-4378-5018
  4. Mattia Marinucci

    Institut Curie, PSL Research University, INSERM, CNRS, Paris, France
    Contribution
    Investigation
    Competing interests
    No competing interests declared
  5. Angelica Gualtieri

    Centre for Endocrinology, William Harvey Research Institute, Barts and the London School of Medicine and Dentistry, Queen Mary University of London, London, United Kingdom
    Contribution
    Formal analysis, Investigation
    Competing interests
    No competing interests declared
  6. Colin Jouhanneau

    Institut Curie, PSL Research University, Animal Transgenesis Platform, Paris, France
    Contribution
    Methodology
    Competing interests
    No competing interests declared
  7. Aurélie Teissandier

    Institut Curie, PSL Research University, INSERM, CNRS, Paris, France
    Contribution
    Formal analysis, Visualization
    Competing interests
    No competing interests declared
  8. Carles Gaston-Massuet

    Centre for Endocrinology, William Harvey Research Institute, Barts and the London School of Medicine and Dentistry, Queen Mary University of London, London, United Kingdom
    Contribution
    Funding acquisition, Visualization, Writing – original draft
    Competing interests
    No competing interests declared
  9. Deborah Bourc'his

    Institut Curie, PSL Research University, INSERM, CNRS, Paris, France
    Contribution
    Conceptualization, Formal analysis, Funding acquisition, Supervision, Writing – original draft, Writing – review and editing
    For correspondence
    deborah.bourchis@curie.fr
    Competing interests
    Reviewing editor, eLife
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-9499-7291

Funding

FP7 Ideas: European Research Council (ERC-Cog EpiRepro)

  • Aurélie Teissandier
  • Deborah Bourc'his

Fondation Bettencourt Schueller

  • Deborah Bourc'his

Ligue Contre le Cancer

  • Juliane Glaser

BTL Charity (GN417/2238)

  • Angelica Gualtieri
  • Carles Gaston-Massuet

Action Medical Research (GN2272)

  • Angelica Gualtieri
  • Carles Gaston-Massuet

DIM Biotherapies

  • Juliane Glaser

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

We would like to thank members of the Bourc’his laboratory for continuous support and stimulation, M Greenberg for critical reading of the manuscript, M Charalambous for methodological and conceptual advices, T Chelmicki and L Marion-Poll for technical help and F El Marjou (Transgenesis Platform of the Institut Curie Animal Facility) for CRISPR in vivo engineering. Luminex assays were performed by the LUMINEX Platform from the CHU Clermont-Ferrand. The laboratory of DB is part of the Laboratory d’Excellence LABEX (LABEX) entitled DEEP (11-LBX0044). This research was supported by the ERC (grant ERC-Cog EpiRepro) and the Bettencourt Schueller Foundation. CGM and AG were supported by grants from Action Medical Research (GN2272) and BTL Charity (GN417/2238). JG was a recipient of PhD fellowships from DIM Biothérapies, Ile-de-France and from La Ligue contre le Cancer.

Ethics

All experimentation was approved by the Animal Care and Use Committee of the Institut Curie (agreement number: C 75-05-18) and adhered to European and National Regulation for the Protection of Vertebrate Animals used for Experimental and other Scientific Purposes (Directive 86/609 and 2010/63).

Senior Editor

  1. Carlos Isales, Medical College of Georgia at Augusta University, United States

Reviewing Editor

  1. Joshua M Brickman, University of Copenhagen, Denmark

Reviewers

  1. Andrew Ward, Bath University, United Kingdom
  2. Stormy Chamberlain, University of Connecticut, United States
  3. Gavin Kelsey, The Babraham Institute, United Kingdom

Publication history

  1. Received: December 10, 2020
  2. Preprint posted: January 30, 2021 (view preprint)
  3. Accepted: January 19, 2022
  4. Accepted Manuscript published: January 20, 2022 (version 1)
  5. Version of Record published: February 2, 2022 (version 2)

Copyright

© 2022, Glaser et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 1,036
    Page views
  • 204
    Downloads
  • 4
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

  1. Juliane Glaser
  2. Julian Iranzo
  3. Maud Borensztein
  4. Mattia Marinucci
  5. Angelica Gualtieri
  6. Colin Jouhanneau
  7. Aurélie Teissandier
  8. Carles Gaston-Massuet
  9. Deborah Bourc'his
(2022)
The imprinted Zdbf2 gene finely tunes control of feeding and growth in neonates
eLife 11:e65641.
https://doi.org/10.7554/eLife.65641

Further reading

    1. Cell Biology
    2. Chromosomes and Gene Expression
    Andria C Schibler, Predrag Jevtic ... Tom Misteli
    Research Article

    The shape and size of the human cell nucleus is highly variable amongst cell types and tissues. Changes in nuclear morphology are associated with disease, including cancer, as well as with premature and normal aging. Despite the very fundamental nature of nuclear morphology, the cellular factors that determine nuclear shape and size are not well understood. To identify regulators of nuclear architecture in a systematic and unbiased fashion, we performed a high-throughput imaging-based siRNA screen targeting 867 nuclear proteins including chromatin-associated proteins, epigenetic regulators, and nuclear envelope components. Using multiple morphometric parameters and eliminating cell cycle effectors, we identified a set of novel determinants of nuclear size and shape. Interestingly, most identified factors altered nuclear morphology without affecting the levels of lamin proteins, which are known prominent regulators of nuclear shape. In contrast, a major group of nuclear shape regulators were modifiers of repressive heterochromatin. Biochemical and molecular analysis uncovered a direct physical interaction of histone H3 with lamin A mediated via combinatorial histone modifications. Furthermore, disease-causing lamin A mutations that result in disruption of nuclear shape inhibited lamin A-histone H3 interactions. Finally, oncogenic histone H3.3 mutants defective for H3K27 methylation resulted in nuclear morphology abnormalities. Altogether, our results represent a systematic exploration of cellular factors involved in determining nuclear morphology and they identify the interaction of lamin A with histone H3 as an important contributor to nuclear morphology in human cells.

    1. Biochemistry and Chemical Biology
    2. Chromosomes and Gene Expression
    Yunfeng Lin, Jia Li ... Shan Yan
    Research Article

    Cells have evolved the DNA damage response (DDR) pathways in response to DNA replication stress or DNA damage. In the ATR-Chk1 DDR pathway, it has been proposed that ATR is recruited to RPA-coated single-stranded DNA (ssDNA) by direct ATRIP-RPA interaction. However, it remains elusive how ATRIP is recruited to ssDNA in an RPA-independent manner. Here, we provide evidence that APE1 directly associates ssDNA to recruit ATRIP onto ssDNA in an RPA-independent fashion. The N-terminal motif within APE1 is required and sufficient for the APE1-ATRIP interaction in vitro and the distinct APE1-ATRIP interaction is required for ATRIP recruitment to ssDNA and the ATR-Chk1 DDR pathway activation in Xenopus egg extracts. In addition, APE1 directly associates with RPA70 and RPA32 via two distinct motifs. Taken together, our evidence suggests that APE1 recruits ATRIP onto ssDNA in an RPA-dependent and -independent manner in the ATR DDR pathway.