Introduction
It is well known that manipulation of embryos in vitro has significant effects on blastocyst gene expression, cell number, growth velocity and potential for implantation.Reference Feuer, Liu and Donjacour 1 By the blastocyst stage, the embryo is constituted by totipotent cells that have differentiated into only two cell types, the inner cell mass and trophectoderm, and only a percentage of inner cell mass cells will contribute to the developing fetus. It is therefore fascinating that an environmental stress occurring within the earliest stages in development can elicit postnatal phenotypes, including some that will not manifest physiologically until much later in adulthood. In mice, exposing the fertilized zygote to suboptimal nutrient conditions in vitro for merely 10 h is sufficient to alter birth weight and postnatal growth.Reference Banrezes, Sainte-Beuve and Canon 2 Other models have shown that mice produced by in vitro fertilization (IVF) exhibit normal physiology at a younger age, but develop glucose intolerance and significantly more fat mass as adults.Reference Feuer, Liu and Donjacour 3 As such, one of the most relevant investigations within the Developmental Origins of Health and Disease field is to more completely understand the mechanism(s) involved in reprogramming and its persistence throughout development. Although epigenetic alterationsReference Reik 4 are likely responsible for maintaining the molecular memory of the ‘environmental insult’ the exact nature of these alterations is only just emerging.
Uncertainty over the long-term health of IVF children has wide clinical relevance,Reference Servick 5 as more than 6.5 million babies have been born following IVF (Dr Fernando Zegers, 30 September 2016, personal communication). Assisted reproductive technology (ART) has been associated with an increased risk for many pregnancy-related complications,Reference Feuer, Camarano and Rinaudo 6 including incidence of congenital malformations, imprinting disorders, preterm birth and low birth weight (a marker of both in utero stress and predisposition to hypertension and diabetes). Longer-term studies evaluating IVF children’s health are limited in number and length of follow-up. Although overall the data are reassuring, there are lingering questions about the effects of IVF on long-term cognitive function and metabolic health. From a metabolic standpoint, ART children display subtle, yet significant changes in blood pressure,Reference Sakka, Loutradis and Kanaka-Gantenbein 7 , Reference Scherrer, Rimoldi and Rexhaj 8 circulating glucose levelsReference Ceelen, van Weissenbruch, Vermeiden, van Leeuwen and Delemarre-van de Waal 9 and fat deposition.Reference Ceelen 10 – Reference Ceelen, van Weissenbruch, Vermeiden, van Leeuwen and Delemarre-van de Waal 12 One study reported that after a 3-day energy-balanced diet, peripheral insulin sensitivity (measured by a hyperinsulinemic-euglycemic clamp) is lower in IVF young adults compared with naturally conceived controls; conversely a 3-day overfeeding diet results in higher systolic blood pressure in IVF-conceived participants.Reference Chen, Wu and Zhao 13
The mouse model has been particularly advantageous for addressing how IVF might affect health into adulthood, given the short generation time and the ease of manipulating mouse embryos. Furthermore, human and mouse embryos share several milestones in development and nutritional requirements.Reference Quinn and Horstman 14 At present, multiple groups have demonstrated that mice produced by IVF exhibit metabolic and cardiovascular abnormalities compared with control animals,Reference Feuer, Liu and Donjacour 3 , Reference Chen, Wu and Zhao 13 , Reference Donjacour, Liu, Lin, Simbulan and Rinaudo 15 – Reference Tan, Wang, Zhang, An and Tian 19 although the phenotypic manifestations vary. In this review, we summarize the current understanding of the postnatal molecular effects of IVF in the mouse model and discuss the hypothesized explanations.
IVF phenotypes are complex
From a historical perspective, investigators first explored whether culturing embryos in vitro produced behavioral abnormalities in the subsequent offspring.Reference Ecker, Stein and Xu 20 , Reference Fernandez-Gonzalez, Moreira and Bilbao 21 These studies showed that mice derived from cultured embryos presented only minor alterations in preweaning behavioral milestones, but as adults exhibited alterations in anxiety, locomotor activity, spatial and implicit memories.Reference Strata, Giritharan and Sebastiano 22 These original experiments inspired investigators to further probe the postnatal effects of IVF, which revealed an impact of in vitro conditions on growth and metabolic physiology.
There are a multitude of phenotypes that have been associated with conception using IVF. There is conflicting evidence over whether IVF affects longevity: one group found no differences in mice fed a standard diet,Reference Sommovilla, Bilker, Abel and Schultz 23 but a different study showed that male IVF mice challenged with a high-fat died exhibit a roughly 25% shorter life span compared with control animals.Reference Rexhaj, Paoloni-Giacobino and Rimoldi 16 In vitro culture can affect organ size, as well as the development of pneumonia, steatosis and kidney inflammation pathologies.Reference Fernandez-Gonzalez, Moreira and Bilbao 21 Embryo culture can also impact blood pressureReference Donjacour, Liu, Lin, Simbulan and Rinaudo 15 , Reference Rexhaj, Paoloni-Giacobino and Rimoldi 16 , Reference Watkins, Platt and Papenbrock 24 and fat deposition.Reference Feuer, Liu and Donjacour 3 Changes in glucose metabolism are frequently observed following IVF, including increased fasting glucose, hyperinsulinemia, glucose intolerance and peripheral insulin resistance.Reference Feuer, Liu and Donjacour 3 , Reference Donjacour, Liu, Lin, Simbulan and Rinaudo 15 , Reference Scott, Yamazaki and Yamamoto 18 , Reference Chen, Wu and Zhao 25 However, there are significant elements that add to the complexity of these phenomena: several phenotypes linked to IVF are condition-specific, strain-specific, sexually dimorphic and may not emerge until later into adulthood.
Condition specificity
The complex processes of IVF and embryo culture introduce a myriad of environmental exposures that deviate from natural conception in vivo, including physical manipulation of gametes and embryos, fluctuations in oxygen concentration, potential deviations in pH, contact with stiffer environments, light exposure, culture medium composition or use of sequential media. Importantly, culture medium composition is not clinically standardized. Different amounts of nutrients (glucose, pyruvate, glutamine, other amino acids, etc.) and possibly the lack of growth factors and cytokines that are present in vivo could play a significant and stress-inducing role in embryonic growth and development.Reference Chronopoulou and Harper 26 – Reference Lane and Gardner 28
Several groups have documented IVF condition-specific effects on embryo physiology and transcription, with more severe preimplantation stress precipitating more deviant phenotypes.Reference Feuer, Liu and Donjacour 1 , Reference Giritharan, Li and De Sebastiano 29 – Reference Rinaudo, Giritharan, Talbi, Dobson and Schultz 32 These effects persist postnatally, and more stressful IVF culture and oxygen conditions have a greater impact on mouse growthReference Banrezes, Sainte-Beuve and Canon 2 and glucose tolerance.Reference Feuer, Liu and Donjacour 3 , Reference Donjacour, Liu, Lin, Simbulan and Rinaudo 15 For example, offspring exposed to suboptimal in vitro conditions [Whitten’s medium and 20% O2 (IVFWM group)] have higher body weights and exhibit glucose intolerance at 19 weeks compared with control mice (in vivo-derived blastocysts transferred to surrogate damns, e.g. ‘flushed blastocyst’ control group); whereas optimized culture conditions [KSOM medium with amino acids and 5% O2 (IVFKAA)] result in normal postnatal growth and glucose homeostasis.Reference Donjacour, Liu, Lin, Simbulan and Rinaudo 15 Interestingly, in spite of the well-described condition-specific effects of culture medium composition and oxygen tension on embryo gene expression,Reference Giritharan, Talbi and Donjacour 30 , Reference Rinaudo, Giritharan, Talbi, Dobson and Schultz 32 there is a shared, concortantly altered 18-gene and nine-pathway signature of preimplantation embryo manipulation that is present in all varieties of in vitro embryo, irrespective of culture environment or method of fertilization.Reference Feuer, Liu and Donjacour 1 The in vitro fingerprint is consistent with an increase in oxidative stress (decreased levels of Hif2α as well as the altered expression of the ubiquitination and protein turnover genes Ube2a, Ube2l3, Lap3, Socs3 and Arrdc4) and changes in placental development (repression of placental growth and/or trophoblast stem cell function genes Eomes, Fgf4, Socs3 and Dusp9).Reference Feuer, Liu and Donjacour 1
Strain specificity
Phenotypes are influenced by genetics and can vary by mouse strain. The same IVFWM and IVFKAA model performed in an inbred C57Bl/6J modelReference Feuer, Liu and Donjacour 3 precipitates more significant changes to glucose metabolism than in outbred CF1×B6D2F1 IVF mice.Reference Donjacour, Liu, Lin, Simbulan and Rinaudo 15 Furthermore, metabolic pathologies present in certain strains of IVF mice may not be observed in others: FVB mice exhibit a hypertensive phenotype following IVF,Reference Rexhaj, Paoloni-Giacobino and Rimoldi 16 whereas outbred CF1×B6D2F1 animals display lower systolic blood pressure.Reference Donjacour, Liu, Lin, Simbulan and Rinaudo 15 Phenotypic variation by strain could be attributed to differences in both genetic code and regulatory mechanisms, including epigenetics. Given that the epigenetic landscape can depend on genetic background, strain is a critical variable in how embryos might differentially respond to similar culture conditions and display different postnatal phenotypes. Other important factors that can contribute to strain effects include potential heterozygote advantage and parental origin effects, which becomes challenging to uncouple for complex phenotypes.Reference Han, Mtango, Patel, Sapienza and Latham 33
Sexual dimorphism
There is also considerable sexual dimorphism in adult IVF phenotypes.Reference Feuer, Liu and Donjacour 3 , Reference Chen, Wu and Zhao 13 , Reference Rexhaj, Paoloni-Giacobino and Rimoldi 16 , Reference Feuer, Donjacour and Simbulan 34 For example, C57Bl/6J IVFKAA females exhibit increased weight and fat deposition, evidence of insulin resistance (increased fasting glucose) and β-cell dysfunction, whereas growth and metabolic physiology is normal in males.Reference Feuer, Liu and Donjacour 3 More significant changes to glucose tolerance have also been described for B6C3F1 females compared with males,Reference Scott, Yamazaki and Yamamoto 18 but also for CF1×B6D2F1 males compared with females,Reference Donjacour, Liu, Lin, Simbulan and Rinaudo 15 further complicating the postnatal IVF landscape. The sexual dimorphism is particularly captivating, as the preimplantation embryo does not appear to produce any gonadal hormones. However, a significant number of genes are differentially expressed between male and female blastocysts (nearly 600 transcripts in mouse and up to one-third of transcripts in bovine),Reference Bermejo-Alvarez, Rizos, Rath, Lonergan and Gutierrez-Adan 35 , Reference Kobayashi, Isotani and Mise 36 indicating that the preimplantation embryo may already be primed to differentially respond to environmental cues by sex.
Latent phenotypes
Some effects of IVF are unmasked by time. As was described above, embryo culture had only minor effects on preweaning behavioral milestones, but adult mice develop changes in anxiety, locomotor activity, spatial and implicit memories.Reference Strata, Giritharan and Sebastiano 22 β-cell dysfunction can also increase with age: our group has shown that glucose-stimulated insulin secretion is normal in younger (12 week) females, but at 29 weeks the IVF islets exhibit higher baseline insulin secretion and do not respond with an appropriate increase in insulin secretion following high glucose exposure.Reference Feuer, Liu and Donjacour 3 This is particularly significant and underlies the importance of longitudinal studies in capturing IVF phenotypes.
Overall, these factors demonstrate the challenges and complexities associated with describing the long-term effects of embryo manipulation, and what its implications for health might hold.
The physiological and molecular effects of IVF on adult tissues
Many groups have reported a range of postnatal metabolic and molecular phenotypes associated with IVF and embryo culture. But given the discrepancies in strain and condition as was described above, it is difficult to link multiple studies into one blanket explanation of how in vitro exposures during preimplantation development affect adult phenotypes. One possibility is to choose one model of IVF and perform untargeted molecular studies in multiple adult tissues, to obtain a more comprehensive snapshot of the long-term effects. To that end, our group has used microarray and metabolomics technologies to interrogate the molecular IVF signatures associated with IVFKAA in females. We used a multi-omics approach because microarray alone does not always capture all gene transcripts or tissue/embryonic isoforms, and metabolomics provides a real-time snapshot of cellular processes and the potential outcomes of transcriptional alterations. In addition, females were selected for their glucose-intolerant phenotype and IVFKAA conditions because they are optimal for mouse embryo developmentReference Perin, Maluf, Nicolosi Foltran Januario and Nascimento Saldiva 37 and more closely resemble the conditions used in human IVF, thereby increasing the likelihood of positive findings and the translational values of the studies, respectively.Reference Wiemer, Anderson, Kyslinger and Weikert 38 In the following section, we present the adult, tissue-specific physiologic and molecular signatures of IVF described in our studies and others’ utilizing the rodent model.
Liver
IVF and in vitro culture have been linked to adult liver weight and the development of liver steatosis in a condition-specific manner.Reference Fernandez-Gonzalez, Moreira and Bilbao 21 Our group found transcriptional evidence of changes to bile acid metabolism, protein ubiquitination, mitochondrial dysfunction and oxidative phosphorylation, as well as gene misexpression that might reflect differences in the cellular response to therapeutics.Reference Feuer, Liu and Donjacour 3 Each of these transcriptional changes were corroborated by metabolomic data showing evidence of altered proteolysis and protein ubiquitination (significant increases in dipeptide concentration), changes in AMP-activated protein kinase (AMPK), adenosine triphosphate (ATP)- and guanosine triphosphate (GTP)- requiring reactions (increased purine metabolites such as adenosine monophosphate (AMP), guanosine monophosphate (GMP), and guanosine) and reduced bile acid pool size. Bile acids have a reciprocal relationship with both glucose and insulin,Reference Li, Francl and Boehme 39 and impaired bile acid synthesis (and subsequent reductions in pool size) significantly decreases energy expenditure and contributes to the pathogenesis of obesity and diabetes.Reference Watanabe, Horai and Houten 40 In addition, we showed that IVF livers displayed a decrease in glucose metabolites with a (potentially compensatory) increase in glycogen intermediates, decreased glutamine and depleted long-chain fatty acids, reflecting altered mitochondrial function. In a separate study, lipidomic profiling of IVFKAA and naturally conceived fetal livers showed an increase in several phospholipids related to glucose metabolism and vascular dysfunction – including lysophosphatidylcholine, phosphatidic acid and lysophosphatidylethanolamine – which might further contribute to the postnatal phenotypes associated with IVF.Reference Li, Xiao and Chen 41
Muscle
Between liver, muscle, adipose tissue and pancreatic islets, microarray analysis revealed that skeletal muscle harbored the most extensive gene misexpression associated with IVF (over 2000 altered genes). However, in spite of the large quantity the changes were modest (only three genes were altered more than two-fold different from controls).Reference Feuer, Liu and Donjacour 3 Over 35 misexpressed genes reflected an upregulation in glucocorticoid receptor (GR) signaling, which is notable because inappropriate glucocorticoid signaling can affect growth, susceptibility to chronic diseases and induction of the stress response. Elevated glucocorticoid exposure is also correlated with predisposition to cardiovascular disease and diabetes.Reference Jaddoe and Witteman 42 Other pathways enriched in the IVF transcriptional signature included those involved in cell–cell communication and movement, as well as nitric oxide (NO) and reactive oxygen species (ROS) production.
Adipose tissue
Our group has shown that IVFKAA females are predisposed to significantly increased fat deposition in adulthood, with over 1000 genes misexpressed in gonadal adipose tissue at 29 weeks of age compared with control females.Reference Feuer, Liu and Donjacour 3 The transcriptional changes overrepresented genes involved in lipid metabolism, energy production such as gluconeogenesis and the tricarboxylic acid (TCA) cycle, as well as insulin receptor signaling, interleukin 1 (IL1) signaling, and activation of retinoid X receptor / liver X receptor (RXR/LXR) and retinoid X receptor / farnesoid X receptor (RXR/FXR) heterodimers. Metabolomic investigation of male and female gonadal fat revealed female-specific increases in pro-inflammatory and ROS metabolites, which is particularly relevant as adipogenesis can be accelerated by oxidizing conditions.Reference Feuer, Donjacour and Simbulan 34 , Reference Imhoff and Hansen 43 , Reference Vigilanza, Aquilano, Baldelli, Rotilio and Ciriolo 44 Because fat is a primary driver of metabolic dysfunction, it is possible that the acquired sex bias in IVF adipose tissue contributes to the sex-specific metabolic phenotypes.
Interestingly, both male and female IVFKAA mice retain fat-exclusive maintenance of transcriptional changes present in IVF blastocysts compared with other adult tissues, including increased expression of the GRReference Simbulan, Liu and Feuer 45 and thioredoxin-interacting protein (Txnip, discussed below).Reference Feuer, Liu and Donjacour 3 These data suggest that adipose tissue may be a locus of sex- and tissue-specific changes associated with IVF.
Pancreatic islets
IVF is frequently associated with higher fasting glucose levels and glucose intolerance in adulthood. Our IVFKAA animals display a lower insulinogenic index – a measure of islet function, calculated by dividing ∆insulin by ∆glucose at 30 min of a glucose tolerance testReference Sone and Kagawa 46 – indicating a reduced ability of IVF mice to secrete insulin in response to glucose, possibly due to β-cell insensitivity to glucose.Reference Donjacour, Liu, Lin, Simbulan and Rinaudo 15 Further investigation showed that IVF females have lower weight-standardized pancreatic weights at 29 weeks of age compared with controls. Microarray analysis of pancreatic islets yielded fewer changes in gene expression than other insulin-sensitive tissues, perhaps because islets are a highly differentiated tissue with many cell types. The most notable change was a significant downregulation in Gapdh expression. Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) coordinates several switches between metabolic function, redox state and apoptosis to orchestrate regulation of and recovery from different insults or stressors.Reference Sirover 47 Furthermore, GAPDH is decreased in patients with diabetes, and GAPDH inhibition following an increase in ROS is the key pathway behind diabetes-induced cellular damage in peripheral tissues and in β-cells.Reference Brownlee 48 The islet data contained other evidence of oxidative stress – including an enrichment in nicotinamide adenine dinucleotide (NADH) repair pathways, oxidized GTP and dGTP detoxification and free radical scavenging – providing one potential mechanism underlying the ROS and glucose-intolerant phenotypes.
Brain
As described above, several groups have reported changes in behavior and anxiety levels in mice conceived by IVF.Reference Fernandez-Gonzalez, Moreira and Bilbao 21 Specifically, IVF mice spend more time in the open arms of an elevated plus maze, reflecting a reduction in anxiety levels. These behavioral differences are particularly evident in males, and are paralleled by altered expression of the genes monoamine oxidase A (MAO-A), corticotropin-releasing factor and γ-aminobutyric acid (GABA) in anxiety-related cortical and subcortical structures and in the cerebellum.Reference Strata, Giritharan and Sebastiano 22 These three genes are very well-described modulators of anxiety and fear behaviors;Reference Millan 49 altered expression of these and related receptors and transporters are correlated with changes in anxiety levels.Reference van Ameringen, Mancini, Farvolden and Oakman 50 Strata et al.Reference Strata, Giritharan and Sebastiano 22 postulated that reductions in hypothalamic MAO-A expression, as well as cortical and hypothalamic expression of GABA receptors and transporter genes in IVF mice might underlie their disinhibited exploratory behaviors.
Vascular tissue
Rexhaj et al. Reference Rexhaj, Paoloni-Giacobino and Rimoldi 16 observed endothelial dysfunction and increased stiffness in the vasculature of FVB male mice generated by IVF, with arterial hypertension that persisted transgenerationally in male offspring. Transgenerational inheritance of the vascular phenotype suggests an epigenetic-based mechanism, but this requires confirmation. The authors performed targeted analyses and found changes in methylation at the promoter of the endothelial nitric oxide synthase (eNOS) gene in the aorta, with related decreased vascular eNOS expression and NO synthesis. A different evaluation of blood pressure similarly found an effect in IVF males, but these CF1×B6D2F1 mice derived from stressful IVF conditions (IVFWM) displayed lower systolic blood pressure and equal diastolic blood pressure compared with in vivo controls.Reference Donjacour, Liu, Lin, Simbulan and Rinaudo 15 Conversely, another group documented increases in systolic blood pressure associated with embryo culture.Reference Watkins, Platt and Papenbrock 24 These discrepancies in results are likely complex, and may be secondary to differences in mouse strain, blood pressure methodology and experimental design (e.g. Rexhaj et al. did not control for litter size).
Heart
Although outbred IVFWM male mice had lower blood pressure than control mice, the end systolic and end diastolic volumes measured by cardiac echocardiography were increased.Reference Donjacour, Liu, Lin, Simbulan and Rinaudo 15 Echocardiographic analysis showed left ventricular hypertrophy, which has also been reported in IVF and intracytoplasmic sperm injection (ICSI) infants.Reference Valenzuela-Alcaraz, Crispi and Bijnens 51 Frozen cardiac weights were heavier in IVF mice, but this did not reach statistical significance. Another group has shown that culture conditions can differentially affect postnatal heart weights.Reference Fernandez-Gonzalez, Moreira and Bilbao 21 Microarray analysis identified misexpression in 1361 genes, but only 16 more than two-fold – among these were Ccl7, Ccl19, Chd7, Cnnm2, Il17ra, Il1rb, Gfpt2 and Ngf. These genes are overrepresented in pathways involved in the inflammatory response, cellular assembly and organization, cancer, as well as cell-to-cell signaling, interaction and movement.Reference Feuer, Liu and Donjacour 1
Of note, similar results have been reported following analysis of cardiac function in children conceived spontaneously or by ART.Reference Zhou, Liu and Gu 52 Although the majority of cardiac geometric morphology parameters are comparable among the study groups, IVF offspring have significant increases in left ventricular relative wall thickness, left ventricular mass index and left ventricular remodeling index after adjustment for age, gender, body surface area and heart rate.
Summary
To summarize, analysis of individual tissues and organs from IVF-generated mice reveal modest, tissue-specific changes and the absence of a common transcriptional fingerprint, in spite of originating from common IVF culture conditions. Tissue and serum metabolomics data similarly confirm the tissue specificity. Together, the most notable findings point to multi-omic and system-wide evidence of oxidative stress following conception by IVF, possibly a ‘fossil record’ of higher oxidative exposure in the embryo. Of note, we have been unable to identify any human studies that investigate the transcriptional effects of IVF in tissues or organs of ART children.
Hypothesis
There are several potential mechanisms – or a combination thereof – by which an environmental stress occurring early in development might affect postnatal and long-term health.
The initial trigger
Given the complexity of the process of IVF, multiple, combinatorial factors – physical, nutritional, oxidative – are likely responsible for inducing the initial molecular changes present in embryos generated in vitro.Reference Feuer and Paolo F. 27 , Reference Lane and Gardner 28
Exposure to this artificial environment will alter multiple transcriptional and metabolic processes in the embryo. One immediate consequence will be a change to the embryo oxidative state, which could subsequently provoke a reprogramming response. There is abundant evidence that embryo culture induces an increase in ROS.Reference Goto, Noda, Mori and Nakano 53 – Reference Cebral, Carrasco, Vantman and Smith 55 In a recent analysis of blastocyst gene expression following exposure to different culture conditions and methods of assisted fertilization, we reported significant changes in genes governed by the transcriptional regulators Hif2α, NFE2L2, CEBPs,Reference Manea, Todirita, Raicu and Manea 56 p53 and GADD45. These or similar factors are capable of inducing epigenetic remodeling that persists throughout development, which through combinatorial interplay with other transcriptional modulators could lead to unique and tissue-specific gene expression signatures in adulthood.Reference Feuer, Liu and Donjacour 1 It is therefore possible that increased mitochondrial stress – and resulting increase in ROS levels – in response to an abnormal preimplantation environment is the initial stimulus that reprograms the embryo. This hypothesis would also explain the metabolomic and transcriptional evidence of altered redox axes and oxidative stress we found in different postnatal insulin-sensitive IVF tissues.Reference Feuer, Liu and Donjacour 3 , Reference Feuer, Donjacour and Simbulan 34
In addition, because the preimplantation embryo undergoes substantial epigenetic remodeling between the zygote and postimplantation stages,Reference Reik 4 , Reference Biechele, Lin, Rinaudo and Ramalho-Santos 57 it is possible that the embryo reaction to this artificial environment results in gene expression changes that are incorporated into the developmental and metabolic program simply because they occur at an epigenetically vulnerable time. There is evidence to both support and refute this hypothesis. Because different in vitro conditions lead to unique changes to blastocyst expression,Reference Feuer, Liu and Donjacour 1 any permanency associated with the acute effects could lead to IVF phenotypes even more varied than currently reported. However, there are also ubiquitous transcriptional changes present in IVF-derived blastocysts in spite of significant variations in culture condition (including oxygen tension, culture medium composition and even the use of ICSI v. traditional IVF). Epigenetic incorporating of these particular changes could underlie the relatively similar postnatal metabolic phenotypes (glucose intolerance, blood pressure alterations).
Consequences
In order to retain the molecular memory of an embryo’s acute exposure to the in vitro environment, several long-term molecular consequences need to occur. We can envision multiple circumstances that explain the postnatal IVF outcomes.
Master regulator gene hypothesis
One possibility is that in vitro manipulation alters the preimplantation expression of one or a selected number of key genes with the potential to propagate the observed phenotypes. These ‘master regulator’ genes would be modified epigenetically such that the alterations persist in the developing individual and govern changes to postnatal metabolic homeostasis. This possibility is appealing for its simplicity and because it has been described in other biological contexts. For example, strategic overexpression of a few critical genes such as the Yamanaka factors can reprogram one cell type into another, including de-differentiation into a pluripotent state.Reference Takahashi and Yamanaka 58
We have investigated this hypothesis in the mouse by analyzing two candidate genes that demonstrate significant blastocyst transcriptional modification after in vitro exposure compared with control embryos: Txnip and the GR Nr3c1.Reference Feuer, Liu and Donjacour 3 , Reference Simbulan, Liu and Feuer 45 We selected Txnip because the literature suggested it might function as a sensitive marker of preimplantation stress. TXNIP operates at the intersection of multiple cellular stress, redox and metabolic pathways: it is the principal inhibitor of the antioxidant thioredoxin; it regulates the PI3K/AKT pathway and is involved in peripheral glucose homeostasis; it regulates β-cell mass by decreasing mitochondrial apoptosis. Not surprisingly, pathological changes in Txnip/TXNIP level or activity are associated with diabetes, apoptosis and oxidative stress. Txnip is regulated by glucose, glutamine, adenosine and radical oxygen species, thereby linking cellular nutritional and redox states.Reference Hui, Andres and Miller 59 – Reference Yu, Goh, Dai and Luo 61 We reported that Txnip/TXNIP expression and protein levels were significantly elevated in IVF embryos, and selectively increased in female IVF fat and muscle but not liver. This tissue-specific dysregulation was associated with increased H4 acetylation at the Txnip promoter,Reference Feuer, Liu and Donjacour 3 indicating that the acute effects of in vitro manipulation can persist throughout development in a locus-specific manner. A valuable experiment to further examine the ‘master regulator gene hypothesis’ would be to manipulate Txnip/TXNIP activity in the preimplantation embryo and determine the effects on adult glucose metabolism.
Similarly, global microarray analysis identified a four-fold upregulation of Nr3c1 in mouse blastocysts conceived by IVF and cultured in suboptimal conditions (IVFWM). Because IVFWM alters postnatal growth trajectory, impairs glucose tolerance and predisposes a male-specific borderline increase in corticosterone levels,Reference Feuer, Liu and Donjacour 3 , Reference Donjacour, Liu, Lin, Simbulan and Rinaudo 15 we investigated whether GR levels might also be altered in adult IVF insulin-sensitive tissues. GR expression was increased in both IVF blastocysts and offspring adipose tissue in males, and confirmed in both inbred (C57Bl/6J) and outbred (CF1×B6D2F1/J) mouse models.Reference Simbulan, Liu and Feuer 45
The variability and tissue specificity of these results indicate that if master regulator genes are indeed present, their cellular framework is elusive. And to confirm the presence of such a regulator, its activity and downstream pathways would have to be further manipulated or normalized in embryos with subsequent evaluation of phenotypes in adulthood – experiments that are both costly and time consuming. Moreover, the many transcriptional changes documented in adult IVF tissues could occur secondary to the altered function of key upstream transcriptional regulators, which would not necessarily be identifiable in a microarray profile.Reference Feuer, Liu and Donjacour 1 Transcriptional regulators are potentially excellent candidates for regulating this reprogramming, because their combinatorial interplay with each other and with other transcriptional modulators can act on different tissues in unique ways. In this manner, altering the function of a transcriptional regulator in the preimplantation embryo could exert a variety of different effects depending upon the cell type, which could explain the tissue specificity observed in the ART expression profiles.
It is also possible that the ‘master regulator’ umbrella extends beyond genes and transcription, and includes other candidates such as epigenetic control regions (ECRs) or regulatory RNAs. Indeed, different ECRs can regulate gene expression in a tissue-specific manner,Reference Skinner, Manikkam, Haque, Zhang and Savenkova 62 and microRNAs have well-described roles in development and cellular differentiation.
Global gene hypothesis
A different possibility is that the acute transcriptional changes that follow preimplantation stress are epigenetically maintained at multiple loci, and a large number of genes are subsequently altered in adult tissues. Because each tissue has a unique developmental program, the same gene could be differentially affected by sex and tissue type, such that only a subgroup of the preimplantation IVF fingerprint remains altered in different adult tissues. This hypothesis would predict that a more universal transcriptional signature is present. Our initial data contradict this view.Reference Feuer, Liu and Donjacour 1 Although we observed a specific 18-gene signature of broad in vitro exposure in the preimplantation embryo, a comparable signature was not present in adult tissues. Although sexual dimorphism might influence these signatures (the pooled blastocysts were not separated by sex), there was no transcriptional concordance between IVF embryos and any individual adult tissue in offspring derived from the same conditions.
It is however possible that the in vitro environment affects several genetic loci that are not immediately required for embryonic development (and as such, would not necessarily be identified by microarray analysis in blastocysts), but are important in postimplantation differentiation and/or postnatal metabolic homeostasis. These loci could contribute to the postnatal transcriptional phenotypes and explain the discordant changes between embryos and adults.
Separately, because there can be great strength in combinatorial changes, it is also possible that the physiological and metabolic phenotypes are derived from transcriptional changes that – while numerous – are so modest in magnitude that they are missed by a standard microarray analysis. This hypothesis requires a more sensitive and comprehensive technology – perhaps also including microRNA – to address the problem.
Conclusion
Until the recent past, it was unclear whether IVF could alter health and metabolism across the offspring’s life course. Since then, we and othersReference Feuer, Liu and Donjacour 3 , Reference Chen, Wu and Zhao 13 , Reference Donjacour, Liu, Lin, Simbulan and Rinaudo 15 – Reference Tan, Wang, Zhang, An and Tian 19 have demonstrated that metabolic and cardiovascular abnormalities are indeed present in IVF mice, which are beginning to be validated in humans. Molecular analyses of adult tissues from IVF mice have provided some insight on the molecular alterations that may function as the substrate responsible for the phenotypic observations. Overall, it would seem that unspecific oxidative changes are ubiquitously present in postnatal IVF tissues. We can therefore envision a cardinal role for mitochondria in the reprogramming mechanism, by which preimplantation embryo manipulation induces mitochondrial increases in ROS production secondary to inadequate nutritional or environmental cues. Consequently, an ‘oxidative emergency’ response is initiated in the IVF embryo that can persist throughout development and endure in adult tissues. Evaluating these hypotheses should be a principal focus of further studies, which warrants additional resources and funding. These results have a clear and important translational value, given the ever-increasing number of IVF children in the world.
Acknowledgments
This work was supported by R01: HD 082039. SKF was supported by the UCSF California Preterm Birth Initiative (PTBi-CA).
