The TP53 fertility network.

The p53 protein and its signal transduction pathway are composed of a set of genes and their protein products that are designed to respond to a wide variety of intrinsic and extrinsic stress signals. Although the important interaction between p53 and LIF is crucial for embryo implantation, current evidence suggests that not only LIF , but other genes may be important in the reproductive stages of decidualization and implantation. To further test this hypothesis, we constructed a network with 18 TP53 related genes involved in decidualization and implantation processes, including LIF , MDM2 and others ( Figure 1 ). Genes related to decidualization and implantation were compiled from the Gene Ontology website database using the AmiGO browser. Association among these genes was tested using the STRING 9 software which tests available known and predicted gene/protein interactions ( Szklarczyk et al. , 2011 ). This “two-step” approach was chosen to minimize the possibility of false associations during the STRING analysis due to co-existence of words. Table S2 summarizes the 18 genes of this network, as well as their interconnections and wide range of functions. It is important to note that neither p63 nor p73 are on this list, since their main reproduction-related functions refer to the control of ovulation and female germ cell integrity in humans, and they apparently are not involved in decidualization or embryo implantation stages. This “ TP53 Fertility Network” illustrates the importance of multiple genes in these specific stages of human fertility and opens a wide range of possibilities for genetic variation studies in genes not yet being investigated with regard to fertility. For example, IGFBP7 (insulin-like growth factor binding protein 7) is predominantly expressed in the vasculature of developing embryos and regulates vascular endothelial growth factor-A-dependent neoangiogenesis ( Hooper et al. , 2009 ). ESR1 (estrogen receptor 1) gene is critical for LIF expression ( Feng et al. , 2011 ) and IL1B (interleukin 1, beta) is involved in a variety of cellular activities that are essential for decidualization, including cell proliferation, differentiation, and apoptosis ( Ben-Sasson et al. , 2009 ). An additional analysis to verify if these 18 genes belong to a specific functional cluster was performed using GeneDecks V3 software. Thirteen of them were functionally clustered as having an involvement in the reproductive system ( CYP27B1, ESR1, LIF, MEN1, PLA2G4A, PLAU, PPARD, PTGS2, SOD1, SPP1, TP53, UBE2A and VDR ). More specifically, seven genes ( CALCA, IL1B, LIF, PPARD, PTGS2, SOD1 and SPP1 ) were associated with embryo implantation (p = 1x10 −16 ) and seven ( CYP27B1, LIF, PLA2G4A, PPARD, PTGS2, SPP1 and VDR ) with decidualization (p = 1x10 −16 ). It is noteworthy that some genes are present in all functional clusters cited above ( e.g. PPARD; peroxisome proliferator-activated receptor delta). This analysis brought additional evidence of the role of these genes in key stages of fertility.

Considering the strict regulation of LIF by p53, it is reasonable to expect that modulation of p53 function by single nucleotide polymorphisms (SNPs) in TP53 and TP53 -related genes may affect fertility. In humans there are many naturally occurring SNPs in genes at critical nodes in the TP53 pathway, including TP53 , MDM2 , MDM4 , and USP7 , all of which have known functional variants that can modify the levels or activity of the p53 protein ( Atwal et al. , 2009 ; Bond et al ., 2004 ). One of the most commonly studied TP53 variants, the non-silent polymorphism Pro72Arg (rs1042522; C/G), is associated with biochemical and functional differences in protein functions, since the protein carrying the Pro72 allele is more efficient in initiating senescence and cell cycle arrest, while the one with the Arg72 allele is more active in inducing apoptosis and suppressing cellular transformation ( Dumont et al. , 2003 ; Thomas et al. , 1999 ). The Pro72 isoform is also observed in other primates, including the chimpanzee, while the Arg72 one is only present in humans, thus suggesting that the C (Pro72) allele may correspond to the ancestral allele. Compared with TP53 Pro72, the Arg72 allele presents higher transcriptional activity toward a subset of p53 target genes, including LIF . The induction of LIF is over 2-fold higher in cells with the Arg72 allele than in cells with the Pro72 allele ( Kang et al. , 2009 ), leading to decreased implantation success. Kay et al . (2006) associated the Pro72 allele with recurrent implantation failure and demonstrated that the Pro72 allele is enriched in women with unexplained infertility from an in vitro fertilization clinic, compared with a fertile control population ( Kang et al. , 2009 ). In Brazil, Ribeiro Junior et al. (2009) associated the Pro72 allele with intense pain in a cohort of endometriotic patients and Bianco et al . (2011) considered that the Pro72Arg polymorphism was not a risk factor for infertility or endometriosis in Brazilian infertile patients. Interestingly, we found that both TP53 Pro72Arg and MDM4 rs1563828 are associated with twinning in Cândido Godói ( Tagliani-Ribeiro et al. , 2012 ), a small town in Brazil remarkable for showing a high frequency of both dizygotic and monozygotic twins ( Tagliani-Ribeiro et al ., 2011 ). An additional remarkable fact regarding Pro72Arg is that the allele frequencies for this SNP vary widely across human populations. For instance, Arg72 frequencies range from ∼20% in some Sub-Saharan populations to ∼80% in northern Europeans, while in Asians the values are intermediate (HapMap and Alfred database, respectively). These distinct allele frequencies promote a level of differentiation (F ST ) of 19% between Caucasians and Yoruba from Nigeria ( Table S1 ). Recently, the complete nuclear genomes of two extinct hominids belonging the genus Homo , Homo neanderthalensis and Denisova specimen were published ( Green et al. , 2010 ; Reich et al. , 2010 ). Based on these genomic data sets, compiled from UCSC Genome Browser, only Pro72 is present in both archaic human sequences. Inference from this observation is that the C → G mutation may have a relatively recent origin, i.e. the Arg72 variant may be Homo sapiens -specific. Additional studies on archaic human species will be needed to confirm this hypothesis. The MDM2 SNP309 (rs2279744; T/G) is another commonly described variant that attenuates the TP53 pathway. It is a gain of function variant that increases the affinity of a sequence in MDM2 for the Sp1 transcription factor leading to increased transcription of the Mdm2 protein, and consequent inhibition or attenuation of the TP53 pathway-mediated tumor suppression functions ( Bond et al ., 2004 ). Interestingly, SNP309 is located in a transcriptional enhancer region of MDM2 regulated by estrogen signaling ( Phelps et al. , 2003 ). Because SNP309 increases the binding affinity for Sp1, a co-activator of multiple hormone receptors, it could potentially affect the hormone-dependent regulation of MDM2 transcription and result in further elevation of the Mdm2 protein levels, as estrogen preferentially stimulates transcription of the 309G allele ( Hu et al ., 2007b ). In addition to TP53 Pro72 and MDM2 309G, other variants in TP53 -related genes ( MDM4 , rs1563828: T/C; USP7 , rs1529916: T/C; and LIF , rs929271: G/T) have been proposed as functional variants with a role in reproduction, showing differential allele frequencies in young infertile women submitted to in vitro fertilization when compared to fertile women ( Kang et al ., 2009 ). From an evolutionary perspective, TP53 Arg72 and MDM2 309G seem to have been positively selected in European and Asian populations, which can be interpreted as a result of adaptive pressures during the dispersion of Homo sapiens from Africa to other continents ( Atwal et al. , 2007 ; Shi et al. , 2009 ; Belyi et al. , 2010 ). Several studies indicate that p53 has evolutionarily conserved functions other than acting as a tumor suppressor, and the existence of p53-like proteins in short-lived organisms that do not exhibit adult cancer incidence, such as flies and worms, adds to the argument that tumor suppression was not the original function for p53 and its pathway ( Lu et al. , 2009 ). In addition, the major impact of p53 in cancer prevention or longevity in humans likely occurs in post-reproductive years, which would exclude a major evolutionary role associated to these functions. Like TP53 and MDM2 , MDM4 and USP7 also appear to have alleles or haplotypes that are under selection pressure and show geographic variations in allele distribution ( Atwal et al. , 2007 ; Shi et al. , 2009 ). In a recent study, Feng et al. (2007) reported that SNPs in the TP63 (rs17506395; T/G) and TP73 (rs4648551 G/A and rs6695978 G/A) genes are associated with infertility in women, independently of age for TP63 and specifically in women aged over 35 years for TP73 . The authors proposed that the possible mechanisms of infertility associated with variations in TP53 might be impaired implantation, whereas variations in TP63 and TP73 may affect the quality of oocytes and induce chromosomal aneuploidy ( Feng et al. , 2007 ). Based on these findings, it is reasonable to assume that alleles in genes of the TP53 family and TP53 pathway with reproductive implications may have been important targets for selection pressure during the human evolutionary history.

The current knowledge on p53 regulation and functions has been the subject of a detailed recent review ( Vousden and Prives, 2009 ). However, most studies have concentrated on p53 as a stress-induced tumor suppressor gene, and little is known about its function in normal cellular processes. The p53 protein accomplishes its function by transcriptionally regulating target genes. In 2002, several genomic DNA sequences were detected where the p53 protein was most likely able to bind and activate transcription ( Hoh et al. , 2002 ). Among these, a potential candidate was the gene encoding the leukemia inhibitory factor ( LIF , OMIM 159540), a secreted cytokine that is critical for blastocyst implantation ( Stewart et al. , 1992 ). This gene contains a putative p53-binding consensus DNA sequence in intron 1, which is conserved in both mouse and human gene sequences ( Hu et al. , 2007a ). In fact, implantation cannot occur unless epithelial cells lining the uterus are exposed to LIF ( Stewart et al. , 1992 ), most likely expressed at the onset of implantation, which occurs at day 4 of pregnancy in mice (day 12 in humans). LIF null mice have a defect in maternal reproduction caused by the complete lack of uterine decidualization at the implantation stage, with consequent failure of blastocyst implantation, which can be rescued by LIF injection at the implantation stage (the 4th day of pregnancy in mice) ( Chen et al. , 2000 ). Hu et al . (2007b) demonstrated that p53 plays a significant role in fertility, since p53-null female mice present reduced uterine expression of LIF and, as expected, reduced maternal reproduction due to impaired implantation functions. Administering LIF to p53 deficient female mice at day 4 of pregnancy significantly increased the pregnancy rate and litter size with improved blastocyst implantation. These findings demonstrate that inactivation of p53 decreases the levels and function of uterine LIF, thus indicating a function for p53 in maternal reproduction through the regulation of LIF. In addition, estrogen is also involved in the regulation of transient LIF expression at the implantation stage ( Chen et al. , 2000 ), mediated through its nuclear receptor alpha (ERα, encoded by ESR1 , OMIM 133430). Feng et al . (2011) demonstrated a significant increase in nuclear ERα levels in endometrial glands at the implantation stage in mice, and concluded that the increased expression of LIF at this stage requires the activation of p53, increased estrogen levels, and activated ERα.

In order to explore certain evolutionary aspects of the network we expanded the analysis on its suitability as a model using two different approaches and taking into consideration inter- (vertebrate) and intra- (human) species variations of the 18 genes that comprise the network. The first approach was to assess the level of conservation of the genes included in the network along evolutionary lineages using comparative analysis between humans and other 21 vertebrate species. Data were compiled in the STRING 9.0 database whereas the level of identity of the amino acid sequences between humans and the others species was obtained using the LALIGN software. The 18 genes presented variable levels of amino acid sequence conservation ( Table S3 ). Protein preservation among the species belonging to the primate order (human, chimpanzee, orangutan, and rhesus monkey) was on average 97%, while among placental mammals (human, chimpanzee, orangutan, and rhesus monkey, mouse, rat, guinea pig, rabbit, cow, cat, dog, horse, pig, and armadillo) it was 85%. In contrast, the degree of protein identity decreased significantly (average of 41%) when the comparison involved only humans and fish species. The results generated from STRING 9.0 show that overall 90% of the network’s connections (edges) were retrieved in primates, while for placental mammals the value was reduced to 80%. However, when all vertebrates were considered, only 42% of the network is recovered. These results suggest that some of the 18 network genes may have acquired novel functions in different taxa, throughout vertebrate evolution, in addition to ancestral functions, a similar situation to that reported previously for the HOX family genes ( Chen et al. , 2010 ). The second approach was to study variation within the 18 genes between human populations. The data were compiled from HapMap and ENSEMBL databases. Using this strategy we identified 10,918 polymorphisms, only 1.4% of which being non-synonymous changes ( Table S4 ). Of these, 82 ( e.g. Pro72Arg) are predicted to be deleterious ( Table S5 ). For most other potentially deleterious polymorphisms no striking difference was found in allele frequencies among continental populations. However, some notable exceptions can be highlighted. Reminiscent of the Pro72Arg polymorphism, the rs5241 SNP located in the CALCA gene shows a frequency of 17% in Africans, whereas it is absent in European-descendents. On the other hand, for another SNP (rs2227564) located in the PLAU gene, the rare allele is only present in Euro-Asian populations (23%–33%), whereas it is absent in Africans ( Table S1 ). Overall, these examples suggest that selection pressure for specific alleles in defined populations has affected only a small subset of the genes involved in the proposed network. Since reproduction is central to the evolutionary process, in all vertebrates, as well as in other organisms, the genome is expected to be optimized for reproductive success. However, even among vertebrates there is an immense diversity in how reproduction occurs, including care and rearing of the offspring ( de Magalhaes and Church, 2005 ; Plunkett et al ., 2011 ). There are many reasons in the evolutionary history of each species, including of our own and of other phylogenetically close species, such as Neanderthal, Denisova and chimpanzee that can explain shared and unique reproductive traits. Our results show for instance, that the human fertility network is not identical in all vertebrate species investigated here. An additional complicating factor is that for humans reproductive strategies can have changed drastically due to cultural practices, as well as in response to environmental pressures ( e.g. climate change). Thus, it is expected that part of this diversity is the result of variable selection pressures encountered by human populations as they progressively expanded over the world. Finally, this analysis adds further support to the idea that there is not a unique major effect gene involved in fertility and that an approach based on a wider gene network and taken under an evolutionary perspective may lead to the delineation of a more comprehensive view of the impact of p53 on the complex biology of fertility. Additional investigations at population level, as well as functional studies are needed to clarify the exact implications of the inter- (vertebrate) and intra- (human) differences highlighted in the present study.