Introduction
Dairy cattle bred primarily for production, with relatively little attention for functional traits, are subfertile (Royal et al., Reference Royal, Mann and Flint2000; Roxstrom et al., Reference Roxstrom, Strandberg, Berglund, Emanuelson and Philipsson2001), as shown by suboptimal ovarian cyclicity (Opsomer et al., 2002), behavioural expression of oestrus (Van Eerdenburg, Reference Van Eerdenburg2006), success rate of artificial insemination (Jorritsma and Jorritsma, Reference Jorritsma and Jorritsma2000; Royal et al., Reference Royal, Mann and Flint2000) and embryo survival (Sheldon et al., Reference Sheldon, Wathes and Dobson2006). Subfertility leads to between-calving intervals that are longer than economically optimal (Huirne et al., 2002), it increases replacement rates (Beaudeau et al., Reference Beaudeau, Seegers, Ducrocq, Fourichon and Bareille2000) and, thus, affects profitability, animal welfare and sustainability of animal production. Decreasing profit margins per kg milk pushes farmers to reduce cost, increase herd size and, consequently, minimize labour input and cost of getting cows pregnant. This restricts options to repair dairy cow fertility through additional management and so genetic improvement will become increasingly important (Veerkamp and Beerda, Reference Veerkamp and Beerda2007). Though variation in dairy cow fertility is in part genetic (Flint, Reference Flint2006), heritability for fertility traits is typically low (Pryce and Veerkamp, 2001). This reflects the large unexplainable residual variation in statistical models predicting traits like calving interval and pregnancy rate at the individual cow level and does not mean genetic information is unimportant: daughters of different sires may have differences in mean calving interval of up to 30 days and the pregnancy rates of daughters sired by extreme Holstein bulls are as high as 7% (Weigel, Reference Weigel2006).
It is often perceived that improving fertility by genetic selection is incompatible with the dairy industry’s aim of achieving ever-higher milk production levels. Focussing on increasing genetic merit primarily for yield reduces genetic merit for fertility (Veerkamp et al., Reference Veerkamp, Beerda and van der Lende2003), but conjoint improvement for reproductive performance and milk yield, say maintaining 70% to 80% of the yearly increase in yield (Veerkamp et al., Reference Veerkamp, Oldenbroek, Van Der Gaast and Van Der Werf2000), is possible (Andersen-Ranberg et al., Reference Andersen-Ranberg, Klemetsdal, Heringstad and Steine2005; Jamrozik et al., Reference Jamrozik, Fatehi, Kistemaker and Schaeffer2005). In line with this, most leading dairy cattle breeding programmes incorporate fertility, as derived from calving dates and insemination dates, in their selection indices (Miglior et al., Reference Miglior, Muir and Van Doormaal2005).
Nowadays, detailed information on the Bovine genetic code (i.e. ∼3 billion nucleic base pairs packaged in 29 autosomes and two sex chromosomes plus ∼16 000 mitochondrial base pairs, (Lewin, Reference Lewin2003)) is available. New high-throughput technologies open up ways to effectively exploit genomic information and accelerate genetic improvement. The aim of the present study is to review and discuss relatively new approaches, like genomic selection and functional genomics, as applied in the field of dairy cow fertility.
Genomic selection
Information on variation between animals in nucleotide sequence, as established by micro-satellites, restriction fragment length polymorphisms (RFLPs) and single nucleotide polymorphisms (SNPs), has been utilized in selection via the candidate gene approach and whole genome scan approach (Veerkamp and Beerda, Reference Veerkamp and Beerda2007). Information on the function of a gene in species other than dairy cows may be used for a candidate gene approach: variation in the gene is linked to phenotype, as applied for genes encoding for gonadotropin-releasing hormone (Schneider et al., Reference Schneider, Tomek and Grundker2006), leptin (Liefers et al., Reference Liefers, Veerkamp, Te Pas, Chilliard and Van der Lende2005) and the bovine luteinizing hormone receptor (Hastings et al., Reference Hastings, Donn, Derecka, Flint and Woolliams2006). In the past years, attention has shifted from identifying variation in individual (causal) genes towards genomic selection.
The whole genome scan approach builds on a large number of polymorphic markers and involves the identification of QTL that are associated with specific phenotypic traits. Of the 55 studies with dairy cattle reviewed by Khatkar et al. (Reference Khatkar, Thomson, Tammen and Raadsma2004), 45 reported on QTL for milk production traits and somatic cell score. The few reporting on reproduction traits were about gestation length, dystocia/stillbirth, success/failure per insemination of daughters and (90 days) non-return rate. In addition, dairy cow genetic markers have been linked to ovulation rate, multiple ovulations (Kappes et al., Reference Kappes, Bennett, Keele, Echternkamp, Gregory and Thallman2000) and twinning (Lien et al., Reference Lien, Karlsen, Klemetsdal, Vage, Olsaker, Klungland, Aasland, Heringstad, Ruane and Gomez-Raya2000). The latter two are associated with QTL on bovine chromosome 7 (BTA7) and may be the same as the one validated for its effect on conception rate (Weller et al., Reference Weller, Golik, Reikhav, Domochovsky, Seroussi and Ron2008). The confidence interval for QTL location typically spans hundreds of genes and the identification of the causal genes involved may require fine mapping of the QTL (Meuwissen et al., Reference Meuwissen, Karlsen, Lien, Olsaker and Goddard2002) by linkage disequilibrium (LD) mapping and the construction of higher density marker maps. Several QTL are suspected to be used by breeding companies, although generally these are not published.
Part of phenotypic variation is linked to SNPs, causally or indirectly by LD, and individual animals can be genotyped on the basis of many tens of thousands of SNPs. Once correlations between SNPs and phenotypes have been established, such information allows detailed genotyping in the absence of further phenotypic records, and genomic selection (Meuwissen et al., Reference Meuwissen, Hayes and Goddard2001) has the ability to reduce the cost and increase the effectiveness of breeding programmes. The Dutch breeding company, CRV, announced the implementation of genomic selection in dairy cattle (using 3000 SNPs) in 2006.
Functional genomics: transcriptomics
Functional genomics builds on information on static aspects of the genome (nucleotide sequence) as produced by genome sequencing projects, but focuses on dynamic aspects such as gene transcription (of DNA into mRNA) and translation. Gene expression divergence may result from variation in the sequence of regulatory regions, control of transcriptional initiation, RNA processing and translation (an estimated 60% of genes in humans use alternative splicing), chromatine structure and DNA methylation (Kehrer-Sawatzki and Cooper, Reference Kehrer-Sawatzki and Cooper2007).
With the application of high-throughput technologies, such as oligo or cDNA microarray-based transcriptome analysis, gene functions are studied in a genome-wide fashion, bridging the gap between sequence and biological function. Fan et al. (Reference Fan, Chee and Gunderson2006) reviewed how microarray platforms for parallel genomic analysis were first developed about 15 years ago (Southern et al., Reference Southern, Maskos and Elder1992), resulting some years later in gene expression profiling assays for measuring mRNA abundance (Schena et al., Reference Schena, Shalon, Davis and Brown1995). mRNA levels are assumed to serve as a proxy for phenotypic variation, although mRNA levels need not always reflect the level of functional proteins associated with the gene (Preuss et al., Reference Preuss, Caceres, Oldham and Geschwind2004). Microarray techniques allow quantitative analyses, which is of great relevance as the effects of gene expression levels may be subtle. Niwa et al. (Reference Niwa, Miyazaki and Smith2000) showed that transcriptional regulators do not function as an on–off control system and the precise level of expression of the POU transcription factor Oct-3/4 determines the fate of embryonic stem cells. Differential expression of 50% above or below normal expression levels in undifferentiated stem cells triggers differentiation to endoderm/mesoderm or trophectoderm, respectively.
Gene expression profiling can be combined with QTL mapping (genetical genomics) and the usefulness of the approach has been demonstrated in mice (Schadt et al., Reference Schadt, Monks, Drake, Lusis, Che, Colinayo, Ruff, Milligan, Lamb, Cavet, Linsley, Mao, Stoughton and Friend2003).
Micorarray-based high-throughput technology has great potential in improving dairy cow fertility (Dawson, Reference Dawson2006; Wolf et al., Reference Wolf, Arnold, Bauersachs, Blum, Frohlich, Hiendleder, Mitko and Reichenbach2006) and facilitates progress in a number of fast-developing research areas such as epigenetics and genomic structural variation (see further on).
Recent transcriptomic studies on dairy cow fertility
Microarray-based gene expression profiling in dairy cattle has been discussed recently (Beerda and Veerkamp, 2006) for studies on oocyte maturation (Dalbiès-Tran and Mermillod, Reference Dalbiès-Tran and Mermillod2003; Vallee et al., Reference Vallee, Gravel, Palin, Reghenas, Stothard, Wishart and Sirard2005), non-regressed and regressed corpus luteum tissue (Casey et al., Reference Casey, Morris, Powell, Sreenan and Fitzpatrick2005), oviduct epithelial cell function (Bauersachs et al., Reference Bauersachs, Blum, Mallok, Wenigerkind, Rief, Prelle and Wolf2003 and Reference Bauersachs, Rehfeld, Ulbrich, Mallok, Prelle, Wenigerkind, Einspanier, Blum and Wolf2004), the endometrium during the oestrous cycle (Bauersachs et al., Reference Bauersachs, Ulbrich, Gross, Schmidt, Meyer, Einspanier, Wenigerkind, Vermehren, Blum, Sinowatz and Wolf2005), pre-implantation embryonic development (El-Halawany et al., Reference El-Halawany, Ponsuksili, Wimmers, Gilles, Tesfaye and Schellander2004; Ushizawa et al., Reference Ushizawa, Herath, Kaneyama, Shiojima, Hirasawa, Takahashi, Imai, Ochiai, Tokunaga, Tsunoda, Tsujimoto and Hashizume2004; Sirard et al., Reference Sirard, Dufort, Vallee, Massicotte, Gravel, Reghenas, Watson, King and Robert2005) and embryo-induced transcriptome changes of the endometrium in the pre- and peri-attachment period (Ishiwata et al., Reference Ishiwata, Katsuma, Kizaki, Patel, Nakano, Takahashi, Imai, Hirasawa, Shiojima, Ikawa, Suzuki, Tsujimoto, Izaike, Todoroki and Hashizume2003; Klein et al., Reference Klein, Bauersachs, Ulbrich, Einspanier, Meyer, Schmidt, Reichenbach, Vermehren, Sinowatz, Blum and Wolf2006). Briefly, the findings indicate that at the ovarian level, processes are typically controlled by genes involved in steroid biosynthesis, oxygen radical metabolism, apoptosis, regulation of cell cycles (including DNA replication and repair), cell structure and tissue remodelling. In the oviducts, variation (e.g. during the oestrus cycle) involves the expression of genes encoding for signal transduction pathways, cell-surface proteins, cell–cell interaction proteins, immune-related proteins and those that promote proliferation. Changes in the endometrium are associated with expression divergence in genes involved in immune function, regulation of cell cycles, apoptosis, tissue remodelling, intercellular communication, cell–cell interaction, cell adhesion molecules, cell motility, intracellular metabolism, protein secretion and modification, and regulation of gene expression.
Hashizume et al. (Reference Hashizume, Ushizawa, Patel, Kizaki, Imai, Yamada, Nakano and Takahashi2007) reviewed bovine gene expression profiles during pregnancy, which involves a cascade of different processes like pre-implantation embryogenesis, maternal recognition of pregnancy, implantation embryogenesis, placentation and foetogenesis. Embryonic mortality is a major contributor to reproductive failure and in about 80% of the cases is caused by suboptimal placentation (Cross et al., Reference Cross, Werb and Fisher1994). Reciprocal signalling and molecular exchange between the mother and the embryo/foetus, at discrete sites that together make up the placentome, is likely to play a critical role. During bovine placentation there is a continuing migration of foetal trophoblast-derived binucleate cells (BNC) and subsequent fusing with maternal uterine epithelial cells into feto-maternal syncytial plaques (Wooding and Flint, Reference Wooding and Flint1994). BNC produce an array of compounds, like steroids, prostanoids, bovine placental lactogens (bPLs), pregnancy-associated glycoproteins (PAGs), bovine prolactin-related protein (bPRP)-1, and the gene expression profiles for BNC-specific bPLs, bPRPs and PAGs vary during pregnancy with apparent strategic effects on implantation, placentogenesis, fetogenesis and overall progression of pregnancy (Hashizume et al., Reference Hashizume, Ushizawa, Patel, Kizaki, Imai, Yamada, Nakano and Takahashi2007).
Bovine embryos grown in vivo are of superior quality compared to those grown in vitro. A subset of expression profiling studies targeting events in early development from oocyte to embryo has been performed in the context of optimizing embryo production and transfer. Mamo et al. (Reference Mamo, Sargent, Affara, Tesfaye, El-Halawany, Wimmers, Gilles, Schellander and Ponsuksili2006) investigated gene expression profiles in in vitro-produced bovine matured oocytes and blastocysts using an 82 target gene cDNA microarray and quantitative real-time polymerase chain reaction (qPCR). Out of the 82 genes, 25 were confirmed as potential candidates for characterizing developmental competence. Clearly, gene expression studies are not restricted to microarray-based approaches and mRNA abundances in bovine oocytes and blastcystes have been determined for a limited number of selected genes using (semi) quantitative PCR (Knijn et al., Reference Knijn, Wrenzycki, Hendriksen, Vos, Zeinstra, van der Weijden, Niemann and Dieleman2005; Lingenfelter et al., Reference Lingenfelter, Dailey, Inskeep, Vernon, Poole, Rhinehart and Yao2007; Warzych et al., Reference Warzych, Wrenzycki, Peippo and Lechniak2007). Wrenzycki et al. (Reference Wrenzycki, Herrmann, Lucas-Hahn, Lemme, Korsawe and Niemann2004) reviewed PCR technology-based gene expression patterns in pre-implantation bovine embyos produced in vitro (IVP), including those derived from somatic nuclear transfer (NT), compared to in vivo-grown embryos. The study evaluated the expression patterns of approximately 100 genes, which constitutes a limited sample as 15000 to 16000 genes may be expressed throughout pre-implantation development in mice (Stanton et al., 2003), and likely other mammals. Gene expression profiles in IVP and especially NT-derived embryos deviate from in vivo-grown embryos and the differences supposedly underlie the abnormalities of which a high birth weight is the predominant feature (large offspring syndrome or LOS). The genes that show expression divergence were suggested as genetic markers for embryo quality and viability.
Whereas IVP-produced embryos may lead to lasting pregnancy in 50% of the cases, for NT this may be 15% or less. Using a 2640 cDNA chip, Pfister-Genskow et al. (Reference Pfister-Genskow, Myers, Childs, Lacson, Patterson, Betthauser, Goueleke, Koppang, Lange, Fisher, Watt, Forsberg, Zheng, Leno, Schultz, Liu, Chetia, Yang, Hoeschele and Eilertsen2005) found 18 genes that were expressed differently in bovine pre-implantation embryos produced by NT or IVP. NT embryos under-expressed three genes that encode for intermediate-filament proteins, possibly foreboding abnormal differentiation and placenta formation, and heat shock protein 27, which may be linked to abnormally high apoptosis (Pfister-Genskow et al., Reference Pfister-Genskow, Myers, Childs, Lacson, Patterson, Betthauser, Goueleke, Koppang, Lange, Fisher, Watt, Forsberg, Zheng, Leno, Schultz, Liu, Chetia, Yang, Hoeschele and Eilertsen2005). Up-regulated genes in NT embryos were suggested to compromise membrane formation (nidogen 2) and increase maternal immune responses to over-expressed MHC-I genes.
Corcoran et al. (Reference Corcoran, Fair, Park, Rizos, Patel, Smith, Coussens, Ireland, Boland, Evans and Lonergan2006) identified differences in gene expression profiles between bovine blastocysts derived from in vivo v. in vitro. Of the 384 genes and expressed sequence tags that were differentially expressed, 85% were down-regulated in the latter. The inferior quality of in vitro-derived embryos was associated with a decreased level of gene expression and, possibly, caused by a deficiency of the machinery associated with transcription and translation (Corcoran et al., Reference Corcoran, Fair, Park, Rizos, Patel, Smith, Coussens, Ireland, Boland, Evans and Lonergan2006). Gene expression profiles can be used to define normal and deviant embryos and from here optimize embryo production protocols. Interestingly, NT- and IVP-derived embryos with widespread dysregulation of genes frequently survive the postnatal period, indicating that mammalian development can tolerate a substantial degree of epigenetic abnormality (Wrenzycki et al., Reference Wrenzycki, Herrmann, Lucas-Hahn, Lemme, Korsawe and Niemann2004).
The research summarized typically is descriptive. The description of changed expressions of (many) genes is often indefinite and non-exclusive, which reflects the biological reality but complicates a clear interpretation of the findings. Computational tools are needed to translate gene expression patterns into metabolic pathways and meaningful biological functions and bioinformatics are increasingly needed for understanding processes underlying reproduction. Noteworthy is the absence of research into gene expression patterns in the brain of dairy cows in relation to fertility. A project has been started to fill this gap and the first findings are reported here. The work is part of collaborate research on the genomics of female fertility, targeting segregating QTL affecting conception rate in dairy cows (Agricultural Research Organization, Israel) and gene expression patterns associated with folliculogenesis (INRA, France), embryonic developmental competence (University of Bonn, Germany and MTT Agrifood Research, Finland) and oestrus/oestrus behaviour (Animal Sciences Group of Wageningen UR, The Netherlands). These and other animal breeding research groups and businesses (33 in total) have joined forces in the EU integrated project SABRE (http://www.sabre-eu.eu/): Cutting Edge Genomics for Sustainable Animal Breeding (6th Framework Programme, Priority 5: Food Quality and Safety).
Genomic regulation of oestrus (behaviour) from the brain
Oestrus behaviour and central mechanisms underlying female sexual behaviour
Oestrus behaviour marks the time of ovulation when cows can be inseminated successfully and oestrus detection is a key factor in the management of fertility: the timing of insemination to the time of ovulation, which ideally is 7 to 18 h before ovulation (Hall et al., Reference Hall, Branton and Stone1959), is critical for the tuning of sperm and oocyte ageing (Hunter and Greve, Reference Hunter and Greve1997). Roughly about 30 h prior to ovulation, cows start expressing oestrus behaviour (Roelofs et al., Reference Roelofs, van Eerdenburg, Soede and Kemp2005). By definition, the standing reflex that cows show when mounted by others identifies heat, but only about 50% or less of the cows may display this (Roelofs et al., Reference Roelofs, van Eerdenburg, Soede and Kemp2005). Relatively little is known about the regulation of oestrus behaviour in farm animals and an experiment was performed to identify genes expressed in the brain of dairy cows that are involved in oestrus and the expression of oestrus behaviour. Different brain areas that are known to play a role in female sexual behaviour have been collected and here findings are presented for a part of the brain containing the ventral tegmental area (VTA).
Studies in rodents show that female reproductive behaviour in mammals hinges on oestrogen actions in the hypothalamus (Pfaff, Reference Pfaff2005), with a central role for the ventromedial hypothalamus (VMH) and involvement of the arcuate nucleus, pre-optic area, paraventricular nucleus and suprachiasmic area. Extra-hypothalamic sites that modulate reproductive behaviour include the limbic system, i.e. amygdala, hippocampus, septum and cingulate gyrus, the stria terminalis and the VTA. Oestrogen primed female rodents, through actions in the VMH, exhibit both appetitive (soliciting) and consummatory (lordosis) behaviour following progestin actions in the VTA. Progesterone from peripheral origin or central biosynthesis is metabolized to dihydroprogesterone (DHP) by 5α-reductase and, subsequently, into 5α-pregnan-3α-ol-20-one (3α,5α-THP) by 3α-hydroxysteroid oxidoreductase (3αHSOR) (Rupprecht, Reference Rupprecht2003). 3α,5α-THP facilitates female sexual behaviour via a mechanism involving the receptors dopamine type 1 (D1, stimulatory), dopamine type 2 (D2, inhibitory) (Frye et al., Reference Frye, Walf and Sumida2004) and γ-aminobutyric acid (GABAA)/benzodiazepine receptor complex (GBR, stimulatory) (Petralia and Frye, Reference Petralia and Frye2005). A key role is assumed for the mitochondrial benzodiazepine receptor (MBR)-mediated increase in neurosteroidogenesis, including that of 3α,5α-THP (Petralia and Frye, Reference Petralia and Frye2005).
Experimental approach
For the identification of genes that are expressed in the brain of dairy cows and that are involved in oestrus (behaviour), a comparison was made between gene expression patterns in brain samples of Holstein Friesian heifers that were euthanized on estimated day 0 (hereafter indicated by 0) of their oestrus cycle with those of heifers slaughtered on estimated day 12 (12), i.e. during the luteal phase when progesterone levels are maximal and before the onset of luteolyse on day 15. This first comparison identifies genes that are involved in regulating oestrus, including the expression of oestrus behaviour. A second comparison was made to narrow down the number of candidate genes that underlie the variability in oestrus expression and identify genes that may be targeted for improving oestrus behaviour. Cows that in general showed a low expression of oestrus behaviour (L) during lactation were compared to those that showed a high expression of such behaviour (H). A total of four comparisons were made. The comparison of group H0 to H12 and L0 to L12 identifies genes that are differentially expressed around oestrus and the luteal phase, and assesses whether this differs between cows that tend to show oestrus behaviour clearly and those that typically do not. The comparison of group H0 to L0 and H12 to L12 indicates whether the H group differs from the L group due to differential gene expression during day 0 or during d12 or for that matter if differences occur regardless of the phase of the oestrus cycle.
Methods
Holstein Friesian heifers (n = 28) were purchased before first calving and group housed in a freestall (research farm the Waiboerhoeve, Lelystad, The Netherlands). During lactation, cows were monitored for oestrus behaviour (continuous observations of half an hour in the morning and afternoon). Behaviours were recorded and scored according to the protocol by Van Eerdenburg (Reference Van Eerdenburg2006): mucous vaginal discharge (3 points), cajoling/flehmen (3), restlessness (5), being mounted but not standing (10), sniffing the vulva of another cow (10), resting with chin on the back of another cow (15), mounting other cows, or attempting to (35), mounting head side of other cows (45), standing heat (100). Points are summed over a rolling 24-h interval and it is assumed that the total score identifies oestrus when accumulating to 50 or more. Milk progesterone concentrations were assessed twice a week and repeated transrectal ultrasonography was performed in the period prior to predicted oestrus. The integrated information was used to determine the moment of euthanasia (by injection of T61), which was done between 77 and 131 days in milk, and calculate mean individual behavioural oestrus expression scores. The latter ranged between 0 and 920 with averages of 375, 414, 87 and 64 for groups H0, H12, L0 and L12, respectively. Efforts were made to euthanize the animals in a stress-free, quick and standardized way, i.e. on the research farm and near the cows’ home section. Brain samples were collected within the hour from euthanasia, immediately stored in liquid nitrogen and maintained at −70°C for long-term storage. Different brain areas were sampled, including a 1-cm-thick slice in a coronal plane taken directly rostral from the colliculus superior in a rostral direction. From the left half of the brain a strip was isolated by cutting from mid-saggital in the lateral direction (1 cm) and from the most inferior in a posterior direction up to the cerebral aquaduct. This means that a relatively large brain sample was taken that included more than only VTA cells (e.g. cells of the substantia nigra and the interpeduncular nucleus). For each experimental group, RNA of seven cows was pooled.
RNA extraction and 20k microarray (Bovine Oligo Microarray Consortium, Jerry Taylor, University of Missouri – Columbia) procedures were performed as described by Niewold et al. (Reference Niewold, Veldhuizen, van der Meulen, Haagsman, de Wit, Smits, Tersteeg and Hulst2007) with only minor modifications. RNA was isolated without sodium citrate/NaCl precipitation and 5 μg RNA was used per labelling. Following hybridization, and before removal of the coverslip, a 5-min delay at room temperature was followed by three subsequent washings of 15 min each in: 2 × SSC + 0.2% SDS (at 42°C), 2 × SSC (room temperature) and 0.2 × SSC (room temperature), after which the procedures conformed to the Micromax TSA protocol (Niewold et al., Reference Niewold, Veldhuizen, van der Meulen, Haagsman, de Wit, Smits, Tersteeg and Hulst2007). Differences in gene expression levels of factors ⩾3 were considered significant.
The study ran from September 2006 to April 2007 and was approved by the Animal Care and Ethics Committee of the ASG-WUR (Lelystad).
Preliminary findings and discussion
Group H0, as compared to H12, showed up-regulation of multiple immunoglobulin superfamily proteins. Immunoglobulins provide the ideal structure for protein–protein interactions and, thus, cell–cell interactions (Rougon and Hobert, Reference Rougon and Hobert2003). Not surprisingly, they play important roles in brain developmental processes, such as axon pathfinding neuronal migration and synapse formation, and the functioning of neuronal networks in adults. Pfaff (Reference Pfaff2005) describes how oestrogens promote female sexual receptive behaviour via increased gene transcription in VMH neurons, causing expansion of dendrites and synapses. Such remodelling of synaptic networks requires signalling by cell–cell interactions, which may be facilitated by the up-regulation of immunoglobulin superfamily proteins that mediate this. Interestingly, the assumed mobilization of immunoglobulin superfamily proteins during oestrus was observed for the H0 group relative to H12, but not for the L0 group relative to L12, suggesting that the mechanism could play a role in the degree of oestrus behaviour expression. Against this interpretation is the fact that the difference between the H0 group and the H12 group reflects low gene expression in the H12 group, not high gene expression in the H0 group. This was because genes encoding for immunoglobulin superfamily proteins were down-regulated in the H12 group as compared to the L12 group.
A gene up-regulated in group H0 as compared to H12 was that encoding for protein phosphatase 1 beta (PP1 beta). Protein phosphatases (PP), including PP1 beta, are widely expressed in the mammalian brain. One of the known PP1 functions is activating Wnt/beta-catenin signalling, a critical event in both cellular proliferation and organismal development (Luo et al., Reference Luo, Peterson, Garcia, Coombs, Kofahl, Heinrich, Shabanowitz, Hunt, Yost and Virshup2007). Compared to the H12 group, the H0 group showed down-regulation of the gene encoding tyrosine hydroxylase, which converts tyrosine into dopa and mediates the rate-limiting step in the catecholamine biosynthesis. Central catecholamine signalling strongly affects behaviour, including female sexual behaviour (Pfaff, Reference Pfaff2005).
Comparing the L0 and L12 groups revealed no differences in gene expression levels.
Relative to the L12 group, the H12 group showed the aforementioned down-regulation of genes encoding immunoglobulin superfamily proteins together with the gene encoding the prohormone, promelanin-concentrating hormone (PMCH). PMHC is proteolytically processed into the orexigenic hormone melanin-concentrating hormone (MCH), as well as into neuropeptide EI (NEI) and neuropeptide GE (NGE). The PMHC gene is involved in the regulation of energy homoeostasis as PMCH-deficient mice have high metabolic rate, are hypophagic and lean whereas mice that over-express the gene are hyperphagic and mildly obese (Marsh et al., Reference Marsh, Weingarth, Novi, Chen, Trumbauer, Chen, Guan, Jiang, Feng, Camacho, Shen, Frazier, Yu, Metzger, Kuca, Shearman, Gopal-Truter, MacNeil, Strack, MacIntyre, Van der Ploeg and Qian2002). The relative high level of PMHC gene expression in cows that tend to poorly express oestrus behaviour may be explained, for example, as a defence mechanism against energy deficit or as a route to establish different energy allocation strategies. The H12 group, relative to the L12 group, showed up-regulation of SCO-spondin. The glycoprotein SCO-spondin, which is an extracellular matrix-like protein and a member of the thrombospondin type 1 repeat (TSR) superfamily (Meiniel and Meiniel, Reference Meiniel and Meiniel2007), is expressed early in the CNS and is involved in developmental processes such as commissural axon pathfinding (Lehmann and Naumann, Reference Lehmann and Naumann2005). The subcommissural organ (SCO) is an ependymal gland located in the roof of the third ventricle that secretes glycoproteins (e.g. SCO-spondin) into the cerebrospinal fluid (CSF), leading to a structure known as Reissner’s fibre (RF). The SCO–RF complex may participate in the clearance of CSF monoamines like catecholamines (Richter et al., Reference Richter, Tome, Yulis, Vio, Jimenez, Perez-Figares and Rodriguez2004).
The comparison of the H0 group to the L0 group showed down-regulation of gamma 2 actin (γ-actin, ACTG2) and desmin (DES). Actins are involved in different cell functions such as cell motility, adhesion and shape (Kaksonen et al., Reference Kaksonen, Toret and Drubin2006). γ-actin influences cell growth and plays a role in the readout of genetic information (Fojo, Reference Fojo2006). Miles et al. (Reference Miles, Andronicos, Baik and Parmer2006) reported on plasminogen-binding proteins identified as rat β/γ-actin on cell surfaces that, upon binding, increased prohormone processing, leading to inhibition of catecholamine release. Desmin plays a role in the structural as well as functional interaction of mitochondria and the cytoskeleton (Capetanaki, Reference Capetanaki2002) and is typically associated with mitochondrial function in muscle cells (Fountoulakis et al., Reference Fountoulakis, Soumaka, Rapti, Mavroidis, Tsangaris, Maris, Weisleder and Capetanaki2005).
Several genes that were differentially expressed between the groups encode for proteins that are involved in general functions like intra-cellular signalling, cell shape, motility and proliferation (γ-actin, DES, PP1 beta). Other findings point to divergences in cell–cell interactions (immunoglobulin superfamily proteins) and catecholamine signalling (tyrosine hydroxylase, SCO-spondin, γ-actin) and energy homeostasis (PMCH). There is no easy way to understand the consistent pattern in gene expression that unambiguously relates to the experimental factors oestrus cycle phase and the cows’ trait to (minimally) express oestrus behaviour, although a more detailed evaluation of the genes and possible functions within the specific brain area is yet to be done and interesting candidate mechanisms are identified. This seems rather unsatisfying, but relates to one of the strong points of gene expression profiling in that it identifies novel directions to explore. The fact that PMCH has been described in more detail than DES does not mean that the former has more biological relevance for oestrus (behaviour) than the latter, but simply that it was more easy to integrate in known concepts. Clearly, the interpretation of the data will be much supported by an increasing level of specificity of the measurements. For example, in terms of homogeneous animals, precise predefined moment of oestrus cycle, exact cell group and standardization of sample collection procedures. Also, data will be more powerful when gene expression profiles are determined at an individual level as these can be statistically linked to detailed phenotypic measurement and, subsequently, tested for multiple factors. With the rapidly decreasing costs of microarray platforms, this has become realistic and is what we will do for tissues from the anterior pituitary, hypothalamus, hippocampus and amygdala.
Developing new genomic research areas
Epigenetics
Epigenetics is about alterations in phenotype due to changes in cellular properties that can be inherited, but do not represent alterations in genotype or, in short, alterations in DNA function without alterations in DNA sequence (Jones and Takai, Reference Jones and Takai2001). Epigenetic modulations of DNA and histones determine the pattern of gene expression and silencing (Jaenisch and Bird, Reference Jaenisch and Bird2003). Histone modifications, like de-acetylation of histones by histone deactylases and site-specific methylation of histone H3, may trigger methylation of cytosine residues by methyltransferases (DNMT3a and DNMT3b), which results in gene silencing that is maintained over cell generations by DMNT1 (Li, Reference Li2002; Jaenisch and Bird, Reference Jaenisch and Bird2003). The epigenetic regulation of gene expression seems inevitable for multicellular organisms as it underlies the development of cell lineage-specific gene expression and phenotypes from one genotype (Jablonka and Lamb, Reference Jablonka and Lamb1998). Also, DNA methylation allows the silencing of virus-derived and potentially destabilizing repeat sequences that make up large proportions of the mammalian genome (International Human Genome Sequencing Consortium, 2001). Epigenetics determines the formation and specification of germlines, the erasure and re-establishment of methylation patterns in embryonic primordial germ cells and, subsequently, re-establishment of sex-specific patterns during gametogenesis and results in different modifications of the paternal and maternal genome after fertilization (Allegrucci et al., Reference Allegrucci, Thurston, Lucas and Young2005). The differential expression of paternal and maternal alleles based on epigenetic differences stemming from the germlines is known as genomic imprinting. The resulting monoallelic expression of imprinted genes is tissue specific, seems restricted to the prenatal period (Moore, Reference Moore2001) and functions to limit the production of transcripts (Allegrucci et al., Reference Allegrucci, Thurston, Lucas and Young2005). Imprinted genes may account for only a fraction of the genome (e.g. ∼0.3% (Miozzo and Simoni, Reference Miozzo and Simoni2002)), but these affect early development strongly and may have a major impact on fertility (Fowden et al., Reference Fowden, Sibley, Reik and Constancia2006). In 2006, Gebert et al. (Reference Gebert, Wrenzycki, Herrmann, Groger, Reinhardt, Hajkova, Lucas-Hahn, Carnwath, Lehrach and Niemann2006) compared bovine oocyte to sperm DNA and reported on a differentially methylated region in the imprinted insulin-like growth factor 2 gene and this allows further work on gene-specific methylation patterns during pre-implantation development. This phenomenon has been hypothesized to cause subfertility in offspring from parents under malnutrition.
RNA interference
Validating gene function through knock-out studies in farm animals using conventional methods is possible (Denning and Priddle, Reference Denning and Priddle2003; Hofmann et al., Reference Hofmann, Zakhartchenko, Weppert, Sebald, Wenigerkind, Brem, Wolf and Pfeifer2004), though typically it will be costly and complicated (Wolf et al., Reference Wolf, Schernthaner, Zakhartchenko, Prelle, Stojkovic and Brem2000). The application of RNA interference (silencing) may, in part, circumvent practical problems. Long double-stranded RNAs and micro-RNA precursors are processed into short (∼22 nucleotides) double-stranded RNAs, i.e. micro-RNAs and short interfering RNAs (Meister and Tuschl, Reference Meister and Tuschl2004). The latter two are processed into micro-RNA containing ribonucleoprotein particles, which repress the translation of target mRNA, and RNA-induced silencing complex, which increases the degradation of target mRNA. Hiendleder et al. (Reference Hiendleder, Bauersachs, Boulesteix, Blum, Arnold, Frohlich and Wolf2005) discuss how RNA interference can be applied without transgenic approaches. Small organisms and cells are treated effectively by soaking them in double-stranded RNA solutions, but in mice short interfering RNAs have been injected effectively in target tissues. The resulting decrease in gene expression may be in the range of 30% to 60% for over a week, and the approach mediated significant effects on plasma metabolite concentrations (Soutschek et al., Reference Soutschek, Akinc, Bramlage, Charisse, Constien, Donoghue, Elbashir, Geick, Hadwiger, Harborth, John, Kesavan, Lavine, Pandey, Racie, Rajeev, Rohl, Toudjarska, Wang, Wuschko, Bumcrot, Koteliansky, Limmer, Manoharan and Vornlocher2004). Possible future applications of RNA interference tools are discussed by Lew et al. (Reference Lew, Jackson and Bellgard2005).
Variable copy numbers
Higher-order architectural features of the genome, as compared to primary sequence information, are linked to genomic instability and susceptibility to genomic rearrangements like deletions, duplications, inversions or translocation of genomic segments, marker chromosomes and isochromosomes. The human genome has been estimated to be composed of ∼5% duplicated sequences that occur both within and between chromosomes (Sharp et al., Reference Sharp, Cheng and Eichler2006). Such DNA segments of 1 kb or larger that are present as variable copy numbers (CNVs) are known to alter gene dosage, interfere with coding sequences and have gene-regulating effects. For example, three instead of two copies of the normal PMP22 gene causes serious neurological disease (Lee and Lupski, Reference Lee and Lupski2006). For a population of humans, Redon et al. (Reference Redon, Ishikawa, Fitch, Feuk, Perry, Andrews, Fiegler, Shapero, Carson, Chen, Cho, Dallaire, Freeman, Gonzalez, Gratacos, Huang, Kalaitzopoulos, Komura, MacDonald, Marshall, Mei, Montgomery, Nishimura, Okamura, Shen, Somerville, Tchinda, Valsesia, Woodwark, Yang, Zhang, Zerjal, Zhang, Armengol, Conrad, Estivill, Tyler-Smith, Carter, Aburatani, Lee, Jones, Scherer and Hurles2006) found that 12% of the genome consisted of copy number variable regions, and since this encompasses more nucleotide content than SNPs, the contribution of CNVs to phenotypic variation may be substantial. Stranger et al. (Reference Stranger, Forrest, Dunning, Ingle, Beazley, Thorne, Redon, Bird, de Grassi, Lee, Tyler-Smith, Carter, Scherer, Tavare, Deloukas, Hurles and Dermitzakis2007) linked gene expressions (14 072 genes) in human lymphoblastoid cell lines to SNPs and CNVs, with the former being investigated at a relatively high resolution as compared to the latter (meaning that the effect of CNV will have been underestimated). There was little overlap between SNPs and CNVs and each captured 83.6% and 17.7% of the genetic variation in gene expression, respectively. Contemporary CNV maps are far from complete (Redon et al., Reference Redon, Ishikawa, Fitch, Feuk, Perry, Andrews, Fiegler, Shapero, Carson, Chen, Cho, Dallaire, Freeman, Gonzalez, Gratacos, Huang, Kalaitzopoulos, Komura, MacDonald, Marshall, Mei, Montgomery, Nishimura, Okamura, Shen, Somerville, Tchinda, Valsesia, Woodwark, Yang, Zhang, Zerjal, Zhang, Armengol, Conrad, Estivill, Tyler-Smith, Carter, Aburatani, Lee, Jones, Scherer and Hurles2006) and increasing power for the detection of the smaller CNVs will likely increase the importance of CNVs in phenotypic variation. DNA microarrays have proved to be a powerful tool in detecting CNVs of 40 to 50 kb and larger.
Nutrigenomics
Nutrients can be seen as dietary signals that influence gene expression and, subsequently, protein and metabolite production (Muller and Kersten, Reference Muller and Kersten2003). Nutritional genomics or nutrigenomics deals with the genome-wide influences of nutrition and studies the effects of dietary components on gene expression patterns (transcriptome), chromatin organization (epigenome), protein expression patterns (proteome) and metabolite profiles (metabolome) (Afman and Muller, Reference Afman and Muller2006). Macronutrients (fats, carbohydrates and proteins), micronutrients (vitamins and minerals) and food components like flavonoids all participate in the binding of nuclear receptors to response elements in gene promoter regions (Muller and Kersten, Reference Muller and Kersten2003) and, through the resulting co-repressor dissociation and co-activator recruitment, facilitate transcriptional activation. Nutrient-mediated activation of nuclear receptors may affect many cell functions as, for example, fatty acid-sensitive nuclear receptor peroxisome proliferator activator receptor-α (PPAR α) targets 3000 to 4000 (hepatic) genes involved in processes such as acute-phase protein responses, cell proliferation, fatty acid oxidation, gluconeogenesis and amino acid metabolism (Mandard et al., Reference Mandard, Muller and Kersten2004). Early work and prospects for nutritional genomics in cattle are discussed by Dawson (Reference Dawson2006), including the effects of endophyte-infected forages and poor-quality feed. The author argues on the basis of dietary-induced changes in gene expression how selenium, through triiodothyronine, modulators of oxidative stress and the thioredoxin carrier system, could affect fertility in cattle. Future nutrigenomic work will result in detailed insight in cows’ nutritional and (reproductive) physiological status and help tailor feeding strategies to improve fertility. For example, as it may no longer be necessary to apply diets in a relatively extreme form over a prolonged period of time in order to assess its effects (Dawson, Reference Dawson2006). Before nutrigenomics results in the tailoring of rations to the needs of groups of individuals (genotypes), some hurdles must be overcome. Mammals handle high concentrations of different nutrients and associated metabolites, with different substances binding multiple targets with different affinities. This complexity obstructs the interpretation of findings from nutrigenomic studies and would require extreme standardization and control of experimental conditions. In general, individual-specific variation in the degree of gene expression may be considerable (Radich et al., Reference Radich, Mao, Stepaniants, Biery, Castle, Ward, Schimmack, Kobayashi, Carleton, Lampe and Linsley2004) and variation in environmental factors for the different experimental animals may cause false positives. Realistically, in the near future complex problems within the field of nutrigenomics may need to be down-sized to small feasible projects that, for example, make use of model systems (Afman and Muller, Reference Afman and Muller2006).
Summarizing conclusions
Increases in genetic merit for female fertility can be combined with high milk yield and nowadays dairy cattle breeding programmes incorporate fertility in their selection indices. Traditional breeding value estimation procedures are combined with gene-assisted selection and, more recently, genomic selection based on genome-wide information. Functional genomics builds on information on static aspects of the genome and is about the implementation of high-throughput technologies like gene expression profiling by microarray platforms. Recent gene expression profiling studies have produced a wealth of data regarding possible genes and physiological processes underlying phenotypic fertility traits. Typically, processes under study involve a large number of genes, which by their transcripts effectuate multiple functions depending on tissue, developmental stage etc. Extracting clear-cut interpretations from the data that reflect the biological reality are complex. Currently, gene expression profiling is already a powerful tool for identifying candidate genes/mechanisms underlying fertility traits, as demonstrated for centrally expressed genes involved in oestrus (behaviour). When aided by advanced computational tools and bioinformatics, it may produce gene expression data that become increasingly self-explanatory. A practical application is the use of gene expression patterns for defining normal and deviant embryos and to optimize embryo production protocols. Studies on alterations in DNA function without alterations in DNA sequence (epigenetics) provide new insights and angles to investigate physiological problems. For example, imprinted genes strongly affect early development and likely have a major impact on fertility. Research into gene-specific methylation patterns during pre-implantation development is underway. RNA interference, as applied without transgenic approaches, may develop into a useful tool to manipulate gene expression and thus validate hypothesized gene functions and physiological mechanisms in farm animals. Nutrigenomics deals with the effects of dietary components on gene expression patterns and such information can help to tailor feeding strategies in order to improve dairy cow fertility. It is recognized that genome-wide influences of nutrition are complex and tackling practical problems by this route may take considerable effort. Recent findings indicate that CNVs may contribute significantly to phenotypic variation. Microarrays allow high-throughput CNV detection and new insights in the genomic regulation of phenotypic fertility traits are to be expected. In the years to come, large amounts of genomic information will become available to be used for the unravelling of genomic and physiological mechanisms underlying fertility traits and fine-tuning of effective selection strategies. Dealing with the information stress may prove challenging. There is great need for bioinformatic solutions to integrate different types of information on, for example, gene expression, DNA methylation, QTL and SNPs, and aid the processing of such data into biologically meaningful interpretations. The integration of large amounts of new genomic information with that on physiological mechanisms is expected to open novel ways to combat issues like subfertility in dairy cattle.
Acknowledgements
Regarding the animal experiment, the support of Jan Bloemert and co-workers of research farm Waiboerhoeve is highly appreciated and we are indebted to Henk Sulkers and Diana Baumann for their assistance. Norbert Stockhofe and Franz-Josef van der Staay are thanked for helping to set up the study and Ad Korevaar for his crucial role in performing it.
The work was financially supported by the international cattle breeding co-operative CRV and has been co-financed by the European Commission, within the 6th Framework Programme, contract No. FOOD-CT-2006-016250. The text represents the authors’ views and does not necessarily represent a position of the Commission who will not be liable for the use made of such information.
