Journal of Mammary Gland Biology and Neoplasia

, Volume 12, Issue 4, pp 305–314

Genomewide Analysis of Secretory Activation in Mouse Models

Authors

  • Palaniappan Ramanathan
    • Centre for Advanced Technologies in Animal Genetics and Reproduction, Faculty of Veterinary ScienceUniversity of Sydney
    • CRC for Innovative Dairy Products, Faculty of Veterinary ScienceUniversity of Sydney
  • Ian Martin
    • Centre for Advanced Technologies in Animal Genetics and Reproduction, Faculty of Veterinary ScienceUniversity of Sydney
  • Peter Thomson
    • Centre for Advanced Technologies in Animal Genetics and Reproduction, Faculty of Veterinary ScienceUniversity of Sydney
    • CRC for Innovative Dairy Products, Faculty of Veterinary ScienceUniversity of Sydney
  • Rosanne Taylor
    • Centre for Advanced Technologies in Animal Genetics and Reproduction, Faculty of Veterinary ScienceUniversity of Sydney
  • Christopher Moran
    • Centre for Advanced Technologies in Animal Genetics and Reproduction, Faculty of Veterinary ScienceUniversity of Sydney
    • CRC for Innovative Dairy Products, Faculty of Veterinary ScienceUniversity of Sydney
    • Centre for Advanced Technologies in Animal Genetics and Reproduction, Faculty of Veterinary ScienceUniversity of Sydney
    • CRC for Innovative Dairy Products, Faculty of Veterinary ScienceUniversity of Sydney
Article

DOI: 10.1007/s10911-007-9052-6

Cite this article as:
Ramanathan, P., Martin, I., Thomson, P. et al. J Mammary Gland Biol Neoplasia (2007) 12: 305. doi:10.1007/s10911-007-9052-6

Abstract

Mouse models have been widely used to elucidate the biology of mammary gland development and secretory activation. Recent advances in the availability of genomic resources for mice will generate a renewed effort to define the genetic basis of lactation phenotypes and help identify candidate gene pathways. Specific aspects of these advances are relevant to the dairy industry and may provide a rationale for improving milk production in the dairy cow. Differences are evident in mammary gland morphology and various characteristics of milk production of inbred mouse strains, but few studies have undertaken any systematic phenotypic analysis of the different inbred strains of mice for lactation performance. Whole genome association analysis using recent strain-specific genotype data and detailed phenotype measurements from available inbred strains, along with transcript profiling of divergent inbred strains for lactation performance, provides a valuable approach to identify putative candidate genes and associated pathways underlying dairy QTL intervals. Here we discuss the utility of integrating mouse phenomic and genomic resources for understanding secretory activation in the mammary gland.

Keywords

Secretory activationwhole genome associationTranscriptome analysis

Abbreviations

QTL

quantitative trait loci

Mprf

maternal performance

Introduction

Many quantitative trait loci (QTL) for the different components of milk production, including milk yield, protein yield, and fat percent, have been identified in dairy cattle and more recently integrated with bovine sequence information [1]. The tools to dissect these QTL regions have also improved with the release of the bovine genome assembly, and web-accessible tools to visualize the genome sequence information (http://bovineqtlv2.tamu.edu/home.php [2]; http://www.animalgenome.org/QTLdb/ [3]). However, as in the case of QTL studies in other species, the confidence intervals of most of these dairy QTLs are large and contain hundreds of genes. Hence, only a few QTLs have been narrowed down to a narrow interval and the causative gene identified. Two notable examples are diacylglycerol transferase 1 (DGAT1) and the Protease Inhibitor (PI) genes, which were identified using a fine mapping strategy and a positional comparative candidate gene analysis approach respectively [4, 5]. Although there have been remarkable increases in milk production over the past three decades, the cost of production has also grown significantly while increased demand for milk products continues. Consequently, the imperative to develop strategies to improve dairy cow health and production remains.

Recent advances in genomics are driving a renewed effort to understand the complex biological mechanisms that control secretory activation/lactogenesis. Milk production fundamentally depends on three distinct processes, namely pre-partum proliferation of the secretory alveoli, cellular differentiation, and synthesis and secretion of milk components. Although these processes have been studied in the context of ruminant species, experimental mouse models remain as an important and strategic tool to unravel the details of molecular mechanisms that drive lactation. This is particularly important in the current era with the dramatic increase in genomic data from bovine, mouse and other species. Comparative genomics, improved tools for quantitative genetics and new methods to integrate information are leading to a marked increase in the power of genomewide analysis of lactation.

An Historical Perspective on Studies of Mouse Lactation

The first experimental study that internal secretions from the corpora lutea were responsible for the growth of the mammary gland was performed in mice [6]. A detailed description including microscopic and macroscopic changes occurring in the mouse mammary gland during different stages of the lactation cycle, namely puberty, pregnancy, lactation and involution, was described in 1933 [7]. Soon after, strain-specific differences in mammary gland morphology were recognized amongst ten different strains of mice [8]. A critical role for steroidal hormones secreted by the pituitary, adrenal cortex and ovaries, in mammogenesis and lactogenesis was demonstrated by means of surgical removal or exogenous supplementation of various hormonal combinations in different strains of mice during the 1950s and 1960s [9, 10]. These studies helped to demonstrate that estrogen (E) and growth hormone (GH) stimulate ductal elongation, while progesterone (P) is necessary for alveolar development. Furthermore, they established that prolactin (PRL), GH, and adrenal steroids are involved in lobuloalveolar differentiation, milk synthesis, and lactation. More recently, additional insights have come from the study of genetic models in which individual hormones or their receptors have been knocked out or overexpressed in mice, providing a detailed framework for understanding hormonal regulation of mammary gland development.

Assessment of Lactation Performance in Mice

During the late 1940s and particularly in Britain, there was a major initiative to enhance livestock production to achieve self-sustainability. One of the priority areas for attention was dairy production. Experimentally, there was a renewed focus on using animal models and the mouse, with its short gestation and high prolificacy, provided a valuable source of phenotypic data. Cumulative litter weight at 12 days of lactation was suggested as a convenient measure of milk production in mice [11]. Initially, there were conflicting reports that 32 or 71.5% of the phenotypic variance measured in litter weight gain was attributed to the dam [12, 13]. Subsequently, it was clearly demonstrated that postnatal influences accounted for 70–80% of the variation in 12-day litter weight, weaning weight and weight gain from birth to weaning, which are all reliable indicators of lactation performance [14, 15]. Lactation curves based on milk yield identified 13–14 day postpartum as peak milk yield in mice [16] further substantiating prior findings based on morphological, histological and biochemical studies [7, 17]. Thus, the first 2 weeks of lactation was recognized as the critical phase during which the weight gain of pups were mostly attributed to maternal performance, since beyond 12 days of age they start consuming solid food. Milk yield was shown to be positively correlated with both postnatal mammary gland development and postpartum dam body weight using mice lines selected for superior nursing ability [18]. Additionally, neonatal growth during the first week of lactation was a better correlate of milk yield than during the second week of lactation, when peak milk yield was attained [19].

Secretory Activation and Gene Knockout Models

In the past decade, screening with high throughput gene expression technologies has implicated the involvement of a greater number of genes in mammary development than had been previously identified. Transgenic technologies have been increasingly used to delineate the role of specific genes in mammary gland development, but relatively few have used similar approaches to examine lactogenesis. This deficiency may be due to the difficulty in assessing lactation defects in transgenic mice, and potential approaches to overcome this have been previously reviewed [20]. A list of studies that have characterized the role of specific genes during lactation using genetically modified mouse strains are summarized in Table 1.
Table 1

List of transgenic mouse studies reporting a lactation phenotype.

Animal model

Lactation phenotype

Reference

β-casein −/−

Can lactate but reduced pup growth

[56]

α-Lalba −/−

Highly viscous milk, pups unable to suckle

[57]

TGFβ-1 overexpression

Failure to lactate; lobuloalveolar defects

[58]

Wa-2 mice (Point mutation in EGFR gene)

Impaired maternal lactation

[59]

Oxytocin −/−

Failure of milk let down

[60]

Prlr −/−

Failure to lactate; defective mammary gland development

[61]

Stat 5a −/−

Failure to lactate; lobuloalveolar defects

[62]

Gal −/−

Failure to lactate; reduced prolactin

[63]

Peg3 +/−

Reduced lactation; reduced oxytocin levels

[64]

Amphiregulin KO (AR−/−), EGFR −/− & TGFα −/− (Triple null females)

Impaired lactation due to reduced ductal outgrowth and alveolar differentiation

[65]

Opgl −/− mice (or) its receptor RANK −/− mice

Failure to lactate; inhibition of lobulo-alveolar development.

[66]

Id2 −/−

Lactation defect; impaired lobulo-alveolar development.

[67]

Socs1 −/− & IFN γ −/−

Precocious lactation and Socs1 +/− rescues lactation in Prlr +/− mice

[68]

Cav1 −/−

Accelerated lobulo-alveolar development and premature lactation

[69]

Xdh +/−

Disrupted lipid secretion, litters die in mid-lactation

[70]

Usf2 −/−

Diminished milk synthesis due to reduced oxytocin levels

[71]

Erbb4 −/−

Lactation failure; impaired epithelial proliferation.

[72]

Tg DN β-catenin chimera (beta- eng)

Lactation failure; impaired lobuloalveolar development

[73]

Akt over expression (MMTV-myr-Akt)

Large lipid droplets causing highly viscous milk and reduced pup growth

[74]

HIF1α −/−

Lactation failure; impaired mammary differentiation and lipid synthesis

[75]

α-catenin −/−

Reduced lactation; enhanced epithelial apoptosis

[76]

PPAR-interacting protein (PRIP) −/−

Reduced lactation; decreased ductal branching and alveolar expansion

[77]

IGF1 over expression

Enhanced milk synthesis

[38]

Jak2 −/−

Failure to lactate; impaired alveologenesis

[78]

Btn1a1 −/−

Disrupted lipid secretion, 50% pup survival at weaning

[79]

Csnk −/−

Failure to lactate; casein micelle destabilisation

[80]

−/− Homozygous gene deletion; +/− heterozygous

Whole Genome Analysis using Inbred Mouse Strains

Mouse QTL Analysis

Inbred strains of mice differ markedly in mammary gland development and milk production, providing a genetic resource to identify genes responsible for the differences. A few studies have characterized a difference in lactation performance between inbred strains of mice and identified putative genomic regions associated with the related traits. Two QTLs for maternal performance (Mprf) were identified on mouse chromosome 2 and 7, which together with 15 epistatic interactions between 23 loci, accounted for approximately 35% of the total phenotypic variance in a cross-breeding experiment between LG/J and SM/J mice [21]. A novel nurturing ability QTL (Naq1), was significantly associated with the inferior nurturing ability of the RR inbred strain of mice at 72 cM on mouse chromosome 5 [22, 23]. The QTL was identified based on the weight gain between days 7 and 12 of lactation, but not at a later stage suggesting that it reflected the decreased lactation capacity of the RR strain of mice. A neonatal growth QTL (Neogq1) based on the individual pup weight gain during the first 7 days of lactation was identified on mouse chromosome 9 using a backcross between excellent (NFR/N) and poor (B.10 Q) breeder mouse strains [24]. Together these studies demonstrate that mapping genetic variation amongst different inbred strains of mice helps to identify critical genomic regions associated with lactation performance. However, the QTL regions are usually large and often contain many genes. The confidence intervals of the QTL regions may be reduced by fine mapping using various mouse resources [2527].

Advanced Intercross Lines

One strategy for fine mapping of QTL regions is to develop an advanced intercross line, which is created by repeated crossing of two strains divergent for a specific phenotype, followed by repeated intercrossing subsequent to the F1 generation that does not involve any brother–sister mating so as to increase the cumulative recombination [25]. Using this strategy, QTL regions spanning 10–20 cM can be narrowed down to 1–2 cM in an F10 generation. The disadvantage is that this process may take 2–3 years, and requires the phenotyping and genotyping of hundreds of animals, or selective genotyping of animals in either extremes based on the phenotype. We utilised a QSi5-CBA advanced intercross line assessed for lactation performance and showed significant patterns of segregation towards the QSi5 mice in the 13th–14th generation (Fig. 1). On the basis that gene expression is genetically controlled, we performed a preliminary analysis of gene expression in the “extremes” to circumvent the requirement for mass genotyping. The AIL population revealed distinct clustering of the two groups based on principal component analysis of the top 500 varying genes. Correlation of the lactation phenotype with normalized expression measures among the replicates in the “extremes” identified 417 probesets with a Pearson’s correlation coefficient greater than 0.8 and a p value < 0.01. This included 219 and 198 probesets over- and under-represented in the superior lactation phenotype AIL group relative to the inferior group. A heatmap of the highly correlated and differentially expressed genes in both groups is given in Fig. 2. A list of candidate genes associated with the trait and over-expressed in the lactating mammary gland is provided in Table 2.
https://static-content.springer.com/image/art%3A10.1007%2Fs10911-007-9052-6/MediaObjects/10911_2007_9052_Fig1_HTML.gif
Fig. 1

Histogram showing the variation in pup growth rate in two inbred strains and an AIL.

https://static-content.springer.com/image/art%3A10.1007%2Fs10911-007-9052-6/MediaObjects/10911_2007_9052_Fig2_HTML.gif
Fig. 2

Correlation of gene expression and lactation phenotype values of the AIL deciles. Top panel of the image showing phenotypic variables arranged across the top in ascending order. Pearsons’s correlation coefficient values are listed on the left and the probeset identification is on the right (Image output from T-Rex in GEPAS).

Table 2

List of mammary-specific differentially expressed genes in the CBA-QSi5 AIL population along with GO biological categories.

Gene symbol

Gene title

GO biological process term

Scd1

Stearoyl-Coenzyme A desaturase 1

Lipid metabolic process

Slc34a2

Solute carrier family 34 (sodium phosphate), member 2

Transport

Lgals12

Lectin, galactose binding, soluble 12

Apoptosis

Egln3

EGL nine homolog 3 (C. elegans)

Apoptosis

Fads2

Fatty acid desaturase 2

Electron transport

St6gal1

Beta galactoside alpha 2,6 sialyltransferase 1

Protein amino acid glycosylation

BC011487

cDNA sequence BC011487

Slc25a1

Solute carrier family 25 (mitochondrial carrier, citrate transporter), member 1

Transport

Tcf7l2

Ttranscription factor 7-like 2, T-cell specific, HMG-box

Transcription

Mod1

Malic enzyme, supernatant (Malic enzyme 1)

Malate metabolic process

Lpl

Lipoprotein lipase

Lipid metabolic process

BC029169

cDNA sequence BC029169

Pik3r1

Phosphatidylinositol 3-kinase, regulatory subunit, polypeptide 1 (p85 alpha)

Negative regulation of cell-matrix adhesion

Tmem176a

Transmembrane protein 176A

Elovl1

Elongation of very long chain fatty acids (FEN1/Elo2, SUR4/Elo3, yeast)-like 1

Fatty acid biosynthetic process

Folr1

Folate receptor 1 (adult)

Folic acid metabolic process

Scarb1

Scavenger receptor class B, member 1

Cell adhesion

Tmem16a

Transmembrane protein 16A

Expansion of Genotype Databases for Inbred Strains

The release of the genome sequence for the C57BL6J mouse strain in 2002 by the publicly funded Mouse Genome Sequence Consortium revealed that the sequence was approximately 90% similar to the human genome sequence, and established that mice have a similar number of genes as humans, further substantiating the usefulness of the mouse as a model organism. Subsequently, sequencing of four commonly used laboratory inbred strains, namely 129Sv1J, AJ, Balb/c, C3H, and NOD, was completed to identify sequence variation within the species. Further large scale sequencing efforts searching for single nucleotide polymorphism (SNP) in various inbred strains by different organisations, including Roche, Celera, Wellcome Trust, NIEHS-Perlegen, and GNF, greatly increased the number of informative polymorphic sites (Table 3). Recently, the first mouse haplotype map composed from 49 inbred strains and approximately 140,000 markers at an average density of 1 SNP for every 500 kb, was released by the Broad Institute. The large SNP datasets are accessible from the Mouse Phenome database (MPD) (http://phenome.jax.org/pub-cgi/phenome/mpdcgi?rtn=docs/home) which facilitates SNP identification in specific genomic regions across all SNP databases.
Table 3

List of mouse genomewide SNP datasets.

Data source

Year

SNP recordsa

No. of strains

GNF

2004

9,594

48

Roche

2004

11,571

18

WTCHG

2005

69,781

67

NIEHS-Perlegen

2005

8,273,434

16

Celera

2006

2,093,327

5

Broad

2006

138,608

49

Table adapted from MPD Mouse SNP database: http://phenome.jax.org/pub-cgi/phenome/mpdcgi?rtn=snps/chooseproj

aRepresents the total number of SNP records that are eligible for inclusion in the MPD SNP database

Mouse Phenome Databases

Simultaneously, a number of organizations considered the need for collating phenotype data on a wide range of traits from different inbred strains and making it publicly available. The Jackson Laboratory, which developed the first computer database for mouse genetics in 1972, in a precursor to the Mouse Genome Database, took a lead in collating phenotypic data from different inbred strains, developing the Mouse Phenome Database. This resource currently has phenotype data for about 500 different traits and continues its expansion [2830]. Similarly, the Eumorphia database (http://www.eumorphia.org) was created for morphological phenotypes [31].

Closely related quantitative traits exhibit a high correlation, perhaps reflecting pleiotrophic effects of the same loci. When considering available phenotype data for assessing lactation traits, consideration should be given to reproductive traits. We determined the correlation for six measured reproductive traits (number of litters born/dam, litter size, percent of productive matings, number of mice born, number of mice weaned and percent weaned), one derived reproductive trait (relative fecundity) and one lactation (individual pup weight gain). All pair wise correlations amongst the different traits are positive. There is an extremely high correlation of 0.99, as may be expected, between mice born and mice weaned as the later is a subset of the former. Similarly, mice per litter (litter size) also have a significant correlation (p < 0.05) of 0.94 and 0.96 with both mice born and mice weaned, respectively. Relative fecundity, which is estimated from number of litters/dam, mice per litter and percent productivity, is significantly correlated with all three having a Pearson’s correlation coefficient of 0.95, 0.8 and 0.63 respectively. Relative fecundity is also significantly correlated with all other traits measured, having a near perfect correlation of 0.99 with both mice born and mice weaned, and a correlation coefficient of 0.63 and 0.61 with lactation yield (pup weight gain) and percent of mice weaned per dam respectively. Lactation yield is also correlated with other reproductive traits, percent weaned and litter size (r = 0.81 and 0.78 respectively; p < 0.05), and a lower correlation with mice born (r = 0.57) and mice weaned (r = 0.62)(0.05 < p < 0.1). Previous studies have reported a correlation between litter size at birth and both 12-day litter weight as well as maternal body weight [32].

Whole Genome Association Mapping

A novel computational method assessing the association between genotype and the phenotype of the inbred strains was suggested as an alternative approach to identify chromosomal regions associated with complex traits, and to refine QTL intervals. This strategy of association mapping was referred to as in silico mapping [33]. Although the study was limited by the small number of strains used and the available genotype data, the advantages of the approach were recognized and seen as an opportunity to identify genes underlying QTL intervals much faster than conventional approaches that utilize mouse resources. A high resolution haplotype map was produced from SNP derived from 16 different inbred strains found in the Roche mouse SNP database. Correlation with phenotypic data resulted in identification of haplotype blocks significantly associated with specific traits, especially monogenic traits [34]. Later 10,990 SNPs distributed uniformly across the mouse genome were genotyped in 48 inbred strains. Use of an inferred haplotype based on patterns observed in blocks of 3 consecutive SNPs was successfully used in mapping both single gene traits and importantly, complex traits [35]. The power of predicting true associations was directly proportional to both the number of SNPs and the number of strains. However, owing to the limited SNP density, this method was still only capable of inferring haplotype association with a resolution of approximately 1 Mb. The subsequent availability of large SNP datasets has overcome this limitation, but spurious SNP associations have been noted as a result of the unequal relatedness among the different inbred strains [36]. A novel cladistic analysis method which takes into consideration the genealogy of inbred strains was partially effective. However, the use of an integrated gene mapping strategy combining information from QTL analysis, regions of shared ancestry and gene expression was effective in identifying candidate genes for complex traits [37]. Recently, rather than association mapping based on a 3-SNP window or haplotype blocks of fixed sizes, haplotype blocks, as defined by Haploview [38], were used for association analysis [39]. The haplotype tag SNP algorithm was effective in identifying associations for platelet count, and these SNPs were subsequently used for selective genotyping of an F11 population, which led to the identification of candidate genes [39]. Other studies using similar approaches have also successfully identified candidate genes for complex traits [40].

We performed whole genome association studies for lactation performance using the haplotype tag SNP method. Eleven different inbred strains of mice were assessed using individual pup weight gain during the first 8 days of lactation. The genotypes of these strains were obtained from the first inbred mouse haplotype map consisting of about 140,000 SNPs distributed uniformly at 1 SNP every 20 kb interval across the mouse genome [41]. The analysis identified haplotype blocks significantly associated with the trait (see e.g. Fig. 3).
https://static-content.springer.com/image/art%3A10.1007%2Fs10911-007-9052-6/MediaObjects/10911_2007_9052_Fig3_HTML.gif
Fig. 3

An example of genomewide haplotype associations for lactation performance on chromosome 5.

Transcriptome Analysis

Over the past decade, microarray technology has allowed researchers to assess expression levels of thousands of genes simultaneously, but to effectively utilize such high-volume data, a range of bioinformatic and statistical tools needed to be created. While initial investigations in this area focused on quantifying “fold change” between states of interest, a range of methods have been derived (see e.g. [42]).

The first studies on expression profiling in the normal mouse mammary gland were to demonstrate the application of a tool, Microarray Explorer tool for data mining [43]. Subsequently, a large body of work emerged using gene expression microarrays for analysis of mammary tissue, mostly in humans and mice, and almost entirely focused on breast cancer (see e.g. [44]). A few mouse studies included information on lactation [43, 45], and the RIKEN expression database includes data from lactation day 10 mouse mammary tissue (http://read.gsc.riken.go.jp/). More recently comparisons have been made of different inbred strains of mice across the major stages of the lactation cycle to identify potential candidate gene clusters and networks. A temporal gene expression profiling experiment was carried out in the highly fecund FVB mouse strain of mice to better understand transition of the mammary gland from a rudimental ductal structure at birth, through the pubertal growth of the epithelial ductal tree, and the development of the alveolar epithelial compartment during pregnancy [45, 46]. A list of genes associated with secretory activation, primarily associated with progesterone withdrawal, was identified by comparing expression differences between late pregnancy and early lactation. Gene expression profiling was also performed in C57Bl/6 and Balb/C mice to identify genes involved in remodelling of the mammary gland during involution, during which the highly developed mammary architecture reverts to a virgin-like state [47, 48]. We used a highly fecund inbred mouse strain (QSi5) [49] for gene expression studies. Comparison of QSi5 and CBA for lactation performance identified candidate genes potentially responsible for the enhanced lactation phenotype of the QSi5 mice (Ramanathan and Williamson, unpublished observations). Together, these experiments implicate a larger number of candidate genes potentially involved in mammary gland regulatory events during the lactation cycle than has been previously identified.

Integrative Approaches

Despite the potential advantages of in silico association mapping in inbred mouse strains, at this time there are limitations to the power of association mapping studies based on available phenotype–genotype resources [36, 50]. This situation will change as detailed genotype data and the haplotype structure of many more strains emerges, along with phenotypic data. However, what is apparent is the utility of combined genome wide analysis using various integrated strategies. Mostly, these involve integration of QTL analysis, association mapping and gene expression profiles.

In one example, a unique nurturing ability QTL Naq1 was identified on mouse chromosome 5 but this did not correspond to any other known QTL. Among 30 genes lying within 1 LOD score confidence interval of the QTL, 12 genes were identified as primary candidate genes based on tissue-specific expression in the mammary gland [23]. A further reduction of this region may be achieved by association mapping of inbred strains.

A second example in which association mapping was included, successfully identified a small number of candidate genes for complex traits [37]. The approach included a novel cladistic analysis, which takes into account the genealogies of the different inbred strains and correlates them with the observed phenotype. A further integration of gene expression data demonstrated the utility of the approach, which captured causative genes amongst a small number of candidates.

Advances in available genome sequence data from multiple species also holds the promise of comparative integration. Already vertebrate genome sequencing projects are completed for the chicken, dog, opossum, chimpanzee, macaque and the platypus, and in the Mammalian Genome Sequencing Project 24 species will be sequenced at two times coverage. Ultimately, this approach will provide another dimension of integration that will heighten the power of integrative genomics analysis.

Systems Comparisons with the Dairy Cow

Completion of the bovine genome sequence, and development of post-genome tools for gene expression and SNP genotyping, has driven recent advances in genomewide approaches to analysis of bovine lactation and dairy herd genetics. The total sequencing effort provided approximately seven times coverage of the genome representing about 2.7 billion base pairs (http://www.hgsc.bcm.tmc.edu/projects/bovine). The sequencing project itself resulted in the discovery of a large number of SNPs. This effort has been extended to re-sequence six breeds, Holstein, Angus, Jersey, Brahmin, Limousine and Norwegian Red, at low coverage, with the aim of identifying genomewide SNPs. In addition, mining of existing expressed sequence tags contributed a putative 17,344 SNPs [51]. Initial analysis of dairy herds using a subset of SNPs revealed the limited diversity of the predominant cattle breed used internationally for milk production, Holstein Friesian (HF) [52], and described the extent of linkage disequilibrium in this breed [53]. Now a primary bovine haplotype map has been created and dairy trait association mapping is undergoing refinement on the basis of additional SNPs [54]. An advantage of undertaking such studies in dairy cattle is the availability of large phenotype databases accompanied by detailed pedigree information and well developed genetic algorithms, based largely around indices based on traits of economic importance. We have recently conducted gene expression profiling experiments across the bovine lactation cycle, and currently we are integrating the genetic and transcriptome datasets using approaches similar to those developed for mice.

Comparison of bovine and mouse data is also providing useful data. An in silico comparison of the differentially expressed genes from three studies of the lactation cycle in different mouse strains with positional candidate genes in dairy QTL identified 212 genes that included well characterized genes associated with mammary gland development [55]. When we integrated differentially expressed genes from our own strain comparison of QSi5 and CBA mice, 28 genes were common to the two groups. Integration of genes involved in secretory activation from studies of knockout mice resulted in 93 genes common to the dairy QTL comparison, and 13 genes found in all three groups (Fig. 4).
https://static-content.springer.com/image/art%3A10.1007%2Fs10911-007-9052-6/MediaObjects/10911_2007_9052_Fig4_HTML.gif
Fig. 4

Comparing gene lists from mouse models of lactation. Secretory activation mouse models (SA DE), candidate genes for dairy QTL (cgQTL) and strain-specific differences in mammary gland (QC).

Concluding Remarks

Mouse models have provided a platform for our current understanding of mammary gland biology, and continue to play an important role in the current genomic era. Advances in mouse genomic data and in statistical approaches to the analysis of whole genome data are having a marked effect on our capacity to dissect events involved in the regulation of mammary gland development and secretory activation/lactogenesis. The continuing development of large scale genotyping and phenotyping datasets, along with the emergence of genomic data for additional species, holds the promise of an integrated systems analysis of mammary gland physiology. Inevitably, this analysis will arise from a generalization of the system, and further analysis of factors influencing phenotype will be required. Mostly, this will involve the non-heritable components that affect phenotype, especially environmental factors or, in the case of dairy cattle, production systems. Either way, mouse models are central to the development of a systems approach, which will facilitate future advances in understanding mammary gland biology.

Copyright information

© Springer Science+Business Media, LLC 2007