| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Development Group (M.J.N., S.R.O., M.G.-G., J.H., K.B., C.J.O.), Cancer Research Program, Garvan Institute of Medical Research, St. Vincent’s Hospital, Sydney, New South Wales 2010, Australia; Division of Biomedical Sciences (T.W.C.H., A.M.W.), University of California, Riverside, California 92521; and University Research Centre Neuroendocrinology (F.C.L., D.W.), Bristol University, Bristol BS2 8HW, United Kingdom
Address all correspondence and requests for reprints to: Christopher J. Ormandy, Development Group, Cancer Research Program, Garvan Institute of Medical Research, St. Vincent’s Hospital, Sydney, New South Wales 2010, Australia. E-mail: c.ormandy{at}garvan.org.au.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONRESULTSDISCUSSIONMATERIALS AND METHODSREFERENCES |
|---|
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONRESULTSDISCUSSIONMATERIALS AND METHODSREFERENCES |
|---|
Proliferative alveolar morphogenesis occurs in response to increased levels of estrogen, progesterone, and prolactin (Prl). An early increase in ductal side branching gives way to the formation of lobuolalveloar structures. During this phase the basic architecture of the gland is established, with lobuloalveoli replacing the previously predominant adipose tissue. The appearance of cytoplasmic lipid droplets signals the onset of lactogenesis, which is divided into the secretory initiation and activation phases (2).
The initiation phase, beginning around midpregnancy, results in the acquisition of limited secretory capacity. Milk protein gene expression commences in a programmed pattern such that Wdnm1 is expressed first, followed by the caseins, whey acidic protein (Wap) and lactalbumin (3). Secretory capacity is gained only by a subset of epithelial cells. Prl, progesterone, and estrogen are permissive for this stage of development, and further development is held in check by high progesterone levels controlled by the placenta (4). This occurs directly in humans via placental progesterone secretion or indirectly in mice via placental lactogen support of the progesterone-producing ovarian corpus luteum.
The secretory activation phase of lactation is triggered by falling progesterone levels, which in mice also trigger parturition, with the result that milk is immediately available to the pups. In humans, secretory activation commences after parturition and so lactation is delayed until 1 d or 2 d postpartum. Prl levels also rise in all species at this time and are essential for lactation (2). The mammary epithelium also synthesizes Prl, which is essential for the increase in epithelial cell proliferation that accompanies secretory activation (5).
Although the hormonal control of the concomitant morphological and functional events of secretory activation have been well described, the underlying alterations occurring in gene expression in the mammary gland that drive these events are not well understood. Experimental models used to examine secretory activation have been limited to the examination of the response of candidate genes to hormonal manipulation of whole animals. The combination of mouse gene knockout and microarray technology (6) now offers a new experimental approach to this question.
We have made two knockout models that experience failure of secretory activation. These are animals with a single functional Prl receptor (Prlr) allele (7) and animals in which the neuropeptide galanin (Gal) is lost (8). Loss of a Prlr allele reduces mammary Prlr expression during pregnancy (9) and, intriguingly, also results in poor maternal behavior, reducing or abolishing milk delivery to the pups (10). Gal is an autocrine/paracrine growth factor for the Prl-secreting pituitary lactotroph cells (8, 11) and, consequently, Gal–/– mice display a failure of estrogen-induced lactotroph proliferation and reduced Prl serum levels. As a result, secretory activation fails in Gal–/– mothers and the pups die (12). The levels of other pituitary hormones, such as GH, LH, FSH, and TSH are normal in Gal–/– mice. Gal and its receptors are also expressed by the mammary epithelium and we have recently demonstrated that Gal exerts a direct developmental effect via the mammary epithelium to modulate Prl action during pregnancy (13).
Several posttranslational modifications of Prl are known to occur (14) with phosphorylation of Prl being quantitatively the most important (15, 16, 17). This phosphorylation can be mimicked by mutation of the normally phosphorylated serine to an aspartate (S179D in most species). Treatment of female rats with S179D resulted in failed lactation and slower onset of maternal behavior (18, 19). The in vitro mechanism of S179D action is currently controversial, with both weak agonist and potent antagonist activities described (20, 21, 22, 23). In vivo treatment with S179D mimics a number of phenotypes previously described in Prlr–/– mice (10, 19, 24, 25, 26). For the purposes of the current study, S179D provides another model of failed secretory activation. Unlike the knockouts, its effects are exerted from the onset of pregnancy, providing a control for any undiscovered prepregnancy developmental effects in the germline models.
We used these experimental models combined with oligonucleotide arrays (Affymetrix, Santa Clara, CA; U74A2 GeneChip) to examine the alterations in the mammary gland transcriptome that result in secretory activation of the mammary gland.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONRESULTSDISCUSSIONMATERIALS AND METHODSREFERENCES |
|---|
). Treatment of Gal+/+ animals with S179D caused lactational failure in four of five females and, despite continued suckling, the stomachs of 30 pups from these four S179D-treated mice did not contain milk. Significantly, treatment of five Gal–/– females with S179D, at rates that produced the lactational failure in Gal+/+ animals, was unable to rescue lactation, and all 40 pups showed no milk in their stomachs, demonstrating no agonist activity of S179D in this assay. S179D treatment prevented the large rise in the level of ß-casein gene expression seen in Gal+/+ animals (Fig. 1B). Levels in S179D-treated animals were comparable to those measured in Gal–/– animals and Prlr+/– animals, indicating that all models passed through secretory initiation but were developmentally blocked during the secretory activation phase. A reduction in expression in response to S179D was also seen in Wdnm1 and Wap milk protein mRNAs (Fig. 1C) and 
- and ß-casein by Western blot (Fig. 1D). The comparison of milk protein content by Western analysis is complicated by the retention of colostrum in the S179D, Gal–/–, and Prlr+/– animals and its expulsion in the wild-type control animals capable of lactation in response to suckling. Colostrum has 5 times the milk protein content of milk and is seen as the strong pink staining of luminal content by hematoxylin and eosin histology in the S179D-treated glands. Morphological examination of the fourth mammary gland at the first day postpartum also showed that S179D treatment produced a block in development at the same developmental stage as that caused by loss of a Prlr allele (29) or loss of Gal (13). Compared with controls (Fig. 1E), lobules had formed, but appeared smaller and less dense than those seen in wild-type animals (Fig. 1F). Histological examination showed that, compared with vehicle-treated controls (Fig. 1G), the S179D-treated animals showed reduced alveolar density and failedlactation, seen as the pink-stained accumulation of colostrum protein within the ducts and alveoli of Fig. 1H. Together these data show that secretory activation had failed in S179D-treated animals.
|
The proportions of Stat5a in total protein were equal in the mammary glands of saline- and S179D-treated Gal+/+ mice; however, the amount of Stat5 that was phosphorylated was markedly reduced in the mammary glands of mice treated with S179D (Fig. 2A). The levels of phosphorylated and total ERK were variable among animal samples, and there was no apparent difference (Fig. 2B). Likewise, the amount of total and phosphorylated Akt, a downstream target of the PI3 kinase-signaling pathway, did not significantly change between the two groups of mice (Fig. 2B). A small diminution in the level of cyclin D1, previously reported to be up-regulated by U-Prl and down-regulated by S179D (20), was detected in some experiments.
|
). Variable levels of Stat5 phosphorylation were seen in Prlr+/– animals capable of lactation sufficient for pup survival, but these were all higher than in the animals that could not lactate. We have previously shown mammary Stat5 phosphorylation in response to Gal in culture (13).
Investigation of the Altered Patterns of Gene Expression Underlying Failed Lactogenesis Produced by Loss of Gal, by Loss of a Prlr Allele, or by Treatment with S179D
We measured the alterations in gene expression during the secretory activation phase of lactogenesis using oligonucleotide expression arrays to compare the global patterns of altered gene expression in our three models of failed lactogenesis: loss of a Prlr allele, loss of Gal, and by treatment with S179D. To perform these experiments, RNA was pooled from four to six replicate animals from each of the seven different genotypes or treatment groups, and expression profiles were obtained using MGU74Av2 Affymetrix GeneChips. The entire experiment, including all animal treatments and RNA pooling, was repeated at a later time to provide complete experimental duplication. The results are available as supplemental data published on The Endocrine Society’s Journals Online web site at http://mend.endojournals.org. To gain a broad overview of the functional groups contained within the set of genes exhibiting altered expression in association with failed lactation, we used OntoExpress (38) to identify gene ontologies with statistically significant overrepresentation in the set of genes showing altered expression associated with failed lactation. These ontologies are shown in supplemental Table 1 (published as supplemental data on The Endocrine Society’s Journals Online web site at http://mend.endojournals.org) where x denotes the significant overrepresentation of that ontology in one or more of the animal models used, and ontologies are organized into functional groups. Genes involved in cholesterol, sterol, fatty acid, and lipid biosynthesis, metabolism, and glycolysis are overrepresented. Interestingly, genes involved in immune responses are also overrepresented to a sufficient degree to be detected at this broad-overview level, probably indicative of successful preventative control of organisms with potential to cause mastitis, as we did not detect any cases of the infection. Other notable ontologies include the IGF-binding proteins, actin cytoskeleton, and arginine metabolic enzymes.
Venn Analysis of the Alterations in Gene Expression Patterns
We undertook a Venn analysis of gene expression among the models of failed lactation. We searched for genes that showed altered expression between Prlr+/+ and Prlr+/– (lactating or nonlactating) in both experimental replicates. We denoted this set as "–Prlr," the genes showing changed expression due to loss of the Prlr. Similarly, we compared Gal+/+ to Gal–/– and identified the genes changing expression due to loss of Gal, denoted "–Gal." We compared Gal+/+ treated with saline to Gal+/+ treated with S179D to identify genes that changed in response to S179D treatment denoted "+S179D." We then combined these sets in the Venn analysis shown in Fig. 3A. There were 7,278 probe sets that were not expressed in the mammary gland of the 12,488 probe sets on the chip. There were 939 probe sets that showed changed expression in at least one of the models from a total of 5210 probe sets with detectable expression. The Venn analysis shows the number of genes with increased (I) or decreased (D) expression in each subset. The squares below the Venn diagram show these data for the intersections of two sets, and the pattern of gene expression change for the 35 genes of the center intersection set can be found in Fig. 5. The false discovery rate for each set is indicated in brackets.
|
|
Prlr was clearly shown to be the dominant force in secretory activation, as loss of a Prlr allele resulted in 654 of the 939 changes detected, with a strong bias toward positive action. The expression of most of these genes (602) was not influenced by S179D, greatly limiting the role of S179D and P-Prl as global modulators of Prl action, but the fact that S179D induced failure of lactation demonstrated that it regulated key genes. Knockout of Gal regulated 109 genes in common with loss of a Prlr allele. This set is indicative of Gal control of pituitary Prl secretion (12) and Gal modulation of Prl action at the mammary epithelial cell (13).
Examination of the +S179D sets shows that S179D shares actions in common with –Gal and –Prlr (central set of 35 genes) and with Prl independent of Gal (set of 52 genes). Where +S179D exerted effects in common with –Prlr (87 genes comprised of sets containing 52 and 35 genes), the observed pattern overwhelmingly demonstrated S179D to be an antagonist of Prl action, as just three of these 87 genes increased with S179D treatment and loss of Prl signaling, whereas 84 responded to a loss of Prl signaling flux and treatment with S179D in a similar way.
Validation of Array Results by Quantitative RT-PCR (QPCR)
We used QPCR on the LightCycler platform to confirm the array results for the nine genes selected from the central set of 35. In all samples the Affymetrix increasing or decreasing calls were confirmed by QPCR, and overall it was apparent that the Affymetrix estimate of the magnitude of change was conservative. Both the Affymetrix and QPCR techniques gave highly reproducible and consistent results (Fig. 3B) as we have observed previously in other studies (13, 25, 39), allowing the use of the Affymetrix profiles as an accurate measure of gene expression level.
Comparison of Milk Protein Expression and Lipid Biosynthetic Enzymes among the Models of Failed Lactation
We compared the changes in the expression of a panel of milk proteins and lipid biosynthetic enzymes to determine whether the arrest in secretory activation had occurred at the same point in development among our models of failed lactation. We used the gene panels defined by Rudolph et al. (6) for the milk proteins (Fig. 4A) and lipid biosynthetic enzymes (Fig. 4B) in this analysis. It is clear from this comparison that all models have proceeded through secretory initiation as all models express all of the milk proteins including -lactalbumin, the last of the milk proteins to be expressed (3). All express the lipid biosynthetic machinery required for the production of lipid droplets, consistent with the histology of these models. Strikingly, the reduction in expression of these genes compared with wild-type animals is very similar among the models, with the exception of casein-
, which showed the most heterogeneous response among the models. Thus these models arrest development at very similar points during secretory activation.
|
) that were commonly altered in the three models of lactational failure represent a small set of key genes involved in lactogenesis. This is a very stringent analysis and relaxation of the conditions by requiring, for example, a gene to be represented in just one of two experimental replicates, expands the central set to many hundreds of genes. Clustering using the small set of 35 unequivocally separated those animals able to lactate from those unable to lactate. Analysis of the function of these genes by extensive literature searches showed that most were involved in the metabolic processes underlying milk production, such as the synthesis of triacylglycerols and cholesterol from glucose, transport of fatty acids from the circulation, and lactose synthesis. Although this finding is not surprising, many of the genes contained in this set have not previously been implicated in lactogenesis and they are now placed within the lactation pathway for the first time (Fig. 5). We have presented these data in a functional scheme based upon extensive literature searching, together with a heat map to indicate expression pattern, to allow a succinct summary of the large amount of information on which this figure is based. Unigene names are used and Unigene should be used as the access point to the gene structure, full gene name, and aliases, and thus the Medline publication data used to construct this figure. The Discussion contains details of the more immediately interesting genes in this set.
Genes Associated with Partial Rescue of Lactation in Prlr+/– Mice
This data set also allowed us to examine the alterations in gene expression that occur between Prlr+/– females that could not lactate and those able to lactate sufficiently for pup survival. Here we are examining the ability of segregating genetic elements within mixed 129 genetic backgrounds to rescue lactation in some individuals. Using an analysis of the data by the robust multiarray average (RMA)/penalized T Statistic method, of the top 100 genes ranked on significance of their P value for changed expression between Prlr+/– animals that lactated compared with those that did not, an extraordinary 25% were found to play a role in the initiation of DNA replication (Fig. 6). These genes are presented according to their function in the assembly of the DNA replication machinery, adapted from the review of Bell and Dutta (41). In addition to these genes, a number of G2/M phase genes were also found. As demonstrated by the heat map, the expression levels of these genes were not only higher than nonlactating animals but were higher than wild-type animals as well. Thus, the most prominent feature of lactational rescue in these animals is abnormally high expression of genes involved in cell proliferation and mitosis.
|
). The wild-type animals of both strains now formed a cluster together. This cluster was closely related to a second cluster containing the Prlr+/– duplicates that were capable of lactation. Distant to these clusters were the Prlr+/– that did not lactate and a fourth cluster containing animals treated with S179D and the two Gal–/–. The Gal–/– mice treated with U-Prl were split, indicating that the U-Prl treatment was less effective in rescuing lactation in one of the treatment groups. It is clear from this analysis that the changed pattern of gene expression produced by S179D treatment far more closely resembles the pattern of change produced by loss of Gal than loss of Prlr.
|
. When we clustered using the genes that changed in the Prlr model (–Prlr), we again produced the same four clusters, but the effects of S179D and loss of Gal were less distinguishable than Prlr genotype (Fig. 7B). Using the +S179D gene set to cluster separated the Gal–/– and S179D-treated samples from the rest (Fig. 7C), and clustering using the –Gal set (Fig. 7D) split the wild type and Prlr+/– that could lactate from the S179D and Prlr+/– that experienced lactational failure. This analysis suggests a close relationship between the transcriptional changes caused by the loss of Gal and those caused by treatment with the S179D mimic of P-Prl. We investigated the P-Prl pituitary secretion of phosphoprolactin in Gal–/– vs. Gal+/+ mice.
The Ratio of Phosphorylated to Unphosphorylated Prl Is Altered in the Pituitaries of Gal–/– Mice
We sought to determine whether expression of Gal regulates the ratio of P-Prl to U-Prl that is released by the pituitary. To eliminate the effects of hormonal status on the degree of Prl phosphorylation, male animals were used. In Gal+/+ male mice, 80.0 ± 4.1% of Prl was present in the unmodified form, whereas 20.0 ± 1.9% was in the phosphorylated form (Fig. 8). Gal–/– mice, however, had 68.9 ± 3.2% of Prl as the unmodified form and 31.1 ± 2.1% as the phosphorylated form (Fig. 1, P < 0.0001 Student’s (unpaired) t test). Thus, the relative ratio of U-Prl to P-Prl was 4:1 in Gal+/+ mice, compared with 2:1 in Gal–/– mice (Fig. 8).
|
, we used the HC11 mouse mammary cell line. These cells initially proliferate to confluent density under the influence of epidermal growth factor during the first 3 d of culture and then differentiate in response to Prl added at d 4 by forming domes and synthesizing ß-casein. Milk proteins, as an assay endpoint, are measured at d 8. We measured the expression level of the 35 genes using Affymetrix chips at d 2, d 3, d 4, and d 8. Of the 35 genes, we detected expression of 28, and their fold change in expression in response to Prl and the P value for this change are shown in Fig. 9A. Genes are named where they showed a significant P value for changed expression, or the case of Cidea and Angptl4 where they showed a large fold change but nonsignificant P value due to a low level of expression at the edge of detectability. The genes Slc39a8, Slc34a2, Car2, Aldo3, Kcnk1, Ctsc, Elovl5, and Ugalt, which showed decreased expression in the models of failed secretory activation, all showed increased expression in response to Prl in HC11 cells with significant P values. Psmb9 and the transcription factors Cebpd and Sox4, which showed increased expression in all models of failed secretory activation, showed decreased expression in response to Prl in HC11 cells. Four genes (Fdps, Acly, Gbp3, and Ctgf) showed the opposite pattern of regulation. Of the remaining 14 genes that were not directly regulated by Prl between d 4 and d 8, Erbb, Folr1, Sqle, Siat, G0s2, and Scd2 showed expression levels that increased from d 2 to d 4 and then reached a plateau, or dropped to zero (G0s2), indicating their regulation during the growth phase of these cells and confirming their role in the proliferative phase of mammary development.
|
), but recovered over d 6 to d 8. We have observed a similar rapid recovery in HC11 cells with a number of other siRNAs, indicating this may be a general feature of these cells. The level of the highly homologous Stat5B protein remained unchanged, demonstrating the specificity of this approach. Reduction of Stat5 caused a fall in the levels of the milk proteins Wdnm-1 (P = 0.018, d 4; and P = 0.06, combined d 6 and d 8) and Csnb (no expression, d 4; and P = 0.05 for combined d 6 and d 8 data) demonstrating that we had reduced Stat5 levels sufficiently to have an effect on the expression of these Stat5-regulated genes. We chose to examine whether reduction of Stat5 could reduce the levels of key lipogenic enzymes Aldo3, the key step in the conversion of glucose to pyruvate, and Scd-2, the rate-limiting step in the production of acyl-Coenzyme As. Knockdown of Stat5 significantly reduced mRNA expression of both these enzymes at d 4 (Aldo3, P = 0.06; and Scd-2, P = 0.000016), and their mRNA levels recovered as Stat5 levels recovered, with no significant expression difference at d 6–d 8. These experiments show that key points in lipogenesis fall under the control of Stat5. Overall, these experiments with the HC11 cells show the existence of direct mechanistic links between Prl, Stat5, and the 35 genes identified as common to our three models of failed secretory activation. A large proportion of the genes identified in Fig. 5 as key lactation genes fall under either the direct control of Prl or under the control of other regulatory processes during HC11 cell growth. |
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONRESULTSDISCUSSIONMATERIALS AND METHODSREFERENCES |
|---|
We have established a new mouse model of failure of secretory activation, using S179D Prl to inhibit this phase of mammary gland development. In experiments reported in Figs. 1, 2, and 4, S179D treatment during pregnancy reduced the expression of a broad range of genes involved in lactation, such as lipid and cholesterol synthesis enzymes, solute carriers, and lipid transport enzymes. Milk proteins were also dramatically reduced. Specifically, ß-casein mRNA showed a large decrease in expression in S179D Prl-treated mice compared with saline-treated mice (Figs. 1 and 4). Analysis of milk protein levels also demonstrated a reduction (Fig. 1). Together these findings indicate that lactation failed due to a failure of differentiation. This conclusion is also supported by the observation that S179D treatment reduces Stat5 phosphorylation (Fig. 3).
These experiments are consistent with a study published during the course of these investigations in which Walker and colleagues (18) reported that S179D inhibited lactation in rats. In this report Walker and colleagues used Northern blotting of S179D-treated mouse mammary HC-11 cells to document a greater induction of ß-casein relative to the ribosomal subunits in response to S179D vs. U-Prl. Although it is initially tempting to directly contrast the induction of ß-casein by S179D in vitro with the repression observed in vivo, such a comparison involves the direct effect of S179D and U-Prl on the HC-11 ß-casein promoter in vitro compared with the developmental effects of S179D and U-Prl on a complex tissue in vivo, making conclusions difficult to draw. The basis of this inconsistency remains to be investigated but could be a consequence of altered short to long receptor ratios (18) or the high levels of insulin and/or hydrocortisone used in the in vitro system.
The in vitro antagonism of Prl action by S179D has been disputed by some (22, 23) and supported by others (20), indicating that in vitro the action of S179D remains to be fully explored. This is in stark contrast to the in vivo situation in which S179D reproduces a number of phenotypes seen in the Prlr knockout model (10, 19, 24, 25, 26), demonstrating a clear Prl antagonist activity. To examine this point, we used global gene expression as an endpoint to further analyze whether S179D acted as an agonist or as an antagonist of Prl-induced alterations in gene expression in our model of S179D-induced failure of lactogenesis. We compared the pattern of altered mammary gene expression caused by treatment with S179D to the alterations produced by the loss of a single Prlr allele, and to the alterations caused by the loss of Gal. All three models exhibit a phenotypically indistinguishable failure of lactogenesis. There were 87 genes (35 + 52) that showed altered expression in response to loss of a Prlr allele and treatment with S179D. Of these, 75 showed decreased expression and nine showed increased expression, in response to loss of a Prlr allele and treatment with S179D, a pattern demonstrating S179D as an antagonist of Prl action. Two genes showed increased expression with S179D, but decreased expression in response to the loss of a Prlr allele, and one gene showed the inverse pattern; both are patterns that indicate S179D was acting as a Prl agonist. Thus S179D is overwhelmingly acting as an antagonist of Prl action on lactogenesis but, at the level of the expression of specific genes, it has detectable agonist activity for three genes. This conclusion must also be viewed in the light of total Prl action. Overall, the loss of a Prlr allele affected the expression of 654 genes, of which just 87 (13%) were altered by S179D. Given the failure of lactogenesis in S179D-treated animals this 13% was clearly a functionally important subset of Prl-regulated genes, but this small subset indicates either that S179D is not a global antagonist of Prl action or that there are developmental effects of the absence of one Prlr allele that result in the disparity. Given the induced failure of lactation, it cannot be argued that this small subset simply results from a suboptimal dose of S179D resulting in submaximal U-Prl antagonism, or a partial agonist activity of S179D.
Another aspect of this comparison was the discovery of 128 genes that responded to S179D treatment, but not to loss of a Prlr allele or loss of Gal. This set contains a high false discovery rate (102), demonstrating that most of these genes are false positives. Thus S179D has a very restricted unique activity (off-target activity). The same caveat applies to the 128 (110 + 18) genes specific to the Gal–/– set that could represent a direct action of Gal that is independent of Prl action, if not for the high false discovery rate. This indicates this effect is small and that almost all of Gal action during secretory activation is via modulation of Prl action by 1) control of serum Prl levels (12), 2) Gal modulation of Prl action at the mammary epithelial cell (13), and 3) possibly also via regulation of Prl phosphorylation (Fig. 8). It is during the transition from the proliferative to secretory initiation phase at midpregnancy that Gal serum levels and mammary Gal receptors are at their highest, conditions most suitable for direct Gal action independent of Prl (13).
To further examine the similarity between the effects of S179D, the loss of Gal, and the loss of a Prlr allele we examined our profiles using hierarchical clustering to group them based on the similarity of changes in their transcript profiles. Whereas the Venn analysis was based on a change in gene expression among our models of failed lactation irrespective of magnitude, hierarchical clustering groups experimental replicates together based upon the level of gene expression, allowing an estimate of overall similarity between expression profiles from their relative position in the computed dendogram. We also used principal-components analysis to cluster the experimental replicates, with very similar results to those found with hierarchical clustering. Gal–/– and S179D-treated glands consistently fell within the same cluster. This approach showed that the pattern of gene expression found in glands experiencing lactational failure due to S179D treatment was very similar to the pattern seen in nonlactating glands from Gal–/– mice and that both were distant from the pattern of gene expression seen in nonlactating Prlr+/– mice. This suggested that alteration in the ratio of U-Prl to P-Prl may form part of Gal’s modulation of lactogenesis.
A functional analysis of the 35 genes common to all three models of failed lactation was undertaken by extensive literature searches, allowing almost all of these genes to be placed in the model of a lactating mammary epithelial cell presented in Fig. 6. This analysis implicated a number of genes in lactation for the first time. Examples include Cidea, a key metabolic gene (43), and the ubiquitin-like Isg15, which is known to prolong the activity of Stat family members by conjugation (44). The key transcription factors Srebf1, controlling lipid metabolism genes (45), Cebp, involved in lipogenic responses and mammary development (46), and Sox4, a progesterone-responsive transcription factor (47), were also found in this set. The latter two transcription factors were expressed at a higher level in nonlactating animals, and our results suggest that their loss of expression is required for secretory activation. Other genes showing this pattern included Erbb3, two probe sets interrogating Angptl4 expression, which is an inhibitor of lipoprotein lipase (48), proangiogenic Ctgf (49), and antiangiogenic Thbs1 (50). A large number of genes have been implicated in the process of secretory activation by a detailed time-course study using wild-type FVB mice (6), and a comparison of data sets will allow Prl-regulated genes to be distinguished from those regulated by the developmental process in general. We used the HC11 mammary epithelial cell model to show that a large proportion of these 35 genes were directly regulated by Prl, and we knocked down Stat5A expression to demonstrate that two key genes in de novo synthesis of lipids from glucose, Aldo3 and Scd2, were responsive to the levels of Stat5A expression, providing a further mechanistic link between our observations and the endocrine control of secretory activation.
Our data sets also allowed us to examine the changes in gene expression that resulted in lactation in Prlr+/– mice. We detected a very strong proliferation signal in lactating Prlr+/– mammary glands, with almost all of the genes involved in the initiation of DNA replication (41), and a number from the G2/M phase of the cell cycle, showing elevation in expression not only above nonlactating Prlr+/– glands but also above Prlr+/+ glands. Notable among this set was the elevated expression of proliferating cell nuclear antigen and Ki67, widely used markers of proliferation. Searching for a potential cause for this we found that epidermal growth factor (Egf) was 4-fold higher and ras family members Rab-18 and K-ras were 2-fold higher in lactating glands. As the Egf signaling pathway results in Stat5 phosphorylation, this provides a potential Prl-independent growth factor signal that could account for the increased levels of Stat5 phosphorylation seen in lactating Prlr+/– animals. Other genes found to be elevated in lactating Prlr+/– glands included many of the key lactational genes shown in Fig. 5. Another gene elevated in this group was synuclein-
, otherwise known as persyn or breast cancer-specific gene 1, the increased expression of which is associated with aggressive breast cancer, increased metastasis, and activation of estrogen-driven transcription (51). Interestingly, apoptosis genes were not prominent in this group.
In summary, treatment of mice with a molecular mimic of phosphorylated Prl resulted in failed lactation and impaired lobuloalveolar development that was associated with reduced Stat5 activation in the mammary gland. Transcriptome analysis of the secretory activation phase of mammary gland development using three models of failed lactation identified potential key regulatory genes for this process. Transcript profiling showed that S179D has actions that are predominantly, but not exclusively, antagonistic to U-Prl-regulated gene expression in the mammary gland. Increased cell proliferation was observed in Prlr+/– females able to lactate, providing mechanistic insight into the partial penetrance of this phenotype. Prl treatment of HC11 cells demonstrated that many of the 35 key genes were under direct Prl regulation and that others were associated with mammary epithelial cell proliferation. Stat5 mediated Aldo3 and Scd2, key enzymes in de novo lipid biosynthesis from glucose. Together these results provide a small list of key genes involved in secretory activation for which a number of applications can be envisaged. For example, these genes provide an excellent starting point for the identification of alleles that may provide enhanced lactational performance in a marker-assisted selection process in commercially valuable agricultural species. Alternatively, their study may help our understanding of lactation failure and other disorders of the breast.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONRESULTSDISCUSSIONMATERIALS AND METHODSREFERENCES |
|---|
Two-Dimensional PAGE
Following decapitation the anterior pituitary was removed, cut into 1-mm pieces, rinsed in PBS to remove material from damaged cells, and incubated in DMEM containing 0.1% BSA for 2 h at 37 C in an atmosphere of water-saturated 5% CO2. At the end of the 2-h incubation period, the medium was removed and frozen before preparation for two-dimensional gel analysis. Two pituitaries were used per 2 ml of incubation medium to allow sufficient Prl accumulation in the samples from the Gal–/– mice. The proteins in the incubation medium were precipitated in 4 vol of –20 C acetone overnight, collected by centrifugation, and then dissolved in urea lysis buffer containing 9 M urea, 5% 2-mercaptoethanol, 4% ampholines (pH 4–6.5) (Sigma Chemical Co., St. Louis, MO). Electrophoresis was performed according to the method of Ho et al. (17). After electrophoresis the gel was silver stained (52), and the spots were identified by reference to standards as described previously (53) and by reference to a corun sample that was subject to Western blot analysis (53). Spot intensity was analyzed using a Kodak image analysis system (Eastman Kodak Co., Rochester, NY).
Phosphorylated and Unmodified Prl Treatment of Mice
On the morning of the observation of a vaginal plug, 6- to 8-wk-old mice were implanted with a 0.25 µl/h, 28-d Alzet miniosmotic pump (Alzet Osmotic Pumps Durect Corp., Cupertino, CA) containing either U-Prl or S179D, the molecular mimic of P-Prl; both hormones were prepared as described elsewhere (21). Either 0.6 or 1.2 µg was delivered per 24 h. On the first day postpartum, maternal behavior of mothers was observed, pups were examined for the presence of milk, and glands were taken for histological analysis.
Histological Analysis
Mammary whole mounts were made by spreading the gland on a glass slide before fixing in a 10% formalin solution. Glands were defatted in acetone before carmine alum (0.2% carmine, 0.5% aluminum sulfate) staining overnight. The whole mount was dehydrated using a graded ethanol series followed by xylene treatment for 60 min and storage and photography in methyl salicylate (54).
mRNA Isolation
The fourth inguinal mammary gland was frozen in liquid nitrogen before storage at –80 C before use. Total RNA was extracted using TRIZOL Reagent (Life Technologies, Gaithersburg, MD) according to the manufacturer’s instructions.
QPCR
QPCR was performed using LightCycler technology (Roche Clinical Laboratories, Indianapolis, IN). Primers were designed on the basis of mismatch to other genes. PCR reactions were performed in 10-µl volumes with 1 µl cDNA, 5 pmol of each primer, and FastStart DNA Master SYBR Green I enzyme mix (Roche) as per manufacturer’s instructions. Relative quantitation of the product was performed by comparing the crossing points of different samples normalized to an internal control (ß-actin). Each cycle in the linear phase of the reaction corresponds to a 2-fold difference in transcript levels between samples. Each reaction was performed in duplicate using pooled RNA from the three to six mammary glands per experiment.
mUgalt2 F a GGTGGTTGGAATAGAAGAGCACAC
mUgalt2 R a CAAGACCGAGACCCAGGAAAAC
mFolr1 F a TGGAGTTGGCGATTAGAGTCTGAC
mFolr1 R a GAGGCAGGTGTCTTGGATAAAGTG
mSiat1 F a TGTAAAATGGGGGTGACAATCC
mSiat1 R a CTCTTGCTGACCTCTTGAAGGAAC
mCyp51 F a AAAGGTAATGGGGTCGTGTAGTTG
mCyp51 R a GCACAGAATACGGGCAATGATAC
mCuta F a TGTCCCAACGAAAAAGTCGC
mCuta R a AAAGGCATCAGGAGCAGGAGAG
mCopz1 F a CAGCACAAGTGGGTTTGGAGTG
mCopz1 R a TGAGGAGAAGGAACACGGCAAG
mCsnd F a TATTACCCATCTACCCCCAGCC
mCsnd R a GAAACCCACAAGCAGACCTAACAC
mCsnb F a TTCACCTCCTCTCTTGTCCTCCAC
mCsnb R a GGGGCATCTGTTTGTGCTTG
mWDMN1 TGACAATGACTACTGCCTGGGC
mWDMN1 TTCCAAAACTGCGTGGGGGC
mWAP F a TGCCTCATCAGCCTCGTTCTTG
mWAP R a CTGGAGCATTCTATCTTCATTGGG
mCIDEA F a GACTTCCTCGGCTGTCTCAATG
mCIDEA R a GAAACTGATTCGTATCCACGCAG
mErbb3 F a TCTACCAAGTGGAACAGGAGAGGC
mErbb3 R a CACCAACAAACGGAGTCTGGAAG
mKeratin18 F a CAAGATCATCGAAGACCTGAGGGC
mKeratin18 R a TGTTCATAGTGGGCACGGATGTCC
mAldo3 F TGCCAGTATGTTACAGAGAAGGTCC
mAldo3 R CCGCTTGATAAACTCCTCAGTAGC
mScd2 F GCTGGGGCGAGACTTTTGTAAAC
mScd2 R TGGCTTCTGGAACAGGAACTGC
mStat5a F CACAGGTGGAAGATTGGGGTTC
mStat5a R CCACTCCCCATCCAAAAACC
mStat5b R CGAATGGAGAAAAGGGATGGTG
mStat5b F GTTCCTCTGCCAGGTAGTCCATAG
Western Analysis
Following RNA extraction from mammary glands using TRIZOL Reagent, protein was extracted according to the manufacturer’s instructions. Protein was separated using SDS-PAGE (Bio-Rad Laboratories, Hercules, CA), transferred to polyvinylidine difluoride (Millipore Corp., Bedford, MA), and blocked overnight with 2% fetal bovine serum, 50 mM sodium phosphate, 50 mM NaCl, and 0.1% Tween 20. Membranes were incubated with one of the following primary antibodies: -milk protein (Accurate Chemical & Scientific Corp., Westbury, NY), -Stat5A (Upstate Biotechnology, Inc., Lake Placid, NY), -phospho-Stat5, -phospho-Erk1/2, -Erk2, -phospho-Akt (S473), -phospho-Akt (T308), -Akt (Cell Signaling Technology, Beverly, MA) or -ß-actin (Sigma). Protein (20 µg) was loaded per lane except for -milk protein where 400 ng of protein was loaded. Specific binding was detected using horseradish peroxidase-conjugated secondary antibodies (Amersham Biosciences, Arlington Heights, IL) with Chemiluminescence Reagent (PerkinElmer, Norwalk, CT) and Biomax Light Film (Eastman Kodak).
Transcript Profiling
Total RNA was extracted using TRIZOL Reagent (Invitrogen Life Science, Carlsbad, CA), purified using RNeasy Mini Kit (QIAGEN, Chatsworth, CA); cDNA synthesis was performed using Superscript II (Invitrogen Life Technologies), and synthesis of biotin-labeled cRNA was performed using BioArray HighYield RNA Transcript labeling kit (Enzo Life Sciences, Farmingdale, NY) and hybridized to Affymetrix MGU74Av2 GeneChips overnight according to manufacturer’s instructions. Arrays were performed in duplicate using four to six glands per treatment group from two separate replicate experiments. Analysis was performed using the Affymetrix GeneChip version 5 software (MAS 5). Data were also analyzed as follows: Signal intensities of each gene were obtained using the RMA function in the Affy package in R (http://www.bioconductor.org). The RMA function, which involves quantile normalization of oligonucleotide signals followed by estimation of the average perfect match signal intensity for each probe set, has been shown to reduce variability and bias when compared with the MAS 5.0 software (40, 55). Differential expression was then assessed by ranked penalized t statistics using lm.series and ebayes functions in the limma package in R (http://www.bioconductor.org).
HC11 Cell Culture
HC11 cells (a kind gift from Dr. Nancy Hynes) were maintained in RPMI 1640 medium (Invitrogen) with 10% heat-inactivated fetal calf serum, 5 µg/ml insulin, and 10 ng/ml Egf (Sigma). Differentiation assays were performed by plating 1.5 x 105 cells in a six-well plate in the above medium for 3d, during which time the cells grow to confluent density, followed by incubation in media without EGF for 24 h and then treatment with 10–6 M dexamethasone (Sigma) and 5 µg/ml ovine Prl (Sigma) in media without EGF for 4 d from d 4 to d 8.
siRNA Transfection of HC11 Cells
siRNA molecules were synthesized using the Silencer siRNA Construction Kit (Ambion, Inc., Austin, TX). Two microliters of 500 nM siRNA (final concentration of 1 nM) was mixed with 200 µl serum free RPMI, 2.5 µl Lipofectamine 2000 (Invitrogen), and 200 µl OPTI-MEM media (Invitrogen) according to the Oligofectamine protocol (Invitrogen). Serum free medium (800 µl) and 200 µl of the siRNA/Lipofectamine complexes were added to each well of a six-well plate 24 h after plating the cells, and 4 h later 500 µl of RPMI/30% fetal calf serum was added to each well.
|
FOOTNOTES |
|---|
Present address for M.J.N.: Developmental Biology Program, Victor Chang Cardiac Research Institute, 384 Victoria Street, Darlinghurst, New South Wales, 2010, Australia.
First Published Online February 10, 2005
1 M.J.N and S.R.O. contributed equally to this work. 
Abbreviations: Egf, Epidermal growth factor; Gal, galanin; GFP, green fluorescent protein; PI3, phosphatidylinositol 3; P-Prl, phosphorylated Prl; Prl, prolactin; Prlr, Prl receptor; QPCR, quantitative RT-PCR; RMA, robust multiarray average; siRNA, short interfering RNA; Stat, signal transducer and activator of transcription; U-Prl, unmodified Prl; Wap, whey acidic protein.
Received for publication July 1, 2004. Accepted for publication February 2, 2005.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONRESULTSDISCUSSIONMATERIALS AND METHODSREFERENCES |
|---|
