Constitutive expression of microRNA-150 in mammary epithelium suppresses secretory activation and impairs de novo lipogenesis

During post-pubertal mammary gland (MG) development, coordinated transcriptional changes directed primarily by estrogen, progesterone and prolactin and their respective receptors govern progression through pregnancy to lactation and involution (Rudolph et al., 2003). During these stages, mammary epithelial cells (MEC) undergo massive proliferatory expansion, differentiation and secretory activation, followed by apoptosis as previously reviewed (Anderson et al., 2007).

Differentiation occurs during mid to late pregnancy in the mouse, when the MECs become competent to synthesize milk, but importantly, actual secretion is held in check until parturition when secretory activation is triggered (Rudolph et al., 2003; Anderson et al., 2007). This trigger results in copious synthesis of milk comprising milk proteins, sugars and fats. Secretory activation may be regulated, in part, at the translational level by expression of microRNAs (miRNAs) that attenuate protein synthesis until biosynthetic enzymes and secretory machinery are needed. Avril-Sassen et al. (2009) performed the first global expression array of miRNAs from whole MG lysates throughout postnatal development and found that the three most populated miRNA expression patterns shared a dramatic decline at the transition between pregnancy and lactation.

Whereas the Caldas group identified distinct patterns of miRNA expression during MG development, it is possible that other cell types such as mammary adipose, lymphatic and immune cells (Rudolph et al., 2009) contributed to the patterns observed. Therefore, to identify MEC-specific miRNAs with the potential to regulate secretory activation and the biosynthetic processes of milk production, we performed simultaneous miRNA and messenger RNA (mRNA) profiling on isolated MECs from mice at pregnancy day (P)14 and lactation day (L)2. We identified a number of miRNAs that decrease in unison between P14 and L2, including miR-150-5p, which had the largest fold-decrease. The coordinated decrease in multiple miRNAs was coincident with the reciprocal upregulation of numerous metabolic pathways crucial for lactation, including lipid modification and fatty acid synthesis noted previously (Rudolph et al., 2003). To test the functional involvement of miR-150 in secretory activation, we utilized transgenic mice that constitutively express miR-150 driven by the whey acidic protein (WAP) promoter in MECs throughout mid-pregnancy and lactation. Our finding that constitutive expression of miR-150 leads to lactation deficiency and pup mortality suggests that the decline in miR-150-5p expression in MECs at late pregnancy contributes to secretory activation, induction of lipogenic mRNA and proteins, and robust activation of de novo lipid synthesis in lactation.

To identify candidate miRNAs that regulate secretory activation specifically in MECs following functional development, simultaneous global mRNA and mature miRNA profiling was performed (analyzed data in Table S1). Thirty-two differentially expressed miRNAs displayed a ≥twofold change between P14 and L2 (Fig. 1A), and the majority (∼80%) were reduced. The observed decreases include miR-150-5p, miR-17-5p and miR-425-5p, which were validated by qRT-PCR (Fig. 1B). In addition, MEC­-specific decreased expression of miR-150-5p in mammary epithelium at L2 compared with P14 was verified by in situ hybridization (ISH) (Fig. 1C).

MEC-specific mRNAs were profiled by microarray and transcripts significantly different between P14 and L2 were identified using an unpaired t-test (analyzed data in Table S2). Ingenuity Pathway Analysis (IPA) on upregulated transcripts identified key pathways involved in secretory activation, particularly cholesterol and lipid biosynthesis (Fig. S1), were increased between pregnancy and lactation, as would be expected. Twenty-three of the 25 significantly decreased miRNAs between P14 and L2 (Fig. 1A) were predicted to target transcripts of genes involved in lipid synthesis, which increased between P14 and L2 and included key members of the de novo fatty acid synthesis pathways as well as several fatty acid desaturase genes (Table S3). Notably, miR-150-5p is predicted to target more lipid synthesis genes than most other miRNA (Table S3). There were 242 TargetScan­-predicted miR-150-5p targets significantly upregulated by at least 1.5-fold or greater at L2 versus P14 (top three fold-changes shown in Fig. 2A). Because v-myb avian myeloblastosis viral oncogene homolog (Myb) is a target of miR-150-5p in B-cells (Xiao et al., 2007), MYB protein was examined by immunohistochemistry (IHC). Although the array data found Myb mRNA to be unaltered, MYB protein was increased in mammary epithelium at L2 compared with P14 (Fig. 2B, top), suggesting that a decrease of miR-150-5p might relieve the inhibition on Myb mRNA translation at lactation, allowing MYB protein to be expressed. As the fatty acid synthase gene (Fasn) was increased between P14 and L2 and was a predicted target, the protein level of FASN was also confirmed by IHC (Fig. 2B, bottom). IPA analysis of the 242 upregulated predicted targets of miR-150-5p revealed that lipid and cholesterol biosynthesis pathways at secretory activation were among the most significantly enhanced pathways correlated with the decline in miR-150-5p at L2 (Fig. 2C).

ISH for miR-150 on C57BL/6 mouse MGs collected previously at days P5, P12, P17 and L1 (Spoelstra et al., 2006) indicated that miR-150-5p expression was low early in pregnancy (P5), increased mid to late pregnancy (between P12 and P17), and decreased just after secretory activation at L1 (Fig. S2A). To investigate a role for the rise in miR-150-5p at mid-pregnancy, we examined previously published time-course microarray data (GEO record GSE4222) (Rudolph et al., 2007) (Fig. S2B). Genes crucial for lactation, such as miR-150-5p predicted targets Elovl5 and Fads1, are expressed at low levels until late pregnancy when they increase (Rudolph et al., 2007) coincident with the decline of miR-150-5p. Conversely, validated miR-150 targets Egr2 (Bousquet et al., 2013) and Myb (Xiao et al., 2007) decreased at mid-pregnancy when miR-150 increased (Fig. S2B).

The decrease in miRNAs such as miR-150-5p at L2 might allow for maximal and accurately timed expression of proteins important for secretory activation. To test this possibility, we used a transgenic mouse model designed to override the natural decline in miR-150 by constitutively expressing miR-150 in the mammary epithelium. Whey acidic protein-driven Cre recombinase (WAP-Cre⁺) transgenic mice (Wagner et al., 1997) were crossed with Rosa-26-flox-Stop-flox-miR-150^fl/− (Stop-150^fl/−) transgenic mice (Fig. 3A,B) created previously (Xiao et al., 2007). Sustained expression of mature miR-150-5p following secretory activation in L2 MECs from WAP-Cre⁺; Stop-150^fl/fl mice was confirmed by qRT-PCR (Fig. 3C). To determine if the miRNA processing machinery was potentially overburdened by constitutive synthesis of miR-150-p, miR-146b-5p, a miRNA that increased at L2 compared with P14 in wild-type mice (Fig. 1A), was analyzed and found to be unaffected by constitutive expression of miR-150 (Fig. S3).

Compared with offspring nursed by control dams (WAP-Cre⁺; Stop-150⁻), pups nursed by WAP-Cre⁺; Stop-150^fl/fl dams had significantly decreased survival by postnatal day (PND)3, which continued to diminish throughout lactation (Fig. 3D, top; P<0.0001). Surrogate litters from control dams confirmed the lactation defect persisted beyond PND3 as surrogate mortality was also significantly higher when nursed by WAP-Cre⁺; Stop-150^fl/fl dams relative to control dams (Fig. 3D, bottom; P<0.0001). Additional evidence of the lactation defect in dams constitutively expressing miR-150 was observed in the milk spots of surviving pups at PND2, which were notably smaller in pups nursed by WAP-Cre⁺; Stop-150^fl/fl dams (Fig. 3E) These data indicate constitutive expression of miR-150-5p in mammary epithelium results in a severe lactation deficiency with death of over 50% of offspring by PND3.

To understand lactation defects at the cellular level, histological MG sections from P18 and L2 were evaluated for morphological changes. MGs from WAP-Cre⁺; Stop-150^fl/fl mice had no difference in alveolar density as late into pregnancy as P18 when compared with that from control mice (Fig. 4A). Furthermore, IHC for the lipid droplet-binding protein adipophilin indicated that WAP-Cre⁺; Stop-150^fl/fl mice have equivalent lipid droplet size and abundance at P18 (Fig. S4). In contrast, there was decreased alveolar density between genotypes at L2 (Fig. 4A).

Reverse phase protein array (RPPA) was used to examine a variety of total and phosphorylated signal transduction proteins in MEC lysates from control and WAP-Cre⁺; Stop-150^fl/fl mice at both P14 and L2 (Fig. 4B). The majority of proteins differentially expressed between genotypes, including the phosphorylated forms of JAK1 (Y1022/Y1023), FAK (also known as PTK2) (Y397), and ER-alpha (also known as ESR1) (S118), were decreased by constitutive expression of miR-150 only at L2 (Fig. 4B), but there was no significant effect on protein expression levels at P14 (Fig. 4B; Table S4).

To explore the cause of morphological changes observed in the bitransgenic mice at L2 (Fig. 4A), histological sections were quantified for nuclei count and area (Fig. 4C). Because there was no difference in nuclei count (Fig. 4C, left), but epithelial area normalized to nuclei count was significantly reduced (Fig. 4C, right; P=0.03), the difference in observed alveolar density is due to the epithelium in control mice becoming distended with milk, whereas epithelium in bitransgenic mice was much less distended.

Prolactin receptor (PRLR) signaling mediated through janus kinase 2 (JAK2) and signal transducer and activator of transcription 5 (STAT5) is essential for lactation (Rudolph et al., 2011). Both total STAT5B (a TargetScan-predicted target of miR-150-5p) and phospho-STAT5 were significantly reduced in L2 MECs from WAP-Cre⁺; Stop-150^fl/fl mice compared with controls (P=0.006 and P=0.005, respectively), but STAT5A protein was not affected (Fig. 4D). The ability of miR-150-5p to target the Stat5b 3′ untranslated region (UTR) was tested with a Stat5b-3′-UTR-luciferase reporter containing the predicted miR-150-5p binding sites (Fig. 4E). Co-transfection of miR-150-5p mimic with plasmid containing the wild-type sequences of the two predicted miR-150-5p binding sites resulted in a significant decrease in reporter activity, but not co-transfection with the Stat5b-3′ UTR-mut plasmid in which residues crucial to miR-150-5p binding were mutated at both sites (Fig. 4E; P=0.0003 and P=0.09, respectively). This demonstrated that mouse Stat5b is a bona fide target of miR-150-5p. Importantly, activated and total janus kinase 2 (JAK2) were similar in L2 MECs from WAP-Cre⁺; Stop-150^fl/fl mice compared with controls, suggesting there was normal PRLR activation upstream of STAT5 (Fig. S5). Thus, our data indicate that targeted reduction of STAT5B, resulting from constitutive expression of miR-150, might impair secretory activation specifically through reduced total STAT5 activity.

Because STAT5 contributes to survival of differentiated mammary epithelium (Cui et al., 2004), cell death was investigated by IHC to detect cleaved caspase 3 (CC3) in sections of L2 MGs from WAP-Cre⁺; Stop-150^fl/fl mice. Although less than 0.2% of MECs were CC3-positive, CC3 staining was significantly higher in MGs from WAP-Cre⁺; Stop-150^fl/fl mice compared with controls (Fig. 4F; P=0.0001), indicating that constitutive expression of miR-150 caused some increased cell death in mammary epithelium.

Global mRNA profiling identified significant gene expression changes due to constitutive expression of miR-150 in MECs (analyzed data in Table S5). In agreement with the bioinformatic predictions shown in Fig. 2C, more than half of the genes involved in lipid and cholesterol synthesis had a ≥twofold decrease due to constitutive expression of miR-150 at L2 (Fig. 5). Of the 22 listed lipid and cholesterol synthesis genes with a ≥twofold decrease, all but two are predicted to be direct targets of miR-150 (TargetScan and RNA22). Notably, these predicted targets include the primary enzymes of the de novo fatty acid synthesis pathway, including ATP citrate lyase (Acly) (Linn and Srere, 1979), acetyl-CoA carboxylase (Acaca, both isoforms) (Harada et al., 2007), Fasn (Smith et al., 2003), and thyroid hormone responsive protein spot 14 (Thrsp) (Rudolph et al., 2014). Further, mitochondrial citrate transporter (Slc25a1), required to shuttle substrate into the de novo fatty acid synthesis pathway (Kaplan et al., 1993) also decreased. Transcripts of genes that modify fatty acids including stearoyl-CoA desaturase 1 (Scd1), fatty acid elongases (Elovl5 and Elovl6), and the fatty acid desaturase (Fads1) were also decreased by miR-150 expression.

FASN is absolutely essential in mammals to synthesize de novo fatty acids from acetyl-CoA and malonyl-CoA substrates (Smith et al., 2003). Because Fasn and Acaca, which encodes the enzyme supplying substrate to FASN, were decreased significantly at L2 in mice with constitutively expressed miR-150 (Fig. 5), protein levels were evaluated by immunoblot and IHC. Immunoblot analysis indicated an ∼85% (P=0.0001) suppression of FASN protein in L2 MEC lysates from WAP-Cre⁺; Stop-150^fl/fl mice compared with control MECs (Fig. 6A). In addition, IHC analysis revealed that whereas FASN levels in the epithelium of control mice increased dramatically from P14 to L2 (Fig. 6B, top), FASN was inhibited at L2 in WAP-Cre⁺; Stop-150^fl/fl mammary epithelium, but not in mammary adipose (Fig. 6B, bottom). This result indicates epithelial cell­-specific expression of exogenous miR-150 coincided with MEC-specific suppression of FASN. The ability of miR-150-5p to target a predicted site in the Fasn 3′ UTR was tested using a Fasn-3′-UTR-luciferase reporter assay (Fig. 6C). Co-transfection of miR-150 mimic resulted in a significant decrease in reporter activity with reporter containing the wild-type Fasn 3′ UTR miR-150-5p target site, but not with the Fasn-3′-UTR-mut plasmid, in which residues crucial to miR-150-5p binding were mutated (Fig. 6C). This demonstrated that mouse Fasn is a bona fide target of miR-150-5p. Whereas ACACA (a RNA22-predicted target of miR-150-5p) increased at the protein level between P14 and L2 in MECs of control mice (Fig. 7A, top), epithelial expression of ACACA was significantly decreased in L2 glands from WAP-Cre⁺; Stop-150^fl/fl mice as shown by immunoblot analysis of MECs (∼80% decrease; Fig. 7B) and IHC (Fig. 7A, bottom). Although the transcript for oleoyl-ACP hydrolase (Olah) did not decrease with constitutive expression of miR-150 (Fig. 5), Olah is a predicted target of miR-150-5p (TargetScan) that is essential for synthesis of medium chain fatty acids (MCFA) in MECs (Smith, 1980). OLAH protein was suppressed by ∼60% in L2 MECs from WAP-Cre⁺; Stop-150^fl/fl mice compared with control mice as quantified by immunoblot (Fig. 7C). Collectively, these results demonstrate that expression of multiple de novo fatty acid synthesis pathway components are suppressed by constitutive expression of miR-150.

To test the functional effects of the reduced levels of multiple proteins in the de novo fatty acid synthesis pathway, amounts of fatty acids were quantified from MECs using gas chromatography-mass spectrometry (GC-MS). MCFAs are known to be exclusively synthesized by the de novo pathway in MECs, owing to MEC-specific expression of OLAH (Smith, 1980). Quantitative lipid mass spectrometry revealed a significant reduction in MCFA, including 10:0, 12:0, and 14:0 (Fig. 8A; P<0.05, P<0.01 and P<0.05, respectively) as well as their sum (inset) in MECs that constitutively express miR-150 compared with control MECs. Furthermore, 16:0, which can originate from either MEC de novo synthesis or be taken up from the serum, was also significantly decreased in MECs from mice with constitutive expression of miR-150 compared with controls (Fig. 8A; P<0.05). Production of 18:0 in MECs can result from elongation of 16:0 and was also significantly reduced in MECs from mice with constitutive expression of miR-150 compared with control mice (Fig. 8A; P<0.01). Although the classes of monounsaturated fatty acids (MUFA) and polyunsaturated fatty acids (PUFA) were unchanged (Fig. 8B) in MECs from mice with constitutive expression of miR-150 compared with control mice, a significant increase in 22:4 n-6 (P<0.01) and a nearly significant increase in 22:5 n-3 (P=0.052) were observed (Fig. 8C). Furthermore, fatty acids known to be taken up by MECs exclusively from dietary sources were not different (Fig. 8D). Collectively, quantitative GC-MS exposed the significant suppression of de novo synthesized fatty acids, further indicating that miR-150-5p profoundly modulates this enzymatic pathway in vivo.

We identified a variety of miRNA in differentiated MECs, including miR-150-5p, that decrease at the transition from pregnancy to lactation. These decreases in miRNA occur reciprocally with multiple mRNA increases in MECs, suggesting that the decline in miRNA could serve to relieve repression of mRNAs crucial for secretory activation and effective lactation. To test this hypothesis we constitutively expressed miR-150-5p, the miRNA with the highest fold-decrease between L2 versus P14, to override its natural decline. Our prediction was that transcripts targeted by miR-150-5p, which would normally be translated when this miRNA declines, would continue to be translationally inhibited with forced miR-150 expression, resulting in impaired lactation. Indeed, constitutive expression of miR-150-5p suppresses transcripts that code for proteins important for de novo fatty acid synthesis (FASN, ACACA and OLAH) and cell survival (STAT5B) in mammary epithelium. Our findings demonstrate that miR-150-5p directly targets the 3′ UTRs of mouse Fasn and Stat5b at predicted binding sites, suggesting regulatory control of two pathways crucial for lactation. Cumulatively, when miR-150-5p is constitutively expressed, synthesis of de novo fatty acids is dramatically reduced, which affects fatty acid content and results in severe lactation deficit and high pup mortality.

Consistent with miR-150-5p targeting Stat5b (Fig. 4D,E), constitutive expression of miR-150 results in a phenotype similar to that of Stat5b knockout mice (Udy et al., 1997). Stat5b knockout mice had impaired milk production and a high incidence of perinatal pup death. As STAT5 activity is important for survival of differentiated MECs (Cui et al., 2004), it is possible that targeted reduction of STAT5B protein by constitutive expression of miR-150 and consequent decreased total STAT5 activity contributed to decreased MEC survival at lactation. Reduced alveolar density (Fig. 4A,C) and a slight increase in cleaved caspase 3 in L2 MGs with constitutive expression of miR-150 (Fig. 4F) were consistent with this theory. Constitutive expression of miR-150 does not alter signaling upstream of STAT5B as no differences were observed in activation of JAK2, a key mediator of PRLR signaling (Fig. S5, Table S4). Prlr mRNA was not suppressed by constitutive expression of miR-150 (Fig. 5). Suppression of STAT5B had no effect on the transcripts encoding β-casein and whey acidic protein (WAP) (Fig. S5), likely because STAT5A, activated during secretory activation (Nevalainen et al., 2002), was not affected by constitutive expression of miR-150 (Fig. 4D). In contrast, suppression of both milk protein and lipogenic genes does occur when both STAT5A and STAT5B are repressed via inhibition of pituitary prolactin secretion (Rudolph et al., 2011; Naylor et al., 2005), or dominant-negative prolactin mimetic (Naylor et al., 2005).

We found that constitutive expression of miR-150-5p suppressed lipid synthesis mRNAs at secretory activation (Fig. 5). It has been suggested that lipogenic gene induction is mediated in part by STAT5 activity (Rudolph et al., 2011). Therefore, it is possible that downregulation of STAT5B and total STAT5 activity by miR-150 contributes to suppression of lipogenic genes at the transcriptional level. However, our data strongly support an additional layer of miR-150-5p-mediated post-transcriptional control of multiple enzymes essential for de novo fatty acid synthesis such as FASN, ACACA and OLAH (Figs 6 and 7).

FASN synthesizes all de novo fatty acids (Smith et al., 2003), ACACA is the primary rate-limiting enzyme in the de novo pathway responsible for supplying malonyl-CoA substrate to FASN (Harada et al., 2007), and OLAH releases the growing fatty acid chain specifically for medium chain fatty acids (MCFA) (Smith, 1980). Interestingly, numerous other predicted targets in this same pathway were suppressed twofold or more by constitutive expression of miR-150 (Fig. 5) including Thrsp, required for MCFA synthesis in MECs (Rudolph et al., 2014), Acly, responsible for synthesis of the substrate acetyl-CoA used for de novo fatty acid synthesis (Linn and Srere, 1979), and Slc25a1, which shuttles substrate into the de novo fatty acid synthesis pathway (Kaplan et al., 1993). These targets, like FASN, might be suppressed even more strongly at the protein level. Likely due to suppression of multiple key components of this pathway, MECs from WAP-Cre⁺; Stop-150^fl/fl mice had a profound deficit in saturated fatty acids (Fig. 8A), with notably less de novo fatty acid synthesized MCFA per mole of triacylglyceride (TAG), including 10:0, 12:0 and 14:0.

Fasn knockout mice (Fasn KO) also exhibit a lactation defect characterized by decreased pup weight, increased pup mortality, and milk containing significantly less 14:0, 16:0, 18:0, and overall fatty acids compared with controls (Suburu et al., 2014). However, constitutive expression of miR-150 additionally affected the MCFA10:0 and 12:0 (Fig. 8A) and demonstrated a more dramatic effect on pup mortality. This broader reduction in MCFA can be attributed to suppression of the aforementioned additional key enzymes involved in de novo fatty acid synthesis. Furthermore, reduction in 18:0 (Fig. 8A) is also consistent with the observed decrease in its synthetic precursor, 16:0 (Fig. 8A), plus decreases in mRNA for fatty acid elongases, such as Elovl6 (Fig. 5) that catalyze the synthesis of 18:0 from 16:0 (Moon et al., 2001). Despite decreased 18:0 substrate levels (Fig. 8A) and the observed reductions of Scd1 and Scd2 mRNA (Fig. 5), their product, 18:1 n-9, remained unaffected, likely due to the high quantity of 18:1 n-9 in our rodent diet. Thus, constitutive expression of miR-150 results in broader defects of milk fat composition than the Fasn KO, affecting both MCFA and long chain fatty acids.

It is likely that functional redundancies evolved such that multiple miRNAs, including miR-150, control lactation. For example, at least six other miRNAs that decline precipitously at secretory activation are predicted to target Fasn, including miR-342-3p, miR-361-5p, miR-425-5p, miR-17-3p, miR-15b-5p and miR-532-5p (Table S3). Functional overlap is also evident in miRNAs predicted to target transcripts of milk protein genes. For example, lactotransferrin (Ltf), is predicted to be targeted by six miRNAs significantly downregulated just prior to lactation, two of which are part of the miR-17/92 cluster (Table S6). Such redundancy could explain why knockout of the miR-17/92 cluster did not affect mammary development (Feuermann et al., 2012). Redundant function of miRNAs likely evolved as compensatory mechanisms for a process as crucial to mammalian survival as lactation. Consequently, forced expression to override the natural decrease in a miRNA might be a more effective method to evaluate the contribution of a specific miRNA to MG development than a knockout approach.

Interestingly, the initial rise in miR-150 at mid-pregnancy coincides with a decrease in direct targets of miR-150 such as Myb and Egr2 (Fig. S2B) that encode proteins associated with proliferation (Miao et al., 2011; Liu et al., 2008), and these transcripts peak with proliferation of mammary epithelial cells (Traurig, 1967). Proliferative expansion precedes differentiation and the increase in miR-150-5p at mid-pregnancy might serve to halt proliferation, whereas its decline prior to secretory activation seems to serve an entirely different purpose, as described in this manuscript. Our RPPA data demonstrates that proteins affected by constitutive miR-150 at L2 are not affected by miR-150 at P14 (Fig. 4B). These are exciting examples of how the targets of a given miRNA might differ depending on the transcriptional drivers of mRNAs and what genes are actively being transcribed at particular stages of development – proliferative expansion, differentiation or secretory activation in the mammary gland.

We speculate that progesterone (P4) (in the additional context of high estrogen during pregnancy) could be a likely candidate for regulating miR-150 via the progesterone receptor, as P4 increases at mid-pregnancy when miR-150 increases, then drops precipitously in mice just prior to parturition when miR-150 declines.

Herein, we describe the effects of preventing the natural decline of a miRNA, miR-150, between late pregnancy and lactation. Our data support the hypothesis that a precipitous decline in a program of miRNAs, such as miR-150, provides a level of post-transcriptional control to fine-tune expression of specific proteins essential for the survival and function of differentiated MECs to achieve successful lactation.

CD1 background mice were purchased from Taconic (Germantown, NY). Stop-150^fl/fl in C57BL/6 background were kindly provided by Changchun Xiao, The Scripps Research Institute (Xiao et al., 2007). WAP-Cre⁺ transgenic mice in FVB were originally generated as described (Wagner et al., 1997). At 8 weeks old, control females (WAP-Cre⁺; Stop-150⁻) or bitransgenic females with constitutive expression of miR-150 at late pregnancy and throughout lactation (WAP-Cre⁺; Stop-150^fl/fl) were impregnated by wild-type males. P1 was identified as the first day a post-coital plug was observed. L1 was identified as the first day litters were present. MGs were harvested from dams at P14, P18 and L2. Pup survival data was collected from PND3 and every other day through lactation day 15 for first litters only. Initial litter size was defined as the total number of pups present at PND3. In fostering experiments, only litters born to control dams within 1 day of the biological litter were used as foster litters starting at PND3. All animal procedures were approved by the Institutional Animal Care and Use Committee of the University of Colorado Anschutz Medical Campus.

Adipose-depleted mouse MECs were isolated from the upper inguinal MGs as described (Rudolph et al., 2009) with modifications (see supplementary Material and Methods).

Total RNA was isolated from MECs using Trizol solution (Thermo Fisher Scientific, Waltham, MA, USA) and purified by Qiagen miRNA columns (Qiagen, Venlo, Netherlands). RNA concentration and purity were assessed in Applied Biosytems Bioanalyzer 2100 (Thermo Fisher Scientific).

A 1 µg aliquot of total RNA from each sample was labeled with FlashTag Biotin RNA Labeling kit (Genisphere, Hatfield, PA, USA) and hybridized onto GeneChip Mouse Gene 1.0 ST, GeneChip miRNA 1.0 ST, and GeneChip Mouse Transcriptome Array 1.0 (Affymetrix, Santa Clara, CA, USA) according to the manufacturer's recommendations and performed in the University of Colorado Cancer Center Microarray Core Facility. The raw data for all three arrays are available at http://www.ncbi.nlm.nih.gov/geo under series records GSE87584 and GSE80666.

Data were extracted from the images, quantile normalized, summarized (median polish) and log₂-transformed with miRNA QC tool software (Affymetrix). Data from mouse miRNAs were imported into GeneSpring GX10 (Agilent Technologies, Santa Clara, CA, USA) by creating a custom miRNA experiment, and differentially expressed miRNAs were identified using unpaired t-test. The P-values were corrected by multiple testing using Benjamini–Hochberg False Discovery Rate (BH-FDR). A 5% FDR cut-off was chosen to identify differentially expressed probe sets.

Affymetrix CEL files from all samples were loaded on to Genespring GX10. Signal intensities for all probe sets were obtained using Robust Multichip Averaging summarization algorithm, involving three steps – background correction, quantile normalization and probe summarization (median polish). Quality control was performed by principal component analysis to identify outliers. Differentially expressed probe sets between P14 and L2 were identified by performing an unpaired t-test to obtain raw P-values, which were subsequently corrected by multiple testing using the BH-FDR method. A 5% FDR cut-off was chosen to identify differentially expressed probe sets.

cDNA was synthesized using TaqMan MicroRNA Reverse Transcription kit (Thermo Fisher Scientific) and miRNA-specific RT primers as per manufacturer's recommendations. PCR for each miRNA was performed using specific TaqMan MicroRNA probe and ABsolute Fast QPCR Low ROX mix (2×) (Thermo Fisher Scientific) or TaqMan Universal PCR Master Mix, no AmpErase UNG (2×) (Thermo Fisher Scientific). Quantification is described in supplementary Materials and Methods.

Lower thoracic glands were fixed in 10% neutral buffered formalin. Tissue processing and paraffin embedding were performed by the University of Colorado Denver Histology Shared Resource. Sections of MGs were analyzed by ISH as described in Cochrane et al., (2010). A hybridization temperature of 53°C and miRCURY LNA microRNA Detection Probes pre-labeled with double digoxigenin and complementary to mature miR-150-5p (CACTGGTACAAGGGTTGGGAGA) (Exiqon, Copenhagen, Denmark) were used.

Representative images of ISH, IHC and Hematoxylin and Eosin (H&E) slides were taken using an Olympus BX40 microscope (Center Valley, PA, USA) with a SPOT Insight Mosaic 4.2 camera and software (Diagnostic Instruments, Inc., Sterling Heights, MI, USA).

Canonical pathway analysis was calculated using Ingenuity Pathway Analysis (Qiagen). Significance was calculated by Fisher's exact test right-tailed. The significance indicates the probability of association of altered molecules in the dataset with the pathway by random chance alone. The percent altered genes indicate the number of statistically significantly altered genes in the pathway divided by the total number of genes that make up that pathway.

Targets of mmu-miR-150-5p were bioinformatically predicted using TargetScan Mouse v7.1, TargetScan Human v7.0 (Lewis et al., 2005), and RNA22 v2.0 (Miranda et al., 2006).

Sections of paraffin-embedded MGs were cut at 4 μm and deparaffinized in xylene, rehydrated with a series of graded ethanols, and subjected to heat-induced epitope retrieval in 10 mM citrate buffer, pH 6.0. Endogenous peroxidase was blocked, and slides were treated with 10% normal goat serum. Antibodies and detection methods are described in supplementary Materials and Methods. For CC3 staining, three separate 100× fields of each slide were analyzed with ImageJ (National Institutes of Health, Bethesda, MD, USA). Color threshold was adjusted manually; Red Green Blue (RGB) for positive-staining nuclei, and Hue Saturation Brightness (HSB) for total nuclei. RGB area was divided by HSB area to calculate percent positive CC3 cells.

H&E stains were purchased from Anatech (Battle Creek, MI, USA) and used per the manufacturer's instructions. Three separate 100× fields of each slide were analyzed. Nuclei were counted manually for each field. Alveolar pixel area was quantified using ImageJ by first subtracting non-epithelial tissue then adjusting a color threshold using HSB color space to include total epithelium. Nuclear density was calculated by dividing nuclei counts with the HSB area.

RPPA printing and analysis of MEC samples was conducted as previously described (Wulfkuhle et al., 2012, 2008; Sheehan et al., 2005). Before use for RPPA analysis, antibody specificity was confirmed by immunoblot and analysis, as previously described (Sheehan et al., 2005).

Protein from each MEC sample (20 µg) was resolved by polyacrylamide gel electrophoresis and transferred to Immobilon-FL membranes (Millipore Continental Water Systems, Bedford, MA, USA). The membranes were blocked with 5% milk in PBS, probed with primary antibodies overnight at 4°C, washed and incubated with appropriate Alexa Fluor secondary antibody (Thermo Fisher Scientific) (see supplementary Materials and Methods section for details of antibodies). Protein was detected using an Odyssey infrared imager (LI-COR Biosciences, Lincoln, NE, USA).

The 3′ UTR DNA fragments of mouse Fasn (1896 bp, NM_007988) and Stat5b (2518 bp, NM_001113563) containing the putative miR-150-5p target sites were amplified by PCR from mouse genomic DNA (primers described in supplementary Materials and Methods) and cloned into the SacI- and XhoI-digested pmirGLO vector downstream of the firefly luciferase cDNA sequence (Promega), resulting in the generation of Fasn-3′-UTR-luc and Stat5b-3′-UTR-luc plasmids. These plasmids were used as a template to generate DNA fragments with mutated miR-150-5p targeting sites (Fasn-3′-UTR-mut, Stat5b-3′-UTR-luc-mut) by using mutated oligos described in the supplementary Materials and Methods. DNA sequencing verified the sequence of both plasmids.

Murine mammary tumor 4T-1 cells were plated at 1×10⁵ cells per well in a 24-well plate. The next day, cells were co-transfected with 50 nM of control RNA or miR-150 mimic (Ambion, Austin, TX, USA) together with 0.5 µg of aforementioned pmirGLO plasmids containing wild-type or mutated putative target sites using Lipofectamine 2000 (Thermo Fisher Scientific). Cells were lysed 48 h later and the dual-luciferase reporter assay system (Promega) was used to measure luciferase activity according to manufacturer's instructions.

Because of the low milk yield from WAP-Cre⁺; Stop-150^fl/fl mammary glands compared with controls, even after administration of a bolus of oxytocin, lipids were extracted from isolated MECs. HPLC grade reagents were purchased from Sigma-Aldrich (St. Louis, MO, USA). Total lipid extraction was performed as previously described (Rudolph et al., 2014) with modifications described in supplementary Materials and Methods.

A 50 μl aliquot of isooctane suspended total lipid was taken to dryness under nitrogen gas, samples were resuspended in 200 μl dichloromethane that contained 15 μl of a 10% nonaethylene glycol monododecyl ether (Sigma-Aldrich) dissolved in dichloromethane (w/v). Samples were incubated for 5 min at 25°C and taken to dryness at 40°C for 25 min to ensure organic solvent was completely evaporated. Pellets contained triglyceride and nonionic surfactant complexes, to which 200 μl of reverse osmosis water was carefully added without mixing and incubated at 40°C for 10 min and followed by a gentle vortex. A standard regression curve was made using 80 nmol of tripalmitin (Sigma Aldrich) combined with 25 μl of 10% nonaethylene glycol monododecyl ether in dichloromethane (w/v), incubated and dried as above, suspended in 100 μl of reverse osmosis water, and dilutions of 20, 10, 5, 2.5, 1.25, 0.625 and 0.3125 nmol tripalmitin were used. Total TAG from the organic fraction was quantified relative to known tripalmitin standard using a modified colorimetric assay (Van Veldhoven et al., 1997) and values are expressed as mM concentrations. Triglyceride Reagent and Free Glycerol Reagent Development were purchased from Sigma-Aldrich and diluted according to the manufacturer's instructions.

Blended stable isotope internal standards containing 100 ng each of [D]3-decanoic acid; [D]3-lauric acid; [D]3-myristic acid; 1,2,3,4-[¹³C]4-palmitic acid; [D]3-stearic acid; [D]4-oleic acid; [D]8-arachidonic acid and [D]5-docosahexanoic acid [purchased from Sigma-Aldrich, Cambridge Isotopes (Andover, MA, USA) or Cayman Chemical Co. (Ann Arbor, MI, USA) with 99 atom % ¹³C and 99 atom % D, respectively] were added to the volume of each sample representing 5 nmol of MEC TAG as quantified above. Samples were taken to dryness under N₂ gas, suspended in 0.5 ml 100% methanol, and were saponified at 45°C for 1 h by adding 0.5 ml of 1M sodium hydroxide mixing at 20 min intervals. Samples were acidified with 0.525 ml of 1 M HCl, vortexed vigorously; fatty acids were extracted twice with 1.0 ml of isooctane, and taken to dryness under N₂ gas. Saponified fatty acids were derivatized at room temperature for 30 min by addition of 30 μl of 1% pentafluorobenzyl bromide in acetonitrile and 30 μl of 1% N,N-diisopropylethylamine in acetonitrile according to Rudolph et al. (2014), after which the samples were taken to dryness under N₂ gas. The resulting pentafluorobenzyl fatty acid esters were suspended in 200 μl isooctane, vortexed, and transferred into vials for gas chromatography mass spectrometry.

We'd like to acknowledge Emanuel Petricoin and Julia Wulfkuhle at the Center for Applied Proteomics and Molecular Medicine at George Mason University for RPPA. The authors also acknowledge the Genomics and Microarray Core and other Shared Resources of Colorado's NIH/NCI Cancer Center Support Grant P30CA046934.