The miR-424(322)/503 cluster orchestrates remodeling of the epithelium in the involuting mammary gland

The mammary gland undergoes continuous remodeling. Llobet-Navas et al. identify the microRNA cluster miR-424(322)/503 as an important regulator of epithelial involution after pregnancy. TGF-β regulates the expression of this miR cluster, which in turn targets BCL-2 and IGF1R. This work suggests a model in which activation of the TGF-β pathway after weaning induces the transcription of the miR-424(322)/503 cluster to down-regulate the expression of key genes.

Abstract

The mammary gland is a very dynamic organ that undergoes continuous remodeling. The critical regulators of this process are not fully understood. Here we identify the microRNA cluster miR-424(322)/503 as an important regulator of epithelial involution after pregnancy. Through the generation of a knockout mouse model, we found that regression of the secretory acini of the mammary gland was compromised in the absence of miR-424(322)/503. Mechanistically, we show that miR-424(322)/503 orchestrates cell life and death decisions by targeting BCL-2 and IGF1R (insulin growth factor-1 receptor). Furthermore, we demonstrate that the expression of this microRNA cluster is regulated by TGF-β, a well-characterized regulator of mammary involution. Overall, our data suggest a model in which activation of the TGF-β pathway after weaning induces the transcription of miR-424(322)/503, which in turn down-regulates the expression of key genes. Here, we unveil a previously unknown, multilayered regulation of epithelial tissue remodeling coordinated by the microRNA cluster miR-424(322)/503.

The mammary gland is a very dynamic organ that passes through continuous tissue remodeling during the female lifetime (Howard and Gusterson 2000; Ip and Asch 2000; Macias and Hinck 2012). The appearances of distinct placodes containing mammary epithelial precursors initiate the limited ingrowth of a rudimentary ductal tree during embryogenesis that undergoes extensive expansion at puberty in the female breast. In adult females, side branches and lobulo–alveolar structures are in a continuous cycle in response to the cycling ovarian hormones. During pregnancy, secretory alveoli develop in response to specialized hormones, forming a very dense lactiferous epithelial tree that invades almost the entire mammary fat pad. During weaning, regression of this architecture occurs, leading to alveoli and secretory duct collapse (a process known as involution).

During the last two decades, key details about the molecular mechanism and regulation of the massive remodeling that occurs during mammary gland involution have emerged (Watson 2006; Stein et al. 2007; Watson and Kreuzaler 2011). It is well established that involution occurs in two distinct phases according to its reversibility. The first 48 h after weaning are characterized by apoptotic cell death without changes in acini architecture. Molecularly, this step involves both the intrinsic and the extrinsic apoptotic pathways and is modulated by the expression levels of members of the pro- and anti-apoptotic BCL-2 family. A few signaling pathways have been identified as upstream modulators of apoptosis, some of the most relevant being TGF-β, IGF (insulin growth factor), LIF, NF-κB, and STAT-3. Cues for these pathways converge to regulate Akt/PKB, which in turn balances cell life and death decisions. The second phase, 2 d after initiation of involution, is characterized by the collapse of the alveolar structure and degradation of the supportive extracellular matrix, which is mediated by metalloproteases.

Despite our current knowledge of mammary gland development, our understanding of the contribution of miRNAs to this process is in its infancy. MicroRNAs (miRNAs) are small noncoding RNAs, ∼22 nucleotides (nt) in length, with critical roles in multiple aspects of cell biology and organism physiology (Bartel 2004, 2009). miRNAs post-transcriptionally regulate gene expression mainly by targeting mRNAs for cleavage, translational repression, and mRNA destabilization. A very limited number of studies have identified and validated miRNAs with a role in any aspect of normal mammary gland biology (Avril-Sassen et al. 2009; Tanaka et al. 2009; Ucar et al. 2010; Yang et al. 2010; Cui et al. 2011; Le Guillou et al. 2012), and even fewer have studied their function in vivo (Ucar et al. 2010; Le Guillou et al. 2012).

Here we identified the miR-424(322)/503 cluster as an important regulator of involution. To investigate the processes influenced by this cluster in vivo, we generated a knockout mouse model in which miR-424(322)/503 was deleted. Although knockout mice were fully viable, detailed analyses of mammary gland development revealed a delayed pattern of involution after pregnancy. The study of miR-424(322)/503 targets identified a group of genes with a key role in apoptosis (BCL-2) and insulin signaling (IGF1R [IGF-1 receptor]). Importantly, we also demonstrated that the changes in the expression of these targets mediated by the miR-424(322)/503 cluster impact cell death as well as the activity of key signal transduction pathways (AKT and ERK1/2) in vitro and in vivo. Furthermore, our studies linked the expression of the cluster to TGF-β, which plays a well-known role in mammary gland involution (Faure et al. 2000; Nguyen and Pollard 2000; Stein et al. 2007). Overall, our studies suggest a model in which the expression of a primary transcript containing the miR-424(322)/503 cluster is up-regulated during involution via TGF-β exposure. Once processed, these mature miRNAs are part of the mechanisms that induce involution by down-regulating the expression of key components of signal transduction and cell death.

Results

The miRNA cluster 424(322)/503 is up-regulated during mammary gland involution in mammary epithelial cells

In order to investigate the landscape of miRNAs expressed in the mammary gland epithelium, we performed miRNA high-throughput sequencing (HTP-seq) studies on six different stages of development (puberty, estrus, diestrus, pregnancy [15 d], lactation [10 d], and involution [3 d]) (Fig. 1A). Estrus and diestrus represent the two most different stages that occur during periodic ovarian cycles. Estrus is characterized by the presence of primary and secondary epithelial ducts with very little alveolar budding. On the other hand, the diestrus stage is characterized by a higher abundance of alveolar buds. Involution of the mammary gland after pregnancy is the phase where the most dramatic changes occur. We thus decided to investigate the miRNAs that were specifically expressed in the mammary epithelial cells during this period. Of 1281 described mature miRNA sequences (miRBase release-19), ∼200 were expressed in at least one developmental stage (complete miRNA profile will be reported elsewhere) (data not shown). When we searched for miRNAs preferentially up-regulated during involution, we found a group of 12 miRNAs with a Z-score ≥1.5 (Fig. 1A). Additionally, we studied by quantitative RT–PCR (qRT–PCR) the expression of these miRNAs across a series of 15 different organs. Interestingly, miR-424(322) was expressed at the highest levels almost exclusively in involuting mammary epithelium, while the rest of the miRNAs in the list were expressed at higher levels in several other organs (Fig. 1B). Furthermore, miR-424(322) was also the most up-regulated miRNA transitioning from lactation to involution (72 h) (Fig. 1C). Nine of the 12 miRNAs that we found preferentially up-regulated during involution have been previously studied during mammary development (Avril-Sassen et al. 2009). While we used HTP-seq of purified mammary epithelial cells to evaluate miRNA expression, the Avril-Sassen et al. (2009) study used a bead-based flow cytometric microarray platform and whole mammary gland RNA extraction without purification. This difference complicates the direct comparison of the data; however, we noticed that six of the nine overlapping miRNAs between both studies showed similar expression patterns (Supplemental Fig. S1A).

Figure 1.

Up-regulation of the miRNA cluster 424(322)/503 in the epithelium during the involution of the mammary gland. (A) Schematic representation of the strategy performed to identify stage-specific miRNAs in the epithelium of the mammary gland and lists of miRNAs differentially up-regulated during involution. Samples were collected at puberty (1 mo), estrus, diestrus, pregnancy (15 d), lactation (10 d), and involution (3 d). (B) Tissue-specific expression analysis of involution-related miRNAs and their relative expression in the involuting mammary epithelium; n = 5. (C) Fold up-regulation of involution-related miRNAs when compared with lactating mammary glands; n = 5. (D) Estimation of the number of molecules of miR-322 per cell during mammary gland development; n = 3. (E) Distribution of the normalized number of reads of miR-424(322) and miR-503 per sample after miRNA HTP-seq analysis in a cohort of 800 primary breast tumors (The Cancer Genome Atlas [TCGA] breast cancer). See also Supplemental Figure S1.

miR-424(322) belongs to the miR-16 family (Liu et al. 2008; Finnerty et al. 2010). Some members of this miRNA family have been previously shown to target known modulators of cell cycle and apoptosis, which are fundamental pathways during involution. We observed that other members of the miR-16 family were expressed in the mammary epithelium; however, their steady-state levels did not increase significantly during involution (72 h) (Supplemental Fig. S1B). Remarkably, evolutionary studies on the miR-16 family have revealed that while all vertebrates studied to date express miR-15a, miR-15b, miR-16, miR-103, and miR-107, only mammals are known to express miR-195, miR-424(322), miR-497, miR-503, and miR-646 (Finnerty et al. 2010), suggesting a specialized function.

Prompted by these data, we decided to investigate in more detail the role and regulation of miR-424(322) in the remodeling of involuting mammary epithelium. A more detailed analysis of the expression of the mature miR-424(322) during early (24 h after weaning) and late (3 d after weaning) involution confirmed that it peaks during late involution (Supplemental Fig. S1C). miR-424(322) is located in chromosome X and in close proximity (∼380 base pairs [bp]) to miR-503 in both humans and mice. Both miRNAs have been proposed to be expressed as a cluster (Forrest et al. 2010), and, in fact, we were able to retrotranscribe a transcript containing both miRNAs (Supplemental Fig. S1D). Interestingly, miR-503 was filtered out in our HTP-seq analysis because of its low expression levels, suggesting that miR-424(322) is preferentially processed from the precursor miRNA (pre-miRNA). By titrating chemically synthesized mature miRNAs, we estimated that the number of molecules of mature miR-322 is about a few hundred per cell but increases to a few thousand (fivefold to 10-fold) during involution (Fig. 1D; Supplemental Fig. S1E). Although the mature form of miR-503 also increased proportionally during involution, the steady-state levels of miR-503 were about a log of magnitude lower than levels of miR-322 (Supplemental Fig. S1F). By comparing the number of reads corresponding to both miRNAs in the HTP-seq data generated by The Cancer Genome Atlas (TCGA) breast cancer study from >800 samples (Fig. 1E; https://tcga-data.nci.nih.gov/tcga), we also confirmed that higher levels of miR-424(322) are maintained in humans. Finally, HTP-seq data available in the miRBase database (http://www.mirbase.org) as well as published studies in normal ovarian tissue (Ahn et al. 2010) confirmed a comparable relative miR-424(322)/miR-503 ratio of expression. Taken together, these data suggest that higher levels of mature miR-424(322) compared with miR-503 are well preserved between humans and mice. The reason for the unbiased levels between these two miRNAs is unclear. In this regard, it has been previously reported that miR-503 is usually unstable and that this instability is mediated by the seed region and the 3′ end (Rissland et al. 2011).

miR-424(322)/503 knockout mice present defects in mammary gland involution and display alveoli hyperplasia after pregnancy

To study the role of the miR-424(322)/503 in the mammary epithelium, we generated a knockout mouse model using zinc finger nuclease (ZFN; Sigma) technology (Fig. 2A; Urnov et al. 2010). Two ZFNs cutting upstream of and downstream from the cluster were engineered and injected into 400 fertilized oocytes. These oocytes were injected into recipient C57BL/6 females, and 54 viable pups were born. Males and females were born at a Mendelian ratio. The genotype of these animals revealed that 50% of the males carried a mutated miR-322/503 allele. In females, 30% were heterozygous, and 8.3% had both alleles mutated. Due to the nature of the mechanisms involved in the repair of the double-strand breaks generated by the ZNFs, there were varying sizes of the deletion in mutant alleles. These deletions ranged from 22 nt to larger fragments of almost 1 kb containing the two miRNAs (Supplemental Fig. S2A,B). For our studies, we used mice containing deletions of both mature miRNAs (alleles 7A and 38) (Fig. 2A). As expected, both knockout alleles showed the same phenotype (see below). Consequently, in this study and for simplification, we do not make any distinction between them and use the general term miR-424(322)/503 knockout. We also obtained alleles where only one of the two miRNAs was deleted. Interestingly, when we analyzed the expression levels of the remaining miRNA, we observed that these levels were severely reduced to barely detectable levels (Fig. 2B). This result suggests that the sequence/structure of the miR-424(322)/503 primary transcript is essential for proper generation of mature miRNAs.

Figure 2.

miR-424(322)^−/−/503^−/− knockout (KO) mice show delayed post-lactation mammary gland involution and alveolar hyperplasia. (A) Modeling of miR-424(322)^−/−/503^−/− double-knockout mice by using the ZFN technology generates multiple deleted alleles. (B) qRT–PCR expression values for miR-424(322) and miR-503 in wild-type animals and miR-424(322)^−/−/503^−/− knockout mice with different deleted alleles. (C) Representative carmine red staining performed on old nulliparous female mammary glands reveals discrete enlarged terminal ductal lobular units (TDLUs). (D) In parous females, TDLUs progress into enlarged multifocal structures compatible with alveolar hyperplasia. (E) Representative stainings of involuting mammary glands at 72 h. H&E, cytokeratin 5, cytokeratin 18, and caspase-3 stainings reveal drastic lobular epithelium death in wild-type (WT) mice, while the mammary epithelium of miR-424(322)/503 knockout animals still presents well-organized and structured acini. (F) The graph shows quantification of caspase-3-positive cells per acini for immunostainings (between 80 and 150 acini were counted per condition). See also Supplemental Figure S2.

Phenotypically, knockout pups were identical to the wild type in appearance, size, and weight. We did not observe any evident differences between knockout and wild-type animals during the first 2 mo of life. At a later age, knockout mice presented with a higher body weight due to increased white fat content (characterization of this phenotype will be reported elsewhere).

Next, we concentrated our studies on the mammary gland epithelium during the different phases of development. A systematic study of the female epithelial mammary tree was done by comparing whole-mount preparations stained with carmine red and quantifying the area occupied by epithelial cells and adipocytes (Chapman et al. 1999; Tiffen et al. 2008). Acini morphology and organization were studied using hematoxyilin/eosin (H&E) staining, and basal (cytokeratin 5) and luminal (cytokeratin 18) markers were studied using immunohistological techniques. These studies revealed no major differences in the organization (size, length, number, and basal/luminal marker distribution) of the ducts and lobules during puberty and during estrus/diestrus of nulliparous animals at an early age (up to 3 mo) (Supplemental Fig. S2C–E). Older virgin knockout females presented a few slightly enlarged terminal ductal lobular units (TDLUs) containing two to three times more acini than age-matched wild-type counterparts (Fig. 2C). Interestingly, parous females displayed mammary glands with multifocal, larger TDLUs compatible with alveoli hyperplasia (Fig. 2D).

Expression of the miR-424(322)/503 cluster increases during involution after pregnancy. Thus, it is plausible that the observed phenotype in parous animals has its origin at this stage. Our studies during pregnancy and involution revealed that during pregnancy, secretory acini in knockout animals displayed normal morphological and functional (milk-producing) characteristics (Supplemental Fig. S2F,G). In addition, we also observed comparable morphological features at early involution (24 h after weaning) between knockout and wild-type animals (Supplemental Fig. S2H,I). At late involution (3 d after weaning), wild-type females displayed massive destruction of the lobular epithelium, which led to the absence of recognizable acini structures; in contrast, knockout females presented lobules with morphologically well-structured acini (Fig. 2E; Supplemental Fig. S2H,I). Remarkably, apoptosis was reduced 3 d after weaning in knockout animals, as indicated by fewer cells displaying activated caspase-3 immunostaining (Fig. 2E,F; Supplemental Fig. S2J).

It is important to mention that both knockout alleles (7A and 38) were crossed with wild-type animals for six generations (a 98.5% clean background) to confirm the association between the genotype miR-424(322)/503 knockout and the phenotype. As expected, the phenotype segregated with the knockout alleles.

The miRNA cluster 424(322)/503 targets the key regulators of mammary gland involution, BCL-2 and IGF1R

The biological functions of miRNAs are mediated by their ability to attenuate the expression of targeted mRNAs. Thus, we investigated the targets of the miR-424(322)/503 cluster. Computational prediction of targets based on seed sequence conservation (Lewis et al. 2005) as well as nonconserved sites (Grimson et al. 2007) is a widely used strategy to identify miRNA targets. We used the TargetScan algorithm (http://www.targetscan.org), which has been shown to have one of the highest specificities in preselecting putative target genes (Shirdel et al. 2011). miR-424(322) and miR-503 have almost identical seed sequences and belong to the same miRNA family (Fig. 3A). Thus, it is not surprising that the predicted target overlap between both miRNAs is very high [>90% of the predicted targets conserved in mammals for miR-503 are included in the list for miR-424(322)] (Supplemental Fig. S3A). Computational predictions of miRNA targets result in hundreds, if not thousands, of candidates. In order to reduce this number of genes, we exploited the fact that up-regulation of the targeting miRNA reduced the expression level of targets with potent binding side sequences (Bartel 2009; Guo et al. 2010). We transduced MCF-10A, an immortalized but nontransformed human mammary epithelial cell (HMEC) line, with lentiviruses expressing the miR cluster or a control vector and compared their expression profiles after 48 h. As expected, this study showed a strong statistically significant reduction in the expression of predicted target genes (P < 10⁻³⁴) (Fig. 3B). Pathway analysis (Ingenuity) of the predicted targets that showed reduction in mRNA levels identified a group of genes involved in cell cycle, apoptosis, TGF-β, and canonical ERK/AKT signal transduction pathways (Supplemental Fig. S3B,C). Since these pathways are key regulators of mammary gland involution (Neuenschwander et al. 1996; Nguyen and Pollard 2000; Schwertfeger et al. 2001; Stein et al. 2007; Macias and Hinck 2012), we decided to further study a series of these putative targets by following a systematic approach.

Figure 3.

BCL2 and IGF1R are bona fide targets of miR-424(322)/503. (A) The illustration represents the multistep strategy designed to identify miR-424(322)/503 targets. (B) Cumulative distribution of predicted miR-424(322)/503 targets and nontargets when their expression levels from control and miR-424(322)/503-up-regulated MCF-10A cells are compared. (C) Cloning of the 3′ UTR and subsequent 3′ UTR-luciferase assays of the selected candidates shows differential targeting activity by miR-424(322)/503 (n = 5) that was further filtered by PAR-CLIP expression analysis via qRT–PCR (n = 5; shown in D). (E) miRNA-binding site mutagenesis and luciferase assays were performed in the most relevant candidates to prove the miR-424(322)/503-binding specificity. (F) MCF-10A cells were lentivirally transduced with an inducible plasmid expressing the miR-424(322)/503 and treated with Dox for 5 d. Western blot analysis showed a decrease in the endogenous protein levels of miR-424(322)/503 targets IGF1R2 and BCL2 but not in CCNE1 (n = 5). (G) Western blot studies of purified mouse mammary epithelium show that IGF1R and BCL2 protein levels decrease during the transition from lactation to involution in wild-type (WT) animals and that post-lactation miR-424(322)/503 knockout animals as well as Tert-immortalized miR-424(322)/503 knockout cells (Tert-KO) present higher expression of these targets than their wild-type counterparts (shown in H). A minimum number of five animals were analyzed. See also Supplemental Figure S3.

First, we assessed the ability of the predicted target sites of these genes to regulate the expression of a reporter gene. We cloned a portion of the untranslated regions (UTRs) containing the predicted miR cluster-binding sites for 20 genes related to the above-mentioned pathways downstream from a luciferase reporter gene (Luc-UTR). We then cotransduced 293T cells with the corresponding Luc-UTR reporter, a renilla luciferase expression vector lacking any UTR used for normalization purposes, and a plasmid expressing the miR-424(322)/503 cluster. The UTRs of 14 out of 20 candidates showed a statistically significant reduction of the Luc-UTR reporter in the presence of the construct expressing the miR-424(322)/503 cluster (Fig. 3C).

Next, we performed additional validation using photoactivatable ribonucleoside-enhanced cross-linking immunoprecipitation (PAR-CLIP) of the RNA-induced silencing complex (RISC) (Hafner et al. 2010). PAR-CLIP experiments were performed in MCF-10A control cells and an MCF-10A variant transduced with lentivirus vector expressing the miR-424(322)/503. After CLIP, qRT–PCR was used to quantify the levels of captured mRNA for the 14 selected targets. Our data revealed a statistically significant enrichment for six of 14 targets, four of them showing more than twofold enrichment (CDC25A, BCL-2, IGF1R, and CCNE1) (Fig. 3D). Interestingly, there was a positive correlation between the enrichment in the PAR-CLIP assay and the Luc-UTR reporter assay. Because the expression of BCL-2 (Metcalfe et al. 1999) and IGF1R (Modha et al. 2004) has been found to be down-regulated during involution and because apoptosis and insulin signaling are well-known regulators of this process, we decided to further investigate these two genes. Additionally, we also studied CCNE1, as it has previously been shown to be targeted by this miR cluster. Thus, we next mutated the predicted conserved miR-424(322)/503-binding sites in the Luc-UTR constructs for these three genes. Modification of the sequence complementary to the miRNA seed region in the mRNA UTR reverted completely the reduction of luciferase signal after up-regulation of the miR-424(322)/503 cluster (Fig. 3E).

Next, we analyzed the impact of the miR-424(322)/503 on the levels of the endogenous proteins. In order to use the most biologically relevant model, we engineered a MCF-10A variant to up-regulate the expression of the miRNA-424(322)/503 using a doxycycline (Dox)-inducible system [MCF-10A-424(322)/503miR-Dox]. Importantly, this system allowed us to up-regulate miR-424(322) and miR-503 to levels comparable with those observed in involuting mouse mammary epithelial cells (MMECs) (Supplemental Fig. S3D). Up-regulation of the miR cluster in MCF-10A-424(322)/503miR-Dox clearly attenuated the expression levels of BCL-2 and IGF1R, but the effect on CCNE1 expression was modest (Fig. 3F). Finally, we studied the expression levels of these three targets in primary MMECs during development as well as in telomerase (Tert)-immortalized mammary epithelial cell lines from wild-type and knockout animals. Western blot studies revealed an evident down-regulation of Igf1r and Bcl-2 during involution that occurs at the highest peak of the mature miR-424(322) and miR-503 expression, while no major effect was seen in Ccne1 (Fig. 3G; Supplemental Fig. S3E). Remarkably, both primary involuting MMECs and Tert-immortalized miR-424(322)/503 knockout mammary epithelial cells presented consistently higher protein levels of these proteins than wild-type counterparts. Again no significant changes were observed in Ccne1 (Fig. 3H).

Expression of the miR-424(322)/503 cluster modulates IGF-1 signal transduction and apoptosis

While the data described above clearly showed the ability of miR-424(322)/503 to modulate the expression of BCL-2 and IGF1R, it was crucial to determine whether this regulation was biologically relevant.

IGF1R is the bona fide receptor for IGF-1 (Pollak 2012). IGF-1 binding activates the receptor kinase, leading to autophosphorylation and phosphorylation of downstream substrates and resulting in the activation of two main signaling pathways, the PI3K–AKT/PKB and the Ras–MAPK pathways. To test the role of miR-424(322)/503 in these pathways, IGF-1 was added to starved MCF-10A-424(322)/503miR-Dox cells after the miR cluster was induced by the addition of Dox to the medium. Down-regulation of IGF1R protein levels mediated by induction of the miR cluster led to a weaker response to IGF-1 stimulation, as demonstrated by reduction in the levels of phospho-IGF1R (Tyr1135 and Tyr1131), phospho-AKT (Ser473), and phospho-ERK1/2 (Thr202/Tyr204) (Fig. 4A). Importantly, wild-type phosphorylation levels in all three proteins were restored when a nontargetable version of IGF1R lacking the 3′ UTR was exogenously expressed (Fig. 4B; Supplemental Fig. S4A). To complement our studies, we performed a similar analysis with Tert-immortalized MMECs and primary MMECs. These revealed that miR-424(322)/503 knockout cells, which present higher levels of Igf1r, were more sensitive to IGF-1 stimulation than their wild-type counterparts (Fig. 4C,D). Altogether, these data demonstrate that modulation of IGF1R by miR-424(322)/503 affects the response to IGF-1 and its downstream canonical signaling.

Figure 4.

Induced miR-424(322)/503 desensitizes cells to IGF-1 and primes cells to undergo apoptosis. (A) Dox-induced miR-424(322)/503 targets endogenous IGF1R, causing desensitization to IGF-1 stimuli, as shown by Western blot. Cells were treated with Dox for 5 d, starved with 0.5% horse serum overnight, and subsequently treated with 50 ng/mL IGF-1 for the specified time points. (B) Engineered pT-424(322)/503 MCF-10A cells stably expressing a nontargetable form of IGF1R (pLX-IGF1R) are able to overcome IGF-1 desensitization caused by Dox-induced miR-424(322)/503. pLX-EV MCF-10A cells stably expressing an empty vector control. (C) Tert-immortalized knockout (Tert-KO) epithelial cells show consistent higher sensitivity to IGF-1 stimulation, as shown by Western blot. (D) Western blots showing increased activation of p-AKT and p-ERK in purified miR-424(322)/503 knockout involuting mammary epithelial cells compared with their wild-type (WT) counterparts. (E) pT-424(322)/503 cells treated with Dox display higher sensitivity to apoptosis induced by 100 ng/mL paclitaxel (PTX) and 10 μg/mL cisplatin, as shown by quantification of cell death by Annexin-V staining and FACs analysis (top panel) and analysis of active caspase-3 and cleaved PARP by Western blot (bottom panel). MCF-10A cells were treated with Dox and subsequently exposed to PTX for 24 h. (F) Expression of a BCL2 cDNA devoid of its 3′ UTR (pLOC-BCL2) blocks apoptotic cell death induced by PTX. (G) Tert-immortalized knockout cells exert higher resistance to PTX, as shown by Annexin-V, caspase-3, and PARP studies. All the experiments were performed in triplicate. See also Supplemental Figures S4 and S5.

BCL-2 is the founding member of the Bcl-2 family of apoptosis regulator proteins (Colitti 2012), and its key role in the intrinsic apoptosis pathway is well established. BCL-2’s prosurvival abilities are largely due to its ability to control mitochondrial membrane permeability by inhibiting proapoptotic proteins. Thus, we tested whether the ability to undergo apoptosis was affected by the miR-424(322)/503 cluster. Apoptosis in MCF-10A-424(322)/503miR-Dox cells was induced by exposure to different chemotherapeutic agents (cisplatin and paclitaxel). Up-regulation of the miR-424(322)/503 cluster by addition of Dox significantly increased apoptosis induced by paclitaxel and cisplatin as shown by Annexin-V FACS analysis, caspase-3, and PARP cleavage, while no differences were observed in control cells (Fig. 4E; Supplemental Fig. S4B–D). Importantly, transduction of a BCL-2 mRNA that lacks the 3′ UTR eliminated this effect (Fig. 4F; Supplemental Fig. S4E–H). As previously done for ERK/AKT signaling studies, we also studied the response to apoptotic stress in our Tert-immortalized MMECs. We found that miR-424(322)/503 knockout cells, which present higher levels of Bcl-2, were significantly more resistant to apoptosis than wild-type counterparts (Fig. 4G; Supplemental Fig. S4I,J). Overall, these data demonstrate that modulation of BCL-2 by the miR-424(322)/503 cluster affects the stress-induced apoptotic response.

Characterization of the 424(322)/503 promoter pri-miR and its modulation by TGF-β

To mechanistically understand the regulation of the miR-424(322)/503 in the mammary epithelium, we studied the genomic organization of the locus. As a first approach, we took advantage of the publicly available information contained in the University of California at Santa Cruz genome (UCSC) browser (http://genome.ucsc.edu). The close proximity between miR-424(322) and miR-503 suggests that the primary transcript containing both miRNAs exists as a cluster. Interestingly, a noncoding RNA (MGC16121) is described in the reference sequence gene bank whose 5′ precisely matches the predicted sequence of the precursor of miR-424(322) [pre-miR-424(322)] after pre-miR has been excised by the nucleoprocessor complex from the primary transcript (Fig. 5A).

Figure 5.

Genomic organization of miR-424(322)/503 locus unveils transcriptional regulation mediated by TGF-β. (A) UCSC-extracted genomic locations of miR-424(322) and miR-503 and putative regulatory elements located upstream in HMECs, including histone marks of transcriptionally active chromatin (H3K4me2, H3K4me3, H3K9ac, and H3K27ac), transcriptional silencing or repression (H3K27me3), and elongation (H3K36me3) as well as marks of elongation and splicing events, represented by genome-wide FOX2 CLIP-seq. The presence of additional regulatory regions such as CpG islands is also depicted. (B) Representation of the conservation status of miR-424(322), miR-503, and the upstream regulatory regions using a vertebrate and placental mammal Markov model. Using this method, 28 vertebrate species were aligned to assess conservation status. Magnification of a small conserved region located upstream (highlighted in red) identifies two conserved SMAD-binding motifs. (C) Analysis by miRNA expression arrays of MCF-10A treated with 10ng/mL TGF-β for 24, 48, and 72 h. (D) Results were further validated by qRT–PCR (top panel) and in situ hybridization (bottom panel). For in situ hybridization, MCF-10A cells transduced with a construct expressing the cluster miR-424(322)/503 were used as a positive control. (E) TGF-β-dependent regulation of miR-424(322) and miR-503 was verified in other HMEC lines (MCF-12A and 184B5), HuMECs, and primary Tert-immortalized MMECs. Cells were treated with 10 ng/mL TGF-β1 for 48 h. (F) Expression analysis by qRT–PCR of miR-15/16 family members in MCF-10A cells upon treatment with 10 ng/mL TGF-β for several days. (G) TGF-β-dependent miR-424(322) and miR-503 expression was demonstrated by transfecting MCF10A cells with siRNA against SMAD4. Scrambled siRNAs and siRNAs against GFP were used as negative controls. The expression of bona fide TGF-β targets (top panel) and miR-424(322) and miR-503 (bottom panel) was performed by qRT–PCR. See also Supplemental Figure S5.

A review of the track displaying maps of chromatin states generated by the Broad Institute/Massachusetts General Hospital ENCODE group using chromatin immunoprecipitation (ChIP) and HTP-seq (ChIP-seq) revealed a region located ∼2 kb upstream of the miR-424/503 cluster and strongly enriched in histone marks related to an active promoter (H3K4me2, H3K4me3, H3K9ac, and H3K27ac) in HMECs (Fig. 5A). Specifically, there was enrichment in a histone mark associated with transcription elongation (H3K36me3) just after the termination of the active promoter marks and spanning ∼5–6 kb in a region containing the miRNA cluster (Fig. 5A). Remarkably, a comparable histone mark pattern was also found in the majority of the cell lines analyzed, with the exception of B cells (Supplemental Fig. S5A). In embryonic stem cells, this putative promoter seems to be poised for active transcription, as it presents both active and negative histone marks (Supplemental Fig. S5B; Bernstein et al. 2006). Finally, FOX2 CLIP-seq (CLIP coupled with HTP-seq) used to study RNAs associated with the splicing regulator FOX2 showed that the H3K36me3-enriched region was, in fact, transcribed in the 5′–3′ orientation, consistent with a primary transcript starting at the putative promoter upstream of the miR cluster (Fig. 5A).

In order to obtain further insights regarding the transcriptional regulation of the locus, we searched for conserved sequence motifs in the putative promoter region (the ∼2-kb region enriched in the above-mentioned histone marks). Multiple alignments of 28 vertebrate species and measurements of evolutionary conservation using a hidden Markov model-based method (phastCons) (Pollard et al. 2010) identified several motifs conserved in vertebrates and mammals (Fig. 5B). Interestingly, one of these motifs represented two SMAD-binding boxes (Fig. 5B; Johnson et al. 1999). SMADs are bona fide components of the TGF-β and BMP pathways (Massague et al. 2005). Significantly, TGF-β ligands have well-known roles in epithelial mammary gland morphogenesis (Nguyen and Pollard 2000; Stein et al. 2007; Macias and Hinck 2012). Upon TGF-β ligand binding to its receptor, receptor-regulated SMAD-2 and SMAD-3 (R-SMADs) are activated by phosphorylation, heterodimerize with SMAD-4, and are then shuttled to the nucleus, where they act as transcription factors. R-SMADs are weak transcriptional activators and commonly cooperate with additional transcriptional modulators, such as E2Fs (Massague et al. 2005). Remarkably, a highly conserved region containing two E2F-binding motifs is located in close proximity (∼100 bases) downstream from the SMAD-3-binding site.

TGF-β ligands have a key role during mammary epithelial development, mainly as potent inhibitors of proliferation (Howard and Gusterson 2000; Macias and Hinck 2012). During involution, expression of TGF-β ligands in the mammary gland, especially TGF-β3, is induced as soon as 3 h after forced weaning, and levels of TGF-β3 continue to increase during the following hours to remain high for several days (Faure et al. 2000; Nguyen and Pollard 2000). Furthermore, expression of TGF-β type II receptor (Faure et al. 2000) and the nuclear internalization of Smad-4 (Nguyen and Pollard 2000) also occur in the mammary epithelium at about the same time. Although the role of TGF-β3 is not fully understood, in vivo evidence from transgenic mice expressing Tgf-β3 under the β-lactoglobin promoter or expressing a dominant-negative form of the Tgf-β receptor-II in the mammary gland (MMTV-DNIIR) as well as from animals where Tgf-β3 knockout mammary epithelial cells were orthotopically transplanted demonstrates the importance of Tgf-β signaling during the early and late involution phases (Nguyen and Pollard 2000; Gorska et al. 2003). Based on the above information, we hypothesized that exposure to TGF-β ligands during involution up-regulates the expression of miR-424(322)/503, which in turn modulates key players of cell cycle, apoptosis, and IGF-1 signaling.

The mammary cell line MCF-10A is a widely used model to study TGF-β response in mammary epithelial cells (Iavarone and Massague 1997; Karakas et al. 2006). Despite being null for the INK4B gene, these cells retain the ability to arrest after being exposed to TGF-β. (Iavarone and Massague 1997). To study the regulation of the miR-424(322)/503 cluster, we exposed MCF-10A cells to TGF-β1 and, using miRNA microarrays, tracked the expression of 799 human miRNAs (Sanger miRbase release 10.1) through time. Both miR-424(322) and miR-503 emerged as the top TGF-β1-up-regulated miRNAs (Fig. 5C). qRT–PCR of the mature miRNAs and in situ hybridization experiments confirmed the microarray data (Fig. 5D) and showed that the observed up-regulation was time- and dose-dependent (Supplemental Fig. S5C). Induction of miR-424(322)/503 was not restricted to MCF-10A cells, and, although it was less pronounced, we also observed induction in two additional nontransformed human cell lines (MCF-12A and 184B5), in primary HMECs (HuMECs), and in MMECs immortalized with TERT (Fig. 5E).

As mentioned above, TGF-β3 is a TGF-β ligand with a key role during involution (Nguyen and Pollard 2000). Thus, we also tested its ability to induce the expression of the miRNA cluster. Our data showed that comparable levels of miR-424(322) and miR-503 up-regulation were induced with both TGF-β1 and TGF-β3 (Fig. 5F). Interestingly, none of the other miR-16 family members were induced after TGF-β1 and TGF-β3 exposure.

To evaluate the role of the canonical TGF-β pathway in the regulation of miR-424(322)/503, we knocked down SMAD4 before exposing MCF-10A cells to TGF-β1 (Fig. 5G). In knockdown cells, the induction of bona fide TGF-β targets such as SMAD7, P21, CTCF, ATF3, and PAI-1 as well as miR-424(322) and miR-503 was significantly reduced. Next we investigated which of the R-SMADs was involved in the regulation of miR-424(322)/503. Both SMAD2 and SMAD3 are activated during involution and when MCF-10A cells are exposed to TGF-β (Supplemental Fig. S5D). Thus, we individually silenced SMAD2 and SMAD3 (Supplemental Fig. S5E) and evaluated the up-regulation of the mature miR-423(322) and miR-503 as well as the pri-miR cluster. Our studies clearly demonstrate that SMAD3 is the R-SMAD responsible for the up-regulation of the pri-miR-424(322)/503 and the production of the miR mature forms upon TGF-β exposure (Supplemental Fig. S5F,G).

To further confirm the role of TGF-β in regulating the expression of miR-424(322)/503, we performed ChIP assays to investigate the presence of SMAD4 in the endogenous putative promoter. To pursue this aim, we created a MCF-10A cell line where HA-tagged SMAD4 protein was constitutively expressed in MCF-10A cells. After ensuring that the exogenously introduced SMAD4 behaved as the endogenous protein did (both were shuttled into the nucleus in the presence of TGF-β) (Supplemental Fig. S6A), we used this cell line to perform ChIP with anti-HA antibodies. These studies revealed that SMAD4 was enriched in the putative miR-424(322)/503 promoter after TGF-β exposure at levels comparable with the bona fide TGF-β target PAI-1 (Fig. 6A). Importantly, the strongest enrichment happened in the region near the SMAD4-binding site.

Figure 6.

Characterization of the miR-424(322)/503 promoter, pri-miR-424(322)/503, and transcriptional regulation by TGF-β in vivo. (A) MCF-10A cells stably expressing HA-tagged SMAD4 or a control empty vector were treated with 10 ng/mL TGF-β for 3 h, and ChIP was performed using an anti-HA antibody. The quantification of genomic DNA enrichment was performed using specific primers surrounding the promoter. The putative SMAD-binding site is indicated by a red square. Quantification of DNA enrichment in the PAI-1 gene promoter was used as a positive control, while β-Actin and Gapdh were used as negative controls. (B) A schematic illustration obtained from the UCSC genome database showing the localization of miR-424(322) and miR-503 and the upstream putative 5.4-kb promoter region containing either the wild-type or mutated SMAD-binding site that was cloned into a reporter vector, as explained in the text. (C) MCF-10A cells were transduced with either the wild-type or mutated promoters and treated with or without 10 ng/mL TGF-β, and the promoter activity was analyzed by qRT–PCR. (D) Representation of the proposed miR-424(322)/503 cluster promoter region containing the SMAD-binding site and the alternative TSSs identified by 5′ RACE in the presence and absence of TGF-β. (E) Expression kinetics of pri-miR-424(322)/503 after treatment of MCF-10A cells with 10 ng/mL TGF-β for the specified time points by qRT–PCR using specific primers surrounding the pre-miR-424(322) and pre-miR-503 (left) or by in situ hybridization using a probe against pri-miR-424(322)/503 (right). (F) qRT–PCR analysis showing that increased mature miR-424(322) and miR-503 levels after treatment of MCF-10A cells with 10 ng/mL TGF-β coincide with a decrease in pre-miR-424(322) and pre-miR-503, respectively. (G) After inhibition of TGF-β signaling in vivo by the administration of the specific inhibitor LY2109761, mammary glands were examined 72 h after weaning for the presence of secretory acini by H&E and for epithelial density by whole-mount carmine red staining. Additionally, mammary epithelial cells were purified, and milk protein production was evaluated by qRT–PCR using specific primers for α-lactalbumin and β-casein genes. (H) Inhibition of the activation of the TGF-β pathway, shown by Western blot and immunohistochemistry for activated phospho-SMAD2/3, abrogates the transcription of pri-miR-424(322)/503 during involution, as shown by qRT–PCR. See also Supplemental Figure S6, E and F, for a similar study using the TGF-β inhibitor LY2157299.

To evaluate the role of the putative SMAD4-binding site in the up-regulation of the miRNA cluster upon TGF-β exposure, we first engineered a reporter construct containing an ∼5.4-kb region upstream of the miRNA cluster followed by the GFP sequence (Fig. 6B). This reporter was integrated into the genome of MCF-10A cells by lentiviral delivery to create a pri-miR reporter cell line (MCF-10A-pri-miR-reporter). Reverse transcription using oligonucleotides complementary to GFP allowed us to study the transcript resulting exclusively from the reporter without interference of endogenous pri-miR (see the Materials and Methods). Upon addition of TGF-β to the MCF-10A-pri-miR-reporter cells, a clear increase in the transcription of the reporter was observed (Fig. 6C). Importantly, when we mutagenized the SMAD4-binding site, this increase was abrogated.

Next, we further characterized the pri-miR transcript controlled by this promoter. First, we identified the transcription start site (TSS) by performing 5′ rapid amplification of cDNA end (RACE). By using a specific primer located upstream of the miR cluster just before the region enriched in promoter-specific histone marks (∼300 bp downstream from the SMAD4-binding site) for the reverse transcription step, we identified a TSS located downstream from the E2F-binding sites (Fig. 6D; Supplemental Fig. S6B). Interestingly, the addition of TGF-β shifted the TSS 181 nt upstream, which suggests the presence of an alternative promoter. As expected, we were able to amplify a region expanding from the miRNA cluster to the TSS by using overlapping PCRs (Supplemental Fig. S6C). qRT–PCR amplification and in situ hybridization confirmed the up-regulation of the pri-miRNA by TGF-β exposure (Fig. 6E; Supplemental Fig. S6C). As expected, the pri-miRNA was detected by in situ hybridization at the mammary epithelium of involuting mammary glands (3 d) (Supplemental Fig. S6D). Notably, kinetic studies of the pri-miRNA, pre-miRNA, and mature miRNA levels after addition of TGF-β revealed that the increase in pri-miRNA and pre-miRNA happens first and that generation of mature miRNAs leads to a reduction of the pri-miRNA and pre-miRNA (Fig. 6E,F). This could be due to the cleavage and degradation of the pri-miRNA after the pre-miRNA is excised, the existence of a negative regulatory feedback loop, or a combination of both mechanisms.

Finally, to test that the regulation of miR-424(322)/503 by TGF-β occurs in vivo during involution, we individually treated lactating females with two different specific inhibitors of TGF-β receptor I and II with demonstrated activity in vivo (LY2109761 and LY2157299) (Melisi et al. 2008; Bouquet et al. 2011; Calon et al. 2012; Bhola et al. 2013) and analyzed the induction of the miR cluster. When lactating females were pretreated with these inhibitors starting 24 h before weaning, these animals showed an evident inhibition of involution 3 d after weaning, as indicated by carmine red and H&E staining of the mammary gland and qRT–PCR levels of milk proteins (Fig. 6G; Supplemental Fig. S6E). Importantly, due to the inhibition of TGF-β signaling, they also presented a strong reduction in the levels of pri-miR-424(322)/503 (Fig. 6H; Supplemental Fig. S6F).

Discussion

Despite the established role of a variety of extrinsic and intrinsic signals regulating mammary gland development (Howard and Gusterson 2000; Macias and Hinck 2012), a restricted number of studies have identified and validated miRNAs with a role in any aspect of normal mammary gland biology (Avril-Sassen et al. 2009; Tanaka et al. 2009; Ucar et al. 2010; Yang et al. 2010; Cui et al. 2011; Le Guillou et al. 2012), and even fewer have studied their function using transgenic animal models (Ucar et al. 2010; Le Guillou et al. 2012). Here we unveiled the miRNA cluster miR-424(322)/503 as an important regulator of the massive reorganization that occurs in the mammary epithelium after pregnancy. Furthermore, through the generation of a knockout mouse model, we characterized mechanistically its role and regulation.

Among the 12 identified miRNAs, miR-424(322) showed the largest fold increase between lactation and involution. Although this miRNA has been studied in the context of monocyte differentiation (Rosa et al. 2007; Forrest et al. 2010) and endothelial cell regulation (Ghosh et al. 2010; Nakashima et al. 2010; Kim et al. 2013), a potential role in mammary epithelial architecture remodeling has not yet been described. The characterization of the mammary epithelium of miR-424(322)/503 knockout females during development revealed a defect during involution. To understand the role of miR-424(322)/503 in this biological context, we considered two pertinent questions: (1) Which are the relevant targets modulated by the cluster in the mammary epithelium? (2) How is the pri-miR regulated?

Through our comprehensive studies in vitro and in vivo, we identified BCL-2 and IGF1R as robust target genes of the miR-424(322)/503 cluster in the mammary epithelium. Interestingly, our data did not validate previous data reporting CCNE1 as a target of this miRNA cluster (Forrest et al. 2010; Nakashima et al. 2010). Since the relevance of the different miRNA targets is context-dependent (Liu and Kohane 2009; Inui et al. 2010), it is important to mention that we used HMECs for our studies and achieved mature miRNA expression levels comparable with the observed during involution in vivo. Notably, when we augmented by lentiviral transduction the intracellular levels of the mature miRNAs several fold (>10 times) over the levels found in vivo, we observed a reliable reduction of CCNE1 protein (Supplemental Fig. S3F). Finally, no major changes were observed in CCNE1 protein levels from lactation to involution. Thus, our data illustrate the importance of the biological context in identifying relevant miRNA targets for different processes, and, consequently, we do not consider CCNE1 a relevant target of the miR-424(322)/503 in the context of mammary epithelial involution and TGF-β exposure.

During the last two decades, key details about the mechanism and regulation of mammary gland involution have been identified (Stein et al. 2007). The use of genetically modified mice with gain or loss of function has provided evidence regarding the implication of Bcl-2 and Igf1r in this process. The Igf1/Igf1r axis plays critical roles in multiple steps of mammary development (Hynes and Watson 2010). During involution, IGF signaling has been found to be inhibited by down-regulation of Igf1r (Modha et al. 2004) and up-regulation of insulin-like growth factor-binding protein 5 (Igfbp-5) (Allan et al. 2004). Importantly, the effect of IGF signaling in life or death fate decisions has been related to its effect on the PI3K–AKT/PKB and the Ras–MAPK pathways (Allan et al. 2004; Stein et al. 2007). In accordance with this, involution has been found to be inhibited in transgenic animals with constitutively active Akt (Schwertfeger et al. 2001; Ackler et al. 2002). BCL-2 family members are also involved in regulating cellular fate after the pregnancy cycle (Colitti 2012). Levels of proapoptotic Bak and Bad increase during lactation, reaching their peak during involution. In contrast, anti-apoptotic Bcl-W and Bcl-2 are strongly down-regulated at the onset of apoptosis (Metcalfe et al. 1999). In the proposed model, mammary epithelial cells are poised for apoptosis with higher levels of the death effectors but are prevented from triggering apoptosis by anti-apoptotic Bcl-2 family proteins until involution, when prosurvival gene expression levels decline (Metcalfe et al. 1999; Stein et al. 2007; Colitti 2012). In agreement with this view, enforced expression of Bak induces rapid mammary apoptosis (Metcalfe et al. 1999), while high levels of Bcl-2 have been shown to potently regulate mammary epithelial cell survival and inhibit alveolar cell apoptosis during involution (Jager et al. 1997; Schorr et al. 1999).

By understanding the signals that initiate the transcriptional up-regulation of the miR cluster, we uncovered its regulation by TGF-β. As TGF-β ligands have well-known roles in epithelial mammary gland morphogenesis (Nguyen and Pollard 2000; Stein et al. 2007), this information provided us with a potential link between the induction of involution and the up-regulation of the miR cluster. It is interesting to discuss that the R-SMAD that controls the up-regulation of the miR-424(322)/503, SMAD3, has been found to control not only transcription but also miRNA processing in a SMAD4-independent function (Davis et al. 2008, 2010). Thus, a question is whether this also applies for the production of mature miR-424(322) and miR-503. SMAD3 has been found to interact with DROSHA and affect miRNA maturation by binding to a conserved 5-base motif (SBE; 5′-CAGAC-3′) present in the pri-miRs. However the pri-miR-424(322)/503 cluster transcript does not contain this binding element. Additionally, individual silencing of both SMAD3 and SMAD4 completely abrogated the up-regulation of both pri-miR-424(322)/503 and individual mature forms. Thus, at this point, transcriptional up-regulation seems to be a critical step, and additional studies will be required to clarify a potential miRNA processing function of pri-miR-424(322)/503 by SMAD3.

Mechanistically, our data suggest a model (Fig. 7) in which milk stasis after weaning rapidly induces the expression of TGF-β3. Activation of the TGF-β pathway induces the transcription of pri-miR-424(322)/503. This primary transcript is processed to preferentially generate mature miR-424(322), which in turn down-regulates the expression of components of the involution response, IGF1R and BCL-2. Our data have unveiled a previously unknown multilayered regulation of epithelial tissue remodeling coordinated by the miRNA cluster miR-424(322)/503.

Figure 7.

Proposed mechanistic model of miR-424(322)/503 function. During mammary gland involution, critical events dictate the reorganization of the mammary epithelial cell compartment. TGF-β, a well-known player, governs involution in part by inducing the expression of a miRNA primary transcript containing the miR-424(322)/503 cluster. The generation of the mature miRNAs, mainly miR-424(322), reduces the expression of targets such as IGF1R and BCL2. This reduction results in attenuation of IGF signaling and sensitization to apoptosis, leading the epithelial cells toward cell death during involution. The above highlights the miR-424(322)/503 cluster as a critical regulator of mammary gland homeostasis through its ability to dictate cell life and death decisions through balancing proliferation, survival, and cell death pathways.

Our work may also have implications regarding tumor biology because several members of the miR-16 family have been involved in human malignancies (Liu et al. 2008; Aqeilan et al. 2010). All miR-424(322)/503 targets studied here have been shown to be associated with human cancers as oncogenes (Yip and Reed 2008; Pollak 2012). The ability of miR-424(322)/503 to modulate target expression raises the question of whether it is also involved in breast tumorigenesis. Our data in this regard reveal that knockout animals accumulate alveoli hyperplasia, but we did not observe tumor development in animals followed up to 1 yr of age. Longer periods of follow-up, multiple pregnancies, or both may be necessary to generate full cell transformation in these animals. Interestingly, miR-424(322)/503 is located in a chromosomal region that has been recently identified to present significant deletions in ∼8% of hormone receptor-positive breast cancers (Dvinge et al. 2013). Furthermore, the promoter region that we identified is placed on a CpG island that could be targeted by hypermethylation as a potential mechanism to down-regulate the expression of the miRNAs during transformation. As miR-424(322)/503 is located in the silenced part of the X chromosome, a single inactivation event is enough to fully abrogate its expression. Interestingly, it has been recently reported that suppressed miR-424(322) expression contributes to the progression of cervical cancer (Xu et al. 2013). Additional research will be necessary to fully define the potential role of miR-424(322)/503 in breast tumorigenesis. In this regard, our work describing the biological role of miR-424(322)/503 during involution and its regulation establishes the basis for future studies.

Materials and methods

Cell culture

Cell lines were obtained from the American Type Culture Collection. Normal breast epithelial MCF-10A and MCF-12 cells were cultured at 37°C/5%CO₂ in DMEM/Ham’s F-12 supplemented with 10% penicillin/streptomycin, 20 ng/mL EGF, 10 μg/mL insulin, 100 ng/mL cholera toxin, 500 ng/mL hydrocortisone, and 5% horse serum. HuMECs (Life Technologies, no. A10565) were grown according to instructions and reagents provided by the manufacturer. 184B5 cells were grown in MEGM medium (Lonza/Clonetics, no. CC-3150) supplemented with 1 ng/mL cholera toxin. Phoenix cells were grown in DMEM–10% fetal bovine serum and 10% penicillin/streptomycin at 37°C/5%CO₂.

TGFβ1 and TGFβ3 were purchased from Sigma (nos. T7039 and T5425, respectively). Human recombinant IGF1 was purchased from PROSPEC (no. CYT-216). Cisplatin (no. P4394) and paclitaxel (no. T7402) were both from Sigma.

miRNA extraction and HTP-seq

To perform mature miRNA and pre-miRNA expression analysis by qRT–PCR, RNA was purified using the miRVANA miRNA isolation kit (Ambion, no. AM1561) according to the instructions provided. Reverse transcription was performed on 50 ng of RNA using the miRNA reverse transcription kit (Roche, no. 4366596). miRNA deep sequencing was conducted at the Columbia University Genome Center’s core facility (http://genomecenter.columbia.edu) using the Illumina platform. RNA sequencing (RNA-seq) was performed at HiSeq single reads with a depth of coverage of 20 million reads per sample.

Luciferase reporter assays

3′ UTRs of specific target genes were cloned downstream from the luciferase reporter in the pMIR-REPORT vector (Life Technologies, no. AM5795M) by PCR from human genomic DNA using restriction enzymes. Relative luciferase units (RLUs) were measured using the Dual-Glo luciferase assay system (Promega, no. E2949). For a detailed description, see the Supplemental Material.

Mouse generation and LY2109761 treatment

To generate single miR-424(322) and miR-503 or double-knockout mice, we used ZNF (Sigma; for a detailed description, see the Supplemental Material). To inhibit TGF-β signaling during involution, lactating wild-type and miR-424(322)^−/−/miR-503^−/− double-knockout females were pretreated by oral gavage twice daily with 50 mg/kg LY2109761 (Selleckchem) for 24 h before weaning. Thereafter, mice were treated again by oral gavage twice daily with 50 mg/kg LY2109761 for 1 and 3 d. Mice were sacrificed, and mammary glands were dissected as mentioned previously. All experiments were performed under the Institutional Animal Care and Use Committee (IACUC) guidelines.

PAR-CLIP analysis

A PAR-CLIP assay to measure AGO2 enrichment in miR-424(322)/503 mRNA targets was performed as described previously (Hafner et al. 2010). Briefly, cells were pretreated with 50 μM 4-thiouridine (Sigma, no. T4509) overnight and cross-linked at 150 mJ/cm² at 365 nm UV on ice. Immunoprecipitation was performed using an anti-AGO2 antibody (Abnova, no. H00027161-M01) overnight at 4°C. RNA was extracted using the miRVANA miRNA isolation kit, and retrotranscribed RNA was subjected to analysis by qRT–PCR using specific primers.

Quantification of the number of miR-424(322) and miR-503 molecules

The number of molecules was estimated using chemically synthesized RNA oligonucleotides that mimic in sequence miR-424(322) and miR-503. RNA oligonucleotides were prepared in serial dilutions corresponding to 1 pg of miRNA per microliter, 0.1 pg of miRNA per microliter, 0.01 pg of miRNA per microliter, and 1 fg of miRNA per microliter. One microliter of each dilution was retrotranscribed in parallel with 50 ng of our samples. The number of molecules per cell was estimated by considering a ratio of 10 pg of RNA per cell and integrating the experimental Ct values into the standard values obtained with the commercial oligonucleotides.

Flow cytometry and cell viability

All flow cytometry experiments were performed on a BD FacsCalibur cell analyzer using the CellQuest Pro software on Mac OSX. To analyze apoptosis induced by cisplatin and paclitaxel, we performed the Annexin-V staining following the FITC Annexin-V apoptosis detection kit guidelines (BD Pharmingen, no. 556570). For a detailed description, see the Supplemental Material.

Whole-mount preparations and immunohistochemistry

To prepare whole-mount carmine red staining, mammary glands were fixed in 4% paraformaldehyde for 1 h, stained with carmine red (Sigma, no. C1022) overnight at room temperature, and ultimately dehydrated and fixed in xilol. To perform immunohistochemical analysis, formalin-fixed paraffin-embedded samples were deparaffinized and rehydrated. Peroxidase inactivation and antigen retrieval were achieved by incubating samples in 1% H₂O₂ and citric buffer. The antibodies used were Ki67 (1:200; Abcam, no. ab15580), anti-cytokeratin 18 antibody (1:300; Abcam, no. ab668), anti-cytokeratin 5 (1:1000; Covance, no. PRB-160P), cleaved caspase-3 (1:500; Cell Signaling, no. 9664), and p-SMAD2/3 (1:100; Cell Signaling, no. 9510).

Gene expression and miRNA expression arrays

For gene expression and miRNA expression arrays, MCF-10A cells were treated with TGFβ, and RNA was extracted using RNeasy and miRVANA extraction kits, respectively, and labeled using the LowInput QuickAmp labeling kit (Agilent, no. 5190-2331). RNAs were hybridized on a human GE 4x44K v2 microarray kit (Agilent, no. G4845A) and a human miRNA V2 oligo microarray (Agilent, no. G4470B), respectively. Gene expression array data have been uploaded and are available in the Gene Expression Omnibus database.

ChIP

MCF-10A untreated and TGFβ cells were cross-linked with 11% formaldehyde solution for 10 min at room temperature, lysed, and sonicated using a Bioruptor Standard sonication device (Diagenode, no. UCD-200 TM; see the Supplemental Material). HA-SMAD4 immunoprecipitation was performed using the DynaMag-2 magnetic particle concentrator (Invitrogen, no. 123.21D), protein G Dynabeads (Invitrogen, no. 100.03D), and an anti-HA antibody (Roche, no. 11867423001) overnight at 4°C on a rotator. Thereafter, samples were collected and washed, and phenol/chloroform/isoamyl alcohol extraction using Phaselock tubes (Prime, no. 2302840) was performed to isolate DNA. DNA enrichment was analyzed by qRT–PCR using specific primers. Data are presented as the Δ-normalized values between SMAD4-HA [Δ (TGFβ-treated) − (untreated)] − EV [Δ (TGFβ-treated) − (untreated)]. For a detailed description, see the Supplemental Material.

In situ hybridization

Briefly, probes against pri-hsa-miR-424(322)/503 were created from human genomic DNA by PCR using specific oligonucleotides and were obtained by T7-driven in vitro RNA synthesis. Probes against mature miR-424(322) and miR-503 were purchased from Exiqon (catalog no. 38195-08), and the assay was visualized using a tyramide signal amplification method (PerkinElmer, NEL741001KT). For a detailed description, see the Supplemental Material.