Coregulation of Genetic Programs by the Transcription Factors NFIB and STAT5

Abstract

Mammary-specific genetic programs are activated during pregnancy by the common transcription factor signal transducer and activator of transcription (STAT) 5. More than one third of these genes carry nuclear factor I/B (NFIB) binding motifs that coincide with STAT5 in vivo binding, suggesting functional synergy between these two transcription factors. The role of NFIB in this governance was investigated in mice from which Nfib had been inactivated in mammary stem cells or in differentiating alveolar epithelium. Although NFIB was not required for alveolar expansion, the combined absence of NFIB and STAT5 prevented the formation of functional alveoli. NFIB controlled the expression of mammary-specific and STAT5-regulated genes and chromatin immunoprecipitation-sequencing established STAT5 and NFIB binding at composite regulatory elements containing histone H3 lysine dimethylation enhancer marks and progesterone receptor binding. By integrating previously published chromatin immunoprecipitation-sequencing data sets, the presence of NFIB-STAT5 modules in other cell types was investigated. Notably, genomic sites bound by NFIB in hair follicle stem cells were also occupied by STAT5 in mammary epithelium and coincided with enhancer marks. Many of these genes were under NFIB control in both hair follicle stem cells and mammary alveolar epithelium. We propose that NFIB-STAT5 modules, possibly in conjunction with other transcription factors, control cell-specific genetic programs.

The discovery of DNA binding activity in nuclear extracts of vertebrate cells to specific sequences in the adenovirus genome (1), the chicken lysozyme gene (2), and the immunoglobulin locus (3) led to the isolation and cloning of the first family of sequence-specific eukaryotic DNA binding proteins. The nuclear factor I (NFI) family consists of four genes, Nfia, Nfib, Nfic, and Nfix (4–6). Genetic studies in mice have established unique and cell-specific roles for each of them (7). Nfib-null mice die perinatally due to impaired lung maturation and neuronal defects (8–10). In hair follicle stem cells (HFSCs), nuclear factor I/B (NFIB) is a negative regulator of melanocyte stem cells (11).

Development and differentiation of mammary tissue during pregnancy are largely driven by progesterone and prolactin, with the latter signaling through the generic transcription factors signal transducer and activator of transcription (STAT) 5A and STAT5B (referred to as STAT5 throughout the manuscript) (12). The up to 10 000-fold induction of genes encoding milk proteins and lipid metabolism in secretory alveolar mammary epithelium during pregnancy is dependent on the presence of STAT5 and its binding to promoter and enhancer sequences (13, 14). However, it is not known to what extent additional transcription factors synergize with STAT5 in establishing cell-specific gene expression. NFIB binds to specific sequences (TGGCA) in the Wap gene promoter (15, 16) and transgenic experiments have demonstrated that these sites are required for promoter activation (17). In contrast, a disabling mutation in the juxtaposed STAT5 binding site mainly abolished increased expression (17). NFI binding sites have also been identified in other genes highly expressed in mammary epithelium, including bovine β-lactoglobulin (18), β1,4-galactosyltransferase (B4galt1) (19), carboxyl ester lipase (Cel) (20, 21), TRPM-2/clusterin (Clu) (22), p53 (Trp53) (23), and the mouse mammary tumor virus (MMTV) promoter (24–26). These findings suggest that the NFI transcription factor family may contribute to the activation of genes during pregnancy. Recent studies have also demonstrated a causative role for NFIB in estrogen receptor–negative breast cancers (27) and in small cell lung cancer (28).

This study tested the hypothesis that genes preferentially expressed in mammary epithelium depend on promoter sequences bound by STAT5 and members of the NFI family.

Materials and Methods

Mice

Mice carrying floxed Nfib (10) and Stat5 (29) genes and Wap-Cre (WC) and MMTV-Cre (MC) (30) transgenes have been described. Nfib^f/f;WC females were mated and mammary tissue was harvested at specified pregnancy stages or within less than 12 hours after parturition. Nfib^f/f;MC animals are not viable, and epithelial development was studied upon transplantation of embryonic anlagen as described below. Mice carrying floxed Nfib and Stat5 alleles, and the WC transgene were generated by breeding. All animals were kept in accordance with National Institutes of Health (NIH) guidelines, and the experiments were approved by the Animal Care and Use Committee of the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK).

Mammary tissue transplantation

Athymic nude mice (3 weeks old) were anesthetized with an intraperitoneal injection of avertin. Fat pads were cleared by excising the proximal part of the inguinal gland containing the mammary epithelium. Mammary buds from embryonic day 13.5 Nfib^f/f;MC embryos were transplanted as described previously (31). Nfib^f/+;MC anlagen were used as controls by transplantation into the contralateral fat pad. To assess complete clearing, the excised tissues were processed for whole-mount staining. Eight weeks after transplantation, fat pads were harvested either from virgin hosts, or the hosts were mated and transplanted tissue was harvested on day 6 or day 13 of pregnancy or the day of parturition within less than 12 hours after delivery.

Histology

Harvested mammary tissues were fixed in 10% formalin, dehydrated through ethanol and xylene, embedded in paraffin, and sectioned. Sections were stained with hematoxylin and eosin by standard methods. For immunostaining, antigen unmasking was performed in a decloaking chamber (Biocare Medical) using Borg Decloaker solution, pH 9.5 (Biocare Medical), at 125°C and 18 to 24 psi for 5 minutes. Sections were blocked for 30 minutes in Tris-buffered saline with Tween 20 with 3% goat serum. Primary antibodies were incubated overnight at 4°C (anti-phosphorylated STAT5, 1:200 [Cell Signaling]; anti-NKCC1, 1:1000 [a gift from Dr Jim Turner, National Institute of Dental and Craniofacial Research, NIH]; anti-perilipin 2 (PLIN2), 1:1000 [a gift from Drs Dean Landos and Alan Kimmel, NIDDK, NIH]; anti–smooth muscle actin, 1:1000 [Sigma-Aldrich]; and anti-E-cadherin, 1:200 [BD Biosciences]). Alexa Fluor 488– or 594–conjugated secondary antibodies (Invitrogen) were applied at a dilution of 1:400 for 30 minutes at room temperature.

RNA-sequencing (RNA-seq) data processing

Poly(A) RNA was purified twice from 1 μg of total RNA. cDNA was synthesized using SuperScript II (Invitrogen) and a TruSeq RNA Sample Preparation Kit (Illumina Inc) and sequenced using a HiSeq 2000 system (Illumina Inc). The single-end reads of biological replicates obtained from each sample were aligned to the mouse reference genome (mm9 assembly) using the TopHat program (32). The total number of the mapped reads on each gene was calculated with the HTSeq program (http://www-huber.embl.de/users/anders/HTSeq/). Transcript abundance was estimated by means of reads per kilobase of transcript per million mapped reads (RPKM) according to the mapped reads on exons as described previously (33). The RNA-seq data are deposited in Gene Expression Omnibus (GEO) under the accession number GSE42878. For Supplemental Figure 1, we obtained RNA-seq data from the GEO repository (GSE29278) (34) and processed them with the same procedure as outlined above.

ChIP-seq analysis

Previously published ChIP-seq (chromatin immunoprecipitation sequencing) data for STAT5 (GSE52016 and GSE40930), the progesterone receptor (PR) (GSE42887), NFIB (GSE42900), and histone H3 lysine dimethylation (H3K4me2) (GSE25105) were analyzed. NFIB antibodies were obtained from Santa Cruz Biotechnology. ChIP-seq analysis was conducted as described previously (13, 14).

Motif analysis

We used two different algorithms to pinpoint transcription factor binding motifs. To identify significantly overrepresented motifs in the promoter of NFIB- and STAT5-regulated genes (see Figure 6), we used the Pscan tool (35) with the JASPAR motif database (36). To scan all 26 237 STAT5 binding sites with 130 known position weight matrices from the JASPAR database, the MOODS algorithm was used with a P value cutoff of .001 (37). Flanking sequences of each STAT5 binding site (500 bp) were scored against the matrices, and only the maximum score within each site per matrix was used for motif score calculation. The background motif score was estimated by the same approach with a randomly generated set. The background score was subtracted before the final motif score.

Figure 6.

NFIB binding motif significantly co-occur near in vivo STAT5 binding sites. A, Common and unique genes, whose expression is dependent on NFIB and/or STAT5. STAT5 RNA-seq data were obtained from the GEO database (GSE37646). KO, knockout. B, Significantly overrepresented motifs were identified in promoters of the NFIB- and STAT5-regulated genes. C, Gene expression levels were determined by RNA-seq and mean expression levels of gene sets are shown. D, The NFI binding motif significantly co-occurs near in vivo STAT5-binding sites. Motif analysis identified NFI and GAS motifs as highly co-occurring near in vivo STAT5 binding sites; 130 defined motifs from the JASPAR database were used. Each spot indicates a single motif and sequence logos of the top 7 motifs (dashed box) are shown (right).

Gene ontology analysis

Genes showing more than 5 RPKM and 2-fold up-regulated expression in the wild type (WT) compared with the absence of NFIB (or STAT5) were categorized into 3 groups according to coregulation (NFIB-specific, STAT5-specific, and NFIB/STAT5-coregulated gene sets). Functional annotation of the genes in each group was assessed by using the DAVID tool (38).

Data visualization

To visualize the next generation sequencing data (see Figures 7 and 8), total mapped reads of RNA-seq were scaled to 1 million and total mapped reads of ChIP-seq data were scaled to 10 million by using the HOMER program (39). Then, the integrative genomics viewer tool was used to load the scaled reads onto the mouse genome (40).

Figure 7.

NFIB and STAT5 are key regulators in mammary gland development. A, Gene ontology (GO) analysis showed that most NFIB- and STAT5-regulated genes were associated with mammary gland development. B, Expression levels of the clustered genes. C, Genome browser images of 3 representative genes are shown. The first and second tracks represent expression levels (tags were normalized to 1 million) of exons in 2 independent studies (Stat5-null and Nfib-null). Expression levels of the genes in knockout (KO; black bar graph) relative to WT (gray line graph) are shown. The rest of tracks show NFIB and STAT5 binding (day 13 of pregnancy) as well as H3K4me2 enrichment (virgin) as peaks. The STAT5 ChIP-seq and H3K4me2 data in mammary gland were obtained from a previous study (GSE40930 and GSE25105). Actual sequences of NFI and GAS motifs located in CRMs (gray bars) are represented (bottom).

Figure 8.

Combinatorial binding of NFIB, PR, and STAT5 is observed on putative enhancers. Gray bars represent joint binding sites of the transcription factors. Genome browser images of Rankl (A), Csn3 (B), and Adamts3 and Mlf1ip (C) are shown with PR, NFIB, STAT5, H3K4me2 (K4me2), and histone H3 lysine 4 trimethylation (K4me3) ChIP-seq peaks. MG, mammary gland tissue.

Results

Deletion of Nfib from mouse mammary epithelium

Immunofluorescence (IF) demonstrated the presence of nuclear NFIB in luminal and basal mammary epithelial cells of virgin, pregnant, and lactating mice (Figure 1), and the staining intensity was similar at all stages of development. In tissue from virgin mice, some stromal cells, possibly of immunological nature, are also positive for NFIB.

Figure 1.

NFIB (red; arrowhead) is present in mammary luminal cells throughout development. A, Virgin. B, Pregnancy day 6 (p6). C, Pregnancy day 13 (p13). D, Day 1 of lactation (L1). E-cadherin (green) outlines luminal epithelial cells. Nuclei were visualized with 4′,6-diamidino-2-phenylindole (blue). The nature of the NFIB-positive stromal cells in mammary tissue from virgin mice is not clear, but they could represent immune cells.

MMTV-Cre (MC) and WAP-Cre (WC) transgenes were used to inactivate the Nfib gene in mammary stem cells and in differentiating alveolar epithelium during pregnancy, respectively. Nfib^f/+;MC females appeared normal and were able to nurse their pups. Nfib^{f/f; MC} mice were not obtained, possibly due to perinatal lethality. To bypass this obstacle, mammary anlagen isolated from Nfib^f/f;MC fetuses were transplanted into cleared fat pads of recipient immunocompromised mice. Mammary tissue was analyzed at days 13 and 18 of pregnancy (Figure 2). Overall, WT and mutant alveolar epithelium had a similar appearance, including the accumulation of milk fat globules, indicating secretory activity and functional differentiation. These results demonstrate that during pregnancy NFIB is not required for alveolar development or expansion or for the initial stages of differentiation.

Figure 2.

Alveolar development during pregnancy in the absence of NFIB. Deletion of Nfib in mammary stem cells using the MMTV-Cre transgene. Alveoli (arrows) are sparser in the fat pads of Nfib^f/f;MC mice (B) than in those of Nfib^f/+;MC mice (A) on day 13 and day 18 (D and C) of pregnancy. Secreted material and lipid droplets are present in tissues of both genotypes.

Nfib^f/f;WC females were fertile and gave birth to normal-sized litters. There was increased pup mortality and only 40% of pups born were weaned. However, the surviving pups grew at the same rate as pups reared by control dams (data not shown). Histological analysis of mammary tissue from Nfib^f/f;WC females at parturition demonstrated the presence of alveoli (Figure 3). However, they were sparser and tended to be smaller in size. IF confirmed the absence of NFIB in most alveoli (Figure 4). The intensity of phosphorylated STAT5 (pSTAT5) staining at day 1 of lactation was similar in control and mutant mice (Figure 4), suggesting that prolactin signaling was intact in the absence of NFIB. WAP, a major milk protein (41), was preferentially present in the alveolar lumina of control tissue and to some extent in the cytoplasm (Figure 4). The cytoplasmic presence of WAP was more prominent in mutant tissue, suggesting impaired secretion. The distribution of PLIN2, an integral component of the lipid droplet membrane, supports the observation that secretion is impaired in mutant tissue (Figure 4). However, secretion is sufficiently intact to produce milk for pups to survive.

Figure 3.

Alveolar development during pregnancy in the absence of NFIB. Deletion of Nfib in differentiating mammary alveolar cells during pregnancy using the Wap-Cre transgene. Alveoli (arrows) are sparser in the fat pads of Nfib^f/f;WC mice (B and D) than in those of Nfib^+/+;WC mice (A and C). A, adipocyte; L, alveolar lumen.

Figure 4.

Immunohistochemical analysis of the epithelial differentiation status of Nfib^f/f;WC epithelium at day 1 of lactation. A and B, NFIB staining (in red; arrow) in control (A) and mutant (B) tissue. The broad arrow indicates incomplete deletion in cells by the WAP-Cre transgene, which is not expressed in ducts and in a mosaic fashion in luminal cells. E-cadherin (green) is used as a counterstain. C and D, pSTAT5 staining in Nfib^+/+;WC (C) and Nfib^f/f;WC (D) epithelium. pSTAT5 (arrow; red fluorescence) is concentrated in nuclei of luminal cells and E-cadherin is stained green. E and F, WAP (red) staining in mammary tissue from Nfib^+/+;WC (E) and Nfib^f/f;WC (F) mice. E-cadherin staining in green indicates epithelial cell membranes. G and H, Staining for PLIN2, a component of the milk fat globule membrane in mammary tissue from Nfib^+/+;WC (G) and Nfib^f/f;WC (H) mice. Green staining of smooth muscle actin (SMA) demonstrates normal development of myoepithelial cells.

The combined function of NFIB and STAT5 in lactation was investigated in mice carrying floxed Stat5 and Nfib alleles and the Wap-Cre transgene. Mammary development in these mice appeared heterogeneous, and IF analyses demonstrated large areas positive for pSTAT5 next to smaller regions that had underdeveloped alveoli devoid of pSTAT5 (Figure 5). This result provides genetic evidence that functional alveolar epithelium cannot be established in the combined absence of NFIB and STAT5 and unrecombined cells preferentially expand during pregnancy.

Figure 5.

Alveolar development at day 1 of lactation in the absence of NFIB and STAT5. A, Whole mount of mammary tissue of Nfib^f/fStat5^f/f;WC dams shows attenuated alveologenesis. B, Heterogeneous alveolar development and incomplete filling of the fat pad is seen. The tissue consists of a mixture of mature, expanded (arrow), and small, underdeveloped (arrowhead) alveoli. C, Presence of pSTAT5 staining (red, arrowheads) in some alveoli next to alveoli that have lost STAT5 (asterisk) indicates mosaic expression of the Wap-Cre transgene. D, Cytoplasmic localization of WAP (red, arrowhead) and lack of WAP in the alveolar lumen (asterisk) suggest impaired secretion. E-cadherin is stained green, and nuclei are stained blue with 4′,6-diamidino-2-phenylindole. E and F, Hematoxylin and eosin–stained section of mammary tissue from Nfib^f/fStat5^f/f;WC dams at the onset of the second lactation shows the presence of expanded alveoli (arrow) interspersed with underdeveloped alveoli (arrowhead).

Identification of NFIB target genes

RNA-seq analyses were performed from mammary tissue harvested from Nfib^f/f;WC and WT mice at parturition. Expression of approximately 370 genes was reduced at least 2-fold in the absence of NFIB (Figure 6A and Supplemental Table 1). These include most of the STAT5 target genes that encode proteins linked to alveolar differentiation during pregnancy (13, 14). An unbiased de novo motif search identified GAS motifs (STAT5 binding sites) and NFIB binding motifs in promoter regions of these genes (Figure 6B). Of the NFIB-regulated genes, 75% (279 of 370) were also under the control of STAT5, and their expression was reduced between 75% and 99% in the absence of either STAT5 or NFIB (Figure 6, A and C). In contrast, expression of the 472 genes only under STAT5 control was reduced up to 80% in the absence of STAT5 but not NFIB. Similarly, reduced expression of the 91 NFIB-specific target genes was only observed in Nfib-null tissue (Figure 6C). Notably, NFIB and STAT5 had differential importance in the expression of milk protein genes (Table 1). Motif analysis was conducted on a 500-bp fragment overlapping all 26 237 in vivo STAT5 binding sites (14) using the MOODS algorithm with predefined 130 motif position weight matrices. The GAS motif was by far the most significantly enriched motif around the STAT5 binding sites (Figure 6D). The next 5 high-ranked motifs were GABPA, FEV, SPI1, ELF5, and SPIB motifs, which contain a half sequence of the GAS motif (TTC or GAA). Upon exclusion of these candidate partial GAS motifs, the known NFI binding motif was the most significantly enriched motif near the genome-wide STAT5 binding sites.

Table 1.

Expression Levels of Genes Coregulated by NFIB and STAT5

Gene	NFIB			STAT5
	WTa	Nulla	Fold change	WTa	Nulla	Fold change
Csn2	143 846	50 688	2.8	193 547	557	347
Csn1s2a	120 513	49 693	2.4	164 023	470	348
Csn1s1	61 562	15 930	3.9	87 949	230	382
Wap	102 755	9085	11.3	109 772	327	335
Csn3	34 488	8794	3.9	57 840	131	440
Glycam1	102 755	8021	12.8	102 504	265	385
Trf	16 334	6734	2.4	21 091	589	35
Expi	7003	2012	3.5	11 627	375	30
Lalba	8155	914	8.9	8626	21	410
Csn1s2b	11 359	712	16.0	13 610	12	1133

^a Normalized expression level (RPKM).

Gene ontology analysis was performed with 472 STAT5-specific, 91 NFIB-specific, and 279 coregulated genes. As expected, overall functional outputs of these genes were mostly related to mammary gland development (Figure 7A). Coregulated genes appear to contain functional outputs of both NFIB- and STAT5-specific gene sets. They also contained a unique group of milk protein genes (Supplemental Table 2). Expression levels of these genes were reduced between 50% and 95% in the absence of either NFIB or STAT5 (Figure 7B). Among these 3 gene sets, the coregulated genes encompassed mammary-specific genes (Supplemental Figure 1). Manual investigation further demonstrated that a STAT5-NFIB cis-regulatory module (CRM), a DNA fragment containing both STAT5 and NFIB binding motifs, is the integral component of many genes expressed in mammary epithelium during pregnancy (Figure 7C). For instance, the Lalba, Wap, and Csn3 genes harbor1 STAT5-NFIB CRM upstream of their transcription start sites, and their expression levels were greatly reduced in the absence of either NFIB or STAT5. All of the STAT5-NFIB CRMs near the 3 representative genes consist of conventional NFIB and GAS motifs. NFIB ChIP-seq experiments confirmed NFIB binding to these CRMs (Figure 7C). NFIB binding was obtained at bona fide NFIB binding motifs (TGGCA) juxtaposed to STAT5 binding sites at the mammary-specific genes.

STAT5 binds not only to promoter proximal sequences but also to putative enhancer sequences (13, 14, 42). Because STAT5, the PR (43), and NFIB are key mammary transcription factor, it can be hypothesized that their joint binding could be a defining feature of at least some mammary enhancers. This hypothesis was explored through the integration of NFIB, STAT5 (13, 14), PR (43), and H3K4me2 (44) ChIP-seq data sets. Binding of these 3 transcription factors was located at known enhancers (Figure 8A), such as the one located 75 kbp 5′ of the Rankl (Tnfsf11) gene (45, 46). Frequently areas recognized by both NFIB and STAT5 contain either a GAS motif or an NFIB motif but not both, suggesting that either of these transcription factors can be the guide or follower. Although the mammary specificity of some milk protein gene promoters has been established in transgenic mice (47–50), genuine enhancers activating genes during pregnancy have not been identified. Based on sequence conservation and DNase I hypersensitivity, an intergenic region in the casein locus has been proposed as an enhancer (51). ChIP-seq–based data demonstrated the presence of H3K4me2 marks and binding of STAT5, NFIB, and PR to a 300-bp sequence in this region (Figure 8B). NFIB-binding motifs (TGGCA) but no GAS motifs were identified, suggesting a role for STAT5 as a cofactor.

Search for NFIB-STAT5 modules in nonmammary cells

If NFIB-STAT5 modules are operational in other cell types, skin and its appendages are qualified candidates because they are evolutionary related to the mammary lineage. NFIB ChIP-seq data from HFSCs (11) were used to search for potential NFIB-STAT5 modules. NFIB occupied 1449 sites in the genome of HFSCs, and expression of some target genes was altered in Nfib-null HFSCs (11). NFIB binding to gene regulatory regions in HFSCs coincided with STAT5 binding in mammary epithelium (Figure 8C). Moreover, these sequences were also recognized by PRs and exhibited H3K4me2 marks as shown for Adamts3 and Mlf1ip (Figure 8C). Notably, genes under NFIB control in HFSCs are also regulated by NFIB and STAT5 in mammary alveolar epithelium. Although STAT5 ChIP-seq data are not available for HFSCs, it can be predicted that, as in mammary epithelium, NFIB-STAT5 modules control the expression of specific gene sets.

Discussion

Twenty years ago, Li and Rosen (17, 52) published landmark studies demonstrating the critical importance of NFI binding sites in the rat Wap gene promoter for its expression in transgenic mice and the relevance of a juxtaposed STAT5 binding site in controlling prolactin-induced activation during pregnancy. Based on these studies, it was hypothesized that the combination of transcription factors might control expression of genes in mammary tissue. This investigation now demonstrates that more than 250 mammary-specific genes are under joint control by STAT5 and NFIB, suggesting that an NFIB-STAT5 module is key in the activation of genes in mammary alveolar epithelium during pregnancy. Moreover, genes under NFIB control in HFSCs are also regulated by NFIB and STAT5 in mammary tissue. These findings establish the concept that the NFIB-STAT5 transcription factor module is critical for the activation of genetic programs in 2 distinct, yet evolutionary related, cell types.

Deletion of Nfib from mammary stem cells and differentiating alveoli had no detrimental impact on the development of mammary alveoli, and the slightly impaired differentiation was reflected by the incomplete activation of milk protein genes. The degree by which steady-state levels of individual mRNAs were reduced possibly reflects the state of epithelial differentiation. Expression of approximately 270 genes during pregnancy is jointly controlled by STAT5 and NFIB with most them being preferentially active in mammary epithelium. In general, the impact of STAT5 on the expression of these genes greatly exceeds that of NFIB because for many of them the presence of STAT5 is mandatory for any induction during pregnancy. In contrast, pregnancy-induced expression is only attenuated in the absence of NFIB (Figure 9). The Janus tyrosine kinase 2/STAT5 pathway is probably intact in the absence of NFIB because expression of a subset of STAT5 target genes remained unaltered.

Figure 9.

Genetically proven model of mammary gland–specific gene regulation by NFIB and STAT5. The 279 genes that are cogoverned by NFIB and STAT5 are specifically expressed in mammary gland tissue (Supplemental Figure 1). Regulatory regions of most of these genes contain STAT5-NFI CRMs, which bind STAT5 in mammary tissue and also contain NFI consensus sequences (left panel). In the absence of NFIB or STAT, expression of coregulated genes is reduced up to 75% or 98%, respectively, compared with the WT (right panel).

Expression of NFIB target genes in mammary epithelium is probably also controlled by additional members of the NFI family. Support for this notion comes from studies with a mutant Wap transgene that was silenced upon disabling of NFI binding sites (17), whereas loss of NFIB does not result in the complete inactivation of the Wap gene. Although NFIB might be the most abundant member in the mammary epithelium, all four Nfi members are expressed (16). Wap transgenes containing only promoter-bound STAT5 binding sites are never fully activated during pregnancy (48, 53), and an additional putative upstream STAT5 site (14) might be required for full activation (49).

The presence of NFIB-STAT5 modules is not only based on the presence of the respective binding motifs but also mostly on the in vivo binding of these two transcription factors, and their biological significance was established in the respective gene knockout mice. NFIB-STAT5 modules are also likely to be active in HFSCs. Transcription factor modules provide an efficient means to control cell-specific genetic programs as has been shown for Smad3 and Oct4, MyoD1 and Pu.1 (54), and Nkx3.1 with Myc (7). In primary mammary cells, expression of the Tnfsf11 (RankL) gene is controlled by progesterone, and PR and STAT5 binding has been localized to at least 5 enhancers within 80 kbp (45). ChIP-seq data analyzed in this study provide evidence for the binding of STAT5, PR, and NFIB and the presence of H3K4me2 marks preferentially at the D5 region located 75 kbp upstream of the transcriptional start site, indicating that this is the key enhancer. An NFIB-STAT5-PR-H3K4me2 module also marks a putative enhancer region within the casein locus (44, 55, 56). Although it is generally believed, and correctly so, that STAT5 and NFIB bind to their respective recognition sites in regulatory regions, their interaction with enhancers is more complex. NFIB, but not STAT5, motifs are found in the Tnfsf11 and casein enhancers, suggesting that NFIB is the driver and STAT5 is the passenger. These findings narrow the knowledge gap on the role of transcription factor modules in regulating cell-specific and hormone-controlled genetic programs.

Additional material

Supplementary data supplied by authors.