Cell-specific and shared regulatory elements control a multigene locus active in mammary and salivary glands

Lee, Hye Kyung; Willi, Michaela; Liu, Chengyu; Hennighausen, Lothar

doi:10.1038/s41467-023-40712-0

Download PDF

Article
Open access
Published: 17 August 2023

Cell-specific and shared regulatory elements control a multigene locus active in mammary and salivary glands

Nature Communications volume 14, Article number: 4992 (2023) Cite this article

1368 Accesses
1 Citations
1 Altmetric
Metrics details

Subjects

Abstract

Regulation of high-density loci harboring genes with different cell-specificities remains a puzzle. Here we investigate a locus that evolved through gene duplication and contains eight genes and 20 candidate regulatory elements, including one super-enhancer. Casein genes (Csn1s1, Csn2, Csn1s2a, Csn1s2b, Csn3) are expressed in mammary glands, induced 10,000-fold during pregnancy and account for 50% of mRNAs during lactation, Prr27 and Fdcsp are salivary-specific and Odam has dual specificity. We probed the function of 12 candidate regulatory elements, individually and in combination, in the mouse genome. The super-enhancer is essential for the expression of Csn3, Csn1s2b, Odam and Fdcsp but largely dispensable for Csn1s1, Csn2 and Csn1s2a. Csn3 activation also requires its own local enhancer. Synergism between local enhancers and cytokine-responsive promoter elements facilitates activation of Csn2 during pregnancy. Our work identifies the regulatory complexity of a multigene locus with an ancestral super-enhancer active in mammary and salivary tissue and local enhancers and promoter elements unique to mammary tissue.

Simultaneous single-cell three-dimensional genome and gene expression profiling uncovers dynamic enhancer connectivity underlying olfactory receptor choice

Article Open access 15 April 2024

Inferring gene regulatory networks from single-cell multiome data using atlas-scale external data

Article Open access 12 April 2024

Evolution of tissue-specific expression of ancestral genes across vertebrates and insects

Article 15 April 2024

Introduction

Mammals arose from the synapsid lineage about 160 million years ago in the middle Jurassic, and all extant mammals rely entirely on milk to nourish their young¹. The transformation of proto-lacteal fluid rich in calcium-binding proteins into nutritious milk was vital for the sustained success of mammals. The innovation of milk and saliva^2,3 was largely driven by the expansion of the secretory calcium-binding phosphoprotein (SCPP) gene family, the acquisition and evolution of genetic regulatory elements conveying gene expression to either mammary or salivary secreting cells and the diversification and repurposing of proteins. The five caseins (CSN1S1, CSN2, CSN1S2A, CSN1S2B, CSN3) are cardinal milk proteins accounting for up to 80% of total milk protein⁴, and their gene arrangement within the casein locus partially reflects its evolutionary history with Csn3 juxtaposed to its ancestral precursor Fdcsp (Follicular dendritic cell secreted protein) and Odam (Odontogenic ameloblast associated) ^5,6.

The extended casein locus is of particular interest as it harbors at least eight genes expressed in mammary or salivary glands or both^3,5,6. While the five Casein genes are dominantly expressed in mammary tissue and highly activated during pregnancy, Fdcsp is preferentially active in salivary tissue and in immune cells, and Odam is expressed in mammary and salivary tissues. The casein locus provides a unique window into the evolution of a multigene locus and co-evolving regulatory elements targeting the expression of individual genes to at least four distinct cell types, mammary, salivary, tooth-associate tissues and specific B cells.

Over the past 25 years, major mammary-centric hormones, receptors, tyrosine kinases and transcription factors have been deleted in mice resulting in dramatic defects in mammary development and function^{7,8,9,10,11,12,13}, frequently leading to lactation failure because of the absence of a differentiated secretory alveolar compartment. However, a major challenge is discerning primary effects from secondary ones. For example, loss of the transcription factor STAT5 prevents the activation of alveolar differentiation programs during pregnancy and the failure to launch milk protein gene expression might be, in some cases, a secondary consequence. Thus, understanding the control mechanisms of genes during normal physiology requires the inactivation of candidate regulatory elements rather than the global or cell-specific ablation of transcription factors.

Our data presented here on the individual and combinatorial deletion of 12 regulatory elements within a locus encoding eight genes provides evidence of a complex super-enhancer active in mammary and salivary tissue, gene-specific local enhancers and a cytokine-responsive promoter that activate genes 10,000-fold in mammary tissue during pregnancy.

Results

Enhancer structure in a locus active in mammary tissue

The extended casein locus spanning approximately 330 kbp encodes at least eight genes (Fig. 1a) with distinct expression patterns. While the five Casein genes (Csn1s1, Csn2, Csn1s2a, Csn1s2b, and Csn3) are known to be expressed almost exclusively in mammary tissue, Fdcsp and Odam are active preferentially in salivary glands. First, we gauged the absolute and relative RNA levels of these genes in mammary glands at day one of lactation (L1) and salivary tissue through RNA-seq (Fig. 1b). Combined, the mRNA levels of the five caseins account for more than 50% of mRNA in mammary tissue, and up to 10⁶ reads were recorded for individual Casein genes. Casein gene expression in salivary tissue is four to five orders of magnitude lower. While expression of Fdcsp is confined to salivary tissue, Odam mRNA levels are similar in both mammary and salivary tissue. Multispecies comparisons (Supplementary Fig. 1a) highlighted the complexity of this locus⁵ and identified an evolutionary conserved region as a candidate regulatory element activating this locus during pregnancy^14,15.

**Fig. 1: Characteristics of the *Casein* locus.**

Next, we measured the expression of these genes throughout mammary development, from non-parous mice to day 10 of lactation (L10) (Fig. 1c). mRNA levels for the five caseins ranged from 100 to 1000 normalized read counts in virgin mice and peaked at approximately 10⁷ normalized read counts at L10, an increase of four to five orders of magnitude. Overall, the most prominent induction was between day six of pregnancy (p6) and L1, the time window of heightened placental lactogen- and prolactin-mediated epithelial cell differentiation. Another 10-fold induction was observed between L1 and L10 (Supplementary Data 1). In contrast, Odam expression was only observed at L1 and L10, and mRNA levels were four to five orders of magnitude lower than the caseins.

The structure of this locus, with salivary genes nestled between mammary genes, makes it an ideal case study on cell-specific and hormone-activated regulatory elements. Candidate regulatory elements were identified based on the presence of H3K27ac marks and transcription factor (TF) occupancy (Fig. 1d, e). As anticipated, H3K27ac marks at the five Casein genes were restricted to mammary tissue. However, H3K27ac coverage upstream of Odam, the ancestral gene of the locus, was detected in both mammary and salivary tissue, pointing to shared regulatory elements operative in both tissues. This region also harbors a 147 bp long evolutionarily conserved region (ECR) identified in mammals⁵.

Digging deeper, we explored binding of mammary-centric TFs¹⁶, such as the cytokine-inducible transcription factor Signal Transducer and Activator of Transcription (STAT) 5, the glucocorticoid receptor (GR), Nuclear Factor IB (NFIB) and mediator complex subunit 1 (MED1) (Fig. 1e). TF binding coincided with H3K27ac and H3K4me1 marks and a total of 20 candidate regulatory elements were identified within the 330 kb locus. In addition, CTCF binds to four distinct sites within this locus¹⁷. The H3K27ac marked region upstream of Odam contains four STAT5-bound regions and has all hallmarks of a super-enhancer (SE) as defined by the Rose algorithm (Fig. 1e, area highlighted in red). RNA polymerase II (Pol II) coverage was most prominent in the Casein genes. Integration of active histone marks, transcription factor binding, and gene expression data suggest that the mammary-salivary locus harbors the highest density of candidate enhancers and highly regulated genes among all multigene loci in mammary tissue (Supplementary Fig. 1b–d, Supplementary Table 1).

Although all 20 candidate regulatory elements were bound by STAT5, a principal TF controlling mammary development and function⁸, only 12 STAT5 peaks coincided with genuine DNA binding motifs (GAS, interferon-Gamma Activated Sequence) (Supplementary Table 2), suggesting that STAT5 binding at other sites is mediated by other TFs, such as NFIB or GR. An inherent diversity in anchor proteins could lead to seemingly identical enhancers, with possibly different and unique activities. STAT5 binding was also detected at the promoter regions of Csn1s1, Csn2, Csns2a and Csn1s2b and coincided with bona fide H3K4me1 enhancer marks^18,19 (Supplementary Fig. 2). As expected, H3K4me3 marks were exclusively associated with promoter regions and Pol II coverage was preferentially over gene bodies.

Super-enhancer activity in mammary tissue during pregnancy

To identify dual regulatory elements controlling genes in mammary and salivary glands, we focused on the candidate SE located between Csn1s2b and Odam, the ancestral gene expressed in both tissues. The SE is composed of four constituent enhancer modules (SE-E1, SE-E2, SE-E3 and SE-E4) spanning a total of 10 kbp (Fig. 2a and Supplementary Fig. 3) and we generated mice carrying individual and combinatorial enhancer deletions (Supplementary Fig. 4). Deletion of the entire 10 kbp SE (∆SE) was confirmed by sequencing and the absence of TF binding and H3K27ac marks in this region (Fig. 2c). Female mice lacking the SE were unable to support their litters due to lactation failure and analyses were conducted in mammary tissue at day 18 of pregnancy (p18), just prior to delivery or within 12 h after delivery (post-partum <12 h, pp <12 h). This was critical as lactation failure results in the loss of cell differentiation and loss of STAT5 phosphorylation and tissue remodeling is observed approximately 24 h after parturition in a mouse model²⁰. Of note, the C57BL/6 strain used in this study has a gestation period of 18.5 days²¹. Although we had hypothesized that the entire shared casein locus would be under SE control, distinct gene-specific differences were observed. Odam mRNA levels were reduced by more than 99%, Csn3 by 98% and Csn1s2b by 93% (Fig. 2b and Supplementary Data 2), which coincided with a decline of H3K27ac, H3K4me3, and Pol II coverage at these genes, but not in Wap, another milk protein gene (Fig. 2c and Supplementary Fig. 3). Csn1s1, Csn2 and Csn1s2a expression was preserved at approximately 50%, 70% and 60% of wild-type (WT) levels, respectively. Since the sequencing depth of ChIP-seq libraries can influence the peak heights of TF binding and of histone modifications, we also present ChIP-seq images of the STAT5 target gene Cish located on chromosome 9 and that are therefore not influenced by the SE located on chromosome 5 (Fig. 2c, right panel) (see Material and Methods for details). In contrast, Csn2 and Csn1s2a mRNA levels were reduced at a lower, yet statistically significant, level. This experiment demonstrates that the SE dictates the expression of three genes but has a limited impact on the other three mammary genes in the shared locus, whose regulation might be controlled by gene-specific enhancers.

**Fig. 2: Differential activation of selected *Casein* genes by the super-enhancer during pregnancy.**

To understand whether the SE is required for the establishment of gene-specific regulatory elements, such as enhancers, we conducted ChIP-seq experiments in tissue lacking the SE. STAT5, GR and NFIB binding at the two candidate Csn3 enhancers remained intact in the absence of the SE (Fig. 2c), suggesting the inability of local enhancers to functionally compensate for the SE. In contrast, STAT5 binding at the Csn1s2b proximal enhancer is lost in ∆SE mammary tissue suggesting that the SE activates this secondary gene-specific enhancer.

A defining feature of milk protein genes is their exceptional response to pregnancy hormones, first and foremost prolactin. While the SE differentially affects Casein genes after a full pregnancy, it might have a more extended function in early pregnancy, prior to the prolactin surges that activate milk protein genes. Expression data obtained from mammary tissue on day six of pregnancy (p6) indicate an expanded SE function extending throughout the entire shared locus. In the absence of the SE (∆SE), expression of Odam and all five Casein genes was reduced by more than 96% (Fig. 2d, e and Supplementary Data 3), suggesting a possible role in the establishment of activating chromatin marks and possibly TF binding in early pregnancy. While H3K27ac marks were present at some Casein genes at p6, they were greatly reduced in the absence of the SE (Fig. 2e).

Casein mRNA levels in mammary tissue from non-parous female mice are equivalent to those obtained in the absence of the SE at p6, suggesting that the SE is biologically inactive at early pregnancy. An earlier study demonstrated enrichment of H3K4me2²² over the region we now call SE-E1, and our own data demonstrate NFIB binding to SE-E1 in virgin tissue (Supplementary Fig. 5). However, we did not detect any H3K27ac marks, suggesting that the SE has no or only marginal activity at that stage.

The 10 kbp SE region harbors four areas bound by mammary TFs, and their capacity to function individually or in combination and contribute to the overall SE activity was not clear. To gain insight into the complexity of this SE, we introduced individual and combined deletions (Supplementary Fig. 4). Deletion of SE-E1 (∆SE-E1), SE-E2 (∆E2) or SE-E4 (∆E4) resulted in the loss of STAT5 binding and H3K27ac at their respective sites. Notably, the establishment of SE-E3 and SE-E4 is dependent on the presence of SE-E2. ChIP-seq data from the STAT5 target gene Cish were added to demonstrate the sequencing depth of the libraries (Supplementary Fig. 4A, right panel). While the loss of SE-E4 (∆E4) had no discernible consequence on any of the Casein genes, Csn1s1, Csn1s2b and Csn3 mRNA levels were reduced between 40–70% in both ∆E1 and ∆E2 tissues. Combined deletion of both SE-E1 and SE-E2 (∆E1/2) silenced the entire SE and mimicked the ∆SE mutation suggesting redundancy between SE-E1 and SE-E2. Of note, ∆E1/2 mice exhibited lactation failure.

Having identified the physiological significance of the SE in mammary tissue, we addressed its regulatory significance in salivary glands. While the loss of SE activity led to a complete silencing of Odam expression, Fdcsp and Csn3 mRNA levels declined by 99% and 88%, respectively (Fig. 3 and Supplementary Data 4), demonstrating its dual specificity.

**Fig. 3: Salivary-specific activation of selected genes in the *Casein* locus by the super-enhancer.**

The Fdcsp gene fails to be activated in mammary tissue

The follicular dendritic cell secreted protein (Fdcsp) gene is expressed in salivary glands and immune cells (https://www.ncbi.nlm.nih.gov/gene/260436) but not in mammary glands. The absence of Fdcsp expression in mammary tissue might be the result of Odam blocking the SE from efficiently reaching and activating the Fdcsp promoter. To test this hypothesis, we deleted the Odam gene, thus transporting the SE within a few kbp to the Fdcsp gene (Fig. 4 and Supplementary Data 5). Despite being in the physical orbit of the SE, the Fdcsp promoter remained silent in lactating mammary tissue (Fig. 4a, c), and expression in salivary tissue was unaltered (Fig. 4b, d). These findings suggest that the promoter is unresponsive to the SE. In contrast, Csn3 mRNA levels increased approximately 2-fold in mammary tissue, suggesting a distance-dependency of SE activity. No expression changes were observed in salivary glands.

**Fig. 4: Distance-dependent function of the super-enhancer.**

To understand the mechanism behind Fdcsp being nonreceptive to the powerful casein SE, we conducted global DNA methylation analyses in mammary tissue at day one of lactation and in liver tissue (Supplementary Fig. 6). The methylation status of the promoter regions (400 bp surrounding the transcriptional start site, TSS) correlated with their activities in mammary tissue. While the highly active Casein genes displayed a low methylation coverage, the Fdcsp promoter was fully methylated, reflecting a repressive state. The methylation status of specific mammary cell populations from Balb/C mice has been investigated²³, and we used these data to confirm the methylation status of the casein locus (Supplementary Fig. 6). In contrast to mammary tissue, the Casein gene promoter regions were highly methylated in liver tissue.

Local enhancers control Csn3 expression

Removal of the SE ablates Csn3 expression without impacting TF binding at the two Csn3 enhancers, thus questioning their physiological roles. We addressed this issue and introduced deletions within the distal (Csn3-E1) and proximal (Csn3-E2) Csn3 candidate enhancers (Fig. 5a). Strong STAT5 binding coinciding with two GAS motifs and an NFIB site occurred at Csn3-E2, and weaker binding was observed at Csn3-E1. GR binding was stronger at Csn3-E1 compared to Csn3-E2, suggesting distinct molecular structures and possibly functions of the two enhancers. While deletion of the distal candidate enhancer (∆E1) resulted in the loss of TF binding and H3K27ac coverage at this site, Csn3 gene expression remained at 55% (Fig. 5b, c). Ablation of the Csn3-E2 enhancer resulted in lactation failure similar to CSN3 deficient mice²⁴, and gene expression was measured in mammary tissue at day 18 of pregnancy (p18), just prior to delivery or within 12 h after delivery (post-partum <12 h, pp <12 h), as in the Csn-SE mutant mice. Deletion of the two GAS motifs in Csn3-E2 (∆E2-S) reduced Csn3 mRNA levels by 98%, revealing the reason for lactation failure. While STAT5 binding was completely abolished, residual binding of NFIB was detected in mammary tissue within 12 h after delivery (pp <12 h). Deletion of the two GAS motifs and the NFIB site (∆E2-S/N) reduced Csn3 mRNA levels even further, coinciding with a complete absence of STAT5, GR and NFIB binding, loss of H3K27ac marks and Pol II loading (Fig. 5b, c). ChIP-seq data from the Cish locus are shown as a reference. Of note, the absence of Csn3-E2 did not adversely affect other Casein genes (Supplementary Fig. 7 and Supplementary Data 6). Interactions between the Csn3 enhancers and the SE had been confirmed by 3C, thus supporting their crosstalk²⁵.

**Fig. 5: Super-enhancer-dependent local enhancers activate *Csn3* expression.**

Local enhancer modifies Csn1s1 expression

Expression of Csn1s1, positioned at the 5′ border of the casein locus, is only marginally influenced by the SE located 200 kbp 3′ of the gene (Fig. 1), suggesting the presence of independent regulatory elements. H3K27ac marks pointed to candidate regulatory elements at −11.5 kb (Csn1s1-E1) and −3.5 kb (Csn1s1-E2) (Fig. 6a). Strong STAT5, GR and NFIB occupancy was detected at site Csn1s1-E2 but less so at Csn1s1-E1, which also had reduced H3K27ac coverage. STAT5 binding was also observed at the Csn1s1 promoter and coincided with a GAS motif at position −100 bp. While deletion of the GR motif in Csn1s1-E1 (ΔE1) resulted in the loss of GR and STAT5 binding at this site and reduced STAT5 binding at the promoter site, Csn1s1 expression at day one of lactation was not impaired (Fig. 6b, c). In contrast, deletion of the NFIB sites in Csn1s1-E2 (ΔE2) resulted in a 65% reduction of Csn1s1 expression and coincided with the loss of TF binding and H3K27ac at both enhancers. Despite the reduction of Csn1s1, lactation in these mice was overtly normal. These findings demonstrate that individually the two Csn1s1 enhancers have a very limited biological activity compared to the Csn3 enhancer described in this study. It is possible that the two enhancers could be partially redundant and work in a superadditive fashion so that only when both enhancers are removed, the effect on Csn1s1 expression would be revealed (see Discussion). Alternatively, the two candidate enhancers could synergize with the STAT5-based promoter element and control Csn1s1 expression during pregnancy and lactation.

**Fig. 6: *Csn1s1* gene expression is modulated by a proximal enhancer.**

Limited biological activity of the Csn2 local enhancers

Although major investments have been made in understanding enhancer function, knowledge of promoter elements and their possible interactions with enhancers is limited. To gain insight into the possible contribution of promoter elements in the regulation of mammary genes, we investigated the regulation of the Csn2 and Csn1s2a genes that are transcribed in opposite directions and whose promoters are separated by 75 kbp (Fig. 7a). Both genes are activated approximately 10,000-fold between virgin and lactating mammary tissue and their combined mRNAs account for approximately 20% of total mRNA during lactation (Fig. 1c and Supplementary Data 1). The mammary transcription factors STAT5, GR and NFIB bind at five candidate enhancers (Csn2-E1 to Csn2-E5) that also coincide with activating histone marks (Fig. 7a). Canonical STAT5 recognition sites (GAS motifs) are present in only four enhancers, suggesting that binding to Csn2-E3 is mediated by another TF, possibly GR or NFIB (Supplementary Table 2). Of note, STAT5 binds to two distinct regions in Csn2-E1, Csn2-E2, and Csn2-E4, suggesting additional internal complexity. STAT5 and GR binding was also associated with the Csn2 promoter, coinciding with a canonical (TTCn₃GAA) and a non-canonical (TTCn₄GAA) GAS motif, as well as with the Csn1s2a promoter. No bona fide GR binding site was detected, suggesting GR binding mediated by STAT5. The H3K27ac patterns differed between enhancer and promoter elements. While TF binding at enhancers was within an H3K27ac ‘gap’, binding at promoters was outside a classical histone-depleted gap. This points to the possibility that TF binding at enhancers and promoters and their interactions with H3K27ac marks is distinctly different and might serve different purposes. It is further possible that promoter regions are more flexible and present higher heterogeneity in TF binding and different chromatin organization due to the gene transcription or 3D organization of these genes can be envisioned.

**Fig. 7: Activity of putative *Csn2* enhancers.**

First, we investigated the physiological role of the three candidate enhancers located upstream of Csn2, Csn2-E1 at −6 kbp, Csn2-E2 at −25 kbp and Csn2-E3 at −35 kbp (Fig. 7a). All three candidate enhancers are bound by STAT5, GR and MED1. We used CRISPR/Cas9 genome engineering to delete these enhancers and monitor their functional significance in mammary tissue on day one of lactation (L1). Neither the deletion of Csn2-E1 nor the combined deletion of Csn2-E2 and Csn2-E3 resulted in a significant reduction of Csn2 mRNA levels (Fig. 7b). The combined deletion of all three candidate enhancers resulted in an approximately 50% decrease in Csn2 mRNA levels. The deletion of enhancers was validated by DNA sequencing and ChIP-seq analyses for activating histone marks, STAT5 and GR (Fig. 7c). Expression of Csn1s2a was not significantly impaired in these mutants, and TF binding to candidate elements associated with Csn1s2a was not affected. The Cish locus was used as a reference for the sequencing depth.

Synergy between promoter and enhancer elements

Based on our results, the three distal enhancers contribute only modestly to the activation of the Csn2 gene, opening the possibility of additional regulatory elements. We therefore focused on the two STAT5 binding sites within the Csn2 promoter region (Fig. 8a) and introduced disabling point mutations into the palindromic sequences of the canonical and non-canonical GAS motifs located within 150 bp of the transcriptional start site. Disabling both GAS sites (ΔP) through deaminase base editing resulted in an approximately 80% reduction of Csn2 mRNA at L1 (Fig. 8b and Supplementary Data 7). These findings begged the question of potential synergy between the cytokine-responsive promoter elements and the three distal enhancers. We addressed this in mice lacking the three enhancers and carrying point mutations in the two promoter-based GAS elements. The combined mutations resulted in an almost complete silencing of the Csn2 gene, with expression levels reduced by four orders of magnitude (99.99%) (Fig. 8b). Loss of these elements coincided with the complete absence of Pol II coverage and H3K27ac marks (Fig. 8c). The presence of the three enhancers and the promoter-based STAT5 sites is required for the full activation of Csn2 expression during lactation (Supplementary Fig. 8a and Supplementary Data 8), but their loss does not affect basal expression in mammary tissue of non-parous (virgin) mice (Supplementary Fig. 8b and Supplementary Data 9). This is further evidence that these regulatory elements elicit hormone responsiveness.

**Fig. 8: Synergy between promoter-based cytokine-response elements and distal enhancers.**

There is consensus that all seven STATs, except for STAT6, activate gene expression through canonical GAS motifs. A single STAT5 peak covers the closely spaced canonical and non-canonical GAS motifs in the Csn2 promoter (Supplementary Fig. 9a), making it difficult to discern if either site or both sites are bound by STAT5 and convey transcriptional activity. To distinguish the relative significance of the two GAS motifs, we introduced mutations individually into the two sites, both in the presence and absence of the three distal enhancers (Supplementary Fig. 9a). In the presence of the distal enhancers, mutation of the canonical site (A) did not impact Csn2 expression, and a 50% reduction was observed upon the simultaneous deletion of the three enhancers (Csn2ΔP-E1/2/3-A) (Supplementary Fig. 9b). Introduction of a disabling mutation into the non-canonical site (B) resulted in a ~80% reduction of Csn2 mRNA levels and the additional absence of the distal enhancers (Csn2-ΔP-E1/2/3-B) resulted in the complete loss of Csn2 expression and levels were four orders lower than in mice with intact regulatory elements. STAT5 binding was still observed upon loss of the canonical GAS motif and completely absent in the non-canonical mutant (Supplementary Fig. 9c). These results provide evidence that the non-canonical GAS motif is a key element in the Csn2 promoter and STAT5 preferentially binds at the non-canonical site.

Discussion

Although mouse genetics has been used to address the functionality of enhancers in single gene loci^{16,26,27,28,29}, key questions pertaining to the contribution of enhancers and super-enhancers (SE) to gene regulation remain to be answered. Specifically, understanding the regulation of complex multigene loci harboring genes expressed in one or more distinct cell types is lacking. Our study provides insight into regulatory mechanisms operative in salivary and mammary gland tissues that share morphological and molecular features during embryogenesis^30,31,32. Specifically, we identified a SE active in mammary and salivary tissues, local gene-specific enhancers and a cytokine-responsive promoter element that activate transcription during pregnancy. The shared locus, with its eight genes linked to lactation, saliva and immune response⁶, is an evolutionary playground that fostered regulatory innovation with 20 enhancer and promoter elements. Odam and its associated SE likely constitute the ancestral unit of the shared locus and regulatory activity expanded from salivary tissue to mammary tissue. However, as this locus expanded, the five newly formed Casein genes acquired their own regulatory elements and three gained partial independence of the SE, at least during lactation (Fig. 9a, b). Although Csn1s2b²⁵ and Csn3 acquired local enhancers essential for their expression in lactating mammary tissue, they still retained their dependence on the SE.

**Fig. 9: Model for the regulation of the shared casein mammary-salivary locus by a dual specific super-enhancer, gene-specific local enhancers and promoters.**

SEs are structurally defined compound regulatory elements which control lineage-specific genes³³. There is ongoing research investigating if individual enhancer modules have unique properties and their contributions to SE activity. A few studies have tackled the complexity of SEs in mice and gained insight into the biological significance of individual components. The well-studied α-globin SE consists of five individual enhancers that act independently and in an additive fashion²⁸. In a recent study, the Higgs group rebuilt the α-globin SE from individual modules and identified elements that were non-functional on their own but facilitated the activities of others³⁴. Like the α-globin locus, the casein SE and the Wap SE¹⁶ are composed of modules with distinct transcriptional capacities. Studies from the Higgs group have also demonstrated, unexpectedly, that the α-globin SE activity is orientation-dependent³⁵ and inversion preferentially activates genes 5′ to the SE. Similarly, the casein SE investigated in our study differentially activates genes located at either end. Based on current knowledge, the regulatory logic of SE is based on mixtures of essential and dispensable modules that can be active in one or more cell types^{16,34,35,36,37,38,39,40,41}. Recent studies have demonstrated that enhancers and promoters can engage CTCF-associated sites and induce distant genes^42,43, a concept that might also be applicable in the casein locus, which harbors four sites occupied by CTCF (Fig. 1a), two bordering the locus and two within the SE¹⁷. We have deleted these sites from the mouse genome and observed only a moderate effect of the SE-bound sites¹⁷.

There is limited knowledge about regulatory mechanisms operative in salivary tissue. Genome-wide histone modification studies in mouse submandibular glands have pointed to putative regulatory regions⁴⁴, but no defined salivary-specific enhancers have been described. ATAC-seq data from salivary tissue have been published⁴⁵, but they did not yield additional information. Key TFs controlling mammary function, such as STAT5^8,9 and ELF5⁷, are dispensable in the salivary gland⁴⁶. The presence of the progesterone receptor is also required for normal mammary gland development⁴⁷, but its role in salivary physiology remains to be determined⁴⁸. The molecular backbone of the casein SE, as defined by H3K27ac marks, extends over 10 kbp, and no salivary-specific TF sites have been identified. While two enhancer modules within the casein SE are key activators in mammary tissue, it is not clear which of the four enhancer elements control salivary gene expression. We also need to consider the possibility that the casein SE activity is facilitated by receptive target gene promoters as indicated by the neighboring Fdcsp gene, which is active in salivary but not in mammary tissue. This specificity is not influenced by the distance of the SE or the presence of the intervening Odam gene, suggesting the presence of unique promoter elements that permit enhancer sensing. Alternatively, differential promoter accessibility⁴⁹ in salivary and mammary tissue could account for cell specificity.

The five Casein genes, namely Csn1s1, Csn2, Csn1s2a, Csn1s2b and Csn3, share common regulatory mechanisms. These include lactogenic hormone response elements, such as GAS motifs bound by the prolactin-activated transcription factor STAT5, glucocorticoid response elements (GREs) and sites recognized by NFIB, another key TF to the 10,000-fold stimulation of casein gene expression during pregnancy and lactation^8,9,16,50. While there are shared regulatory mechanisms for the five Casein genes, the presence of distinct differences provides flexibility and specificity in the regulation of each gene’s expression and ultimately contributes to the complexity of milk production. Most notably, the presence of the super-enhancer within the regulatory landscape of Csn1s2b and Csn3 suggests that these genes rely on a highly cooperative and robust network, which includes local enhancers, for their expression. On the other hand, Csn1s1, Csn2 and Csn1s2a are largely regulated by a different set of local gene-specific enhancers with minimal dependence on the super-enhancer.

The combination of local gene-specific enhancers and promoters, all bound by STAT5 and other mammary-centric TFs, is likely the regulatory key to casein gene expression. For example, Csn2 expression is under the combined control of three distal enhancers and a single promoter element, with their individual ablation having a very limited biological impact. The neighboring Csn1s1 gene also features two STAT5-bound enhancer elements, and while their individual deletion had a limited biological impact, it can be hypothesized that, like in the Csn2 gene, the enhancer and promoter elements synergize to ensure maximum transcriptional control.

Although STAT5 is the key switch activating Casein genes during pregnancy and lactation, the presence of additional control elements within the promoter and enhancer regions might modulate their activation. Both the Csn1s2b and Csn3 genes display functional and structural features not seen in other Casein genes. Csn1s2b contains a unique intronic enhancer bound by STAT5 and other mammary TFs that contribute to the priming of the single distal enhancer and the promoter prior to parturition, resulting in a distinctive temporal expression distinct from that of the other Casein genes²⁵. Lastly, while four Casein genes (Csn1s1, Csn2, Csn1s2a and Csn1s2b) are characterized by STAT5-bound promoter regions, the Csn3 promoter does not feature a GAS motif, nor does it bind STAT5 or other mammary-centric TFs. This is probably not surprising since Csn3 is not a classical calcium-sensitive casein gene and has a distinct evolutionary history.

At least 13 STAT5-based enhancers, including the four component super-enhancer, and five distinct promoters control the five Casein genes, a testimony to the regulatory complexity of different building blocks, thus ensuring the maximum production of milk needed for the survival of the young and the species. Although all enhancers are bound by STAT5 and an array of mammary TFs, only a subset displays a classical STAT5 binding motif, suggesting that a seemingly equivalent TF occupancy can be achieved through a combination of TFs, including GR and NFIB.

While the Wap gene relies on a tripartite SE for its activation, the Csn2 gene is a unique and distinct example of synergy between distal enhancer elements and promoter elements, both of which respond to cytokines through the mammary TF STAT5. A new class of regulatory elements, named Epromoters^51,52, function as both promoter and enhancer. Epromoters are hubs for TF machinery and are typically associated with stress-response genes, such as those activated by interferons (IFN) through the TFs STAT1/2. A key feature of Epromoters is their capacity to activate neighboring genes independent of the presence of additional enhancers⁵². Epromoters are frequently found in loci harboring co-regulated genes, such as the Oas locus, that otherwise do not contain enhancers⁵².

Despite similarities between Epromoters and the Csn2 promoter, there are distinctive differences (Fig. 9c). Although both Epromoters and the five Casein gene promoters, including Csn2 investigated in this study, harbor cytokine-response elements bound by TF STATs, their positioning appears to be different, with those in the Casein genes being within 100 bp of the TSS and therefore likely integral part of the promoter. A distinct difference between the two regulatory elements is the capacity of Epromoters to activate neighboring genes in the absence of additional enhancers, while the Csn2 promoter element is fully dependent on the presence of distal enhancers. Moreover, unlike Epromoters, which activate neighboring genes at great distance, the Csn2 promoter activity is confined to its own gene, and neighboring genes are regulated independently. The activity of Epromoters has been validated in cell lines using CRISPR-Cas9-induced deletions spanning several hundred base pairs⁵², which leaves the possibility that additional elements contributed to their activity. In contrast, our study introduced single base pair mutations into GAS motifs, demonstrating the specific and defining requirement of STAT5 in the activation of Csn2. Validating the specificity of TF binding through the introduction of point mutations is desirable since it has been shown that phantom TF binding occurs at active promoters⁵³.

Regulation of the Csn2 gene has been investigated in HC11 tissue culture cells by Kabotyanski and Rosen^54,55. Prolactin treatment resulted in the recruitment of STAT5 to the promoter region and a candidate enhancer at −6 kb⁵⁴, and looping between these elements was observed⁵⁵. Efficient RNA polymerase II recruitment required the simultaneous administration of prolactin and glucocorticoids⁵⁴, which is supported by our study demonstrating STAT5 and GR binding to these elements. While the −6.4 kb enhancer identified by the Rosen group coincides with the Csn2-E1 enhancer in our study, Csn2-E2 and Csn2-E3 have not been reported in HC11 cells, pointing to clear differences between in vivo mouse studies and tissue culture cells. A study using cell culture and gene transfection chloramphenicol acetyl transferase (CAT) assays as readout described an element in the upstream region of the bovine beta Casein gene (now Csn2) that responds to the extracellular matrix and prolactin^56,57. Another investigation also using gene reporter assay described a similar element in the human beta casein gene⁵⁸. While these elements activate reporter genes in cell lines, deletion of the corresponding elements in the mouse casein locus has limited significance, as shown in our in vivo study, emphasizing clear differences between cell culture studies and in vivo mouse genetic studies.

The Csn2 promoter appears to be a hub for transcription factors, such as the progesterone receptor⁵⁹ and C/EBPs⁶⁰. Progesterone suppresses Csn2 expression⁶¹, and the receptor is essential for mammary gland development⁴⁸ suggesting a wider role in the mammary genome. As the complexity of pregnancy-induced genetic programs and the magnitude of gene activation cannot be replicated in tissue culture cells, experimental mouse genetics provides an opportunity to gain insight into physiological processes. Having said this, global or cell-specific deletions of regulatory proteins, such as transcription factors, frequently elicit global effects, which prevents a clear understanding of their roles in individual genes. The inactivation of specific regulatory elements through genome engineering provides more definitive answers.

Gene families regulated by IFNs, and milk protein genes share features that likely co-evolved during evolution. Repurposing structural genes, promoters and enhancers is a driving force in evolution, fostering innovation and the establishment of new genes and regulatory concepts^62,63. Both, IFN-regulated genes and milk protein genes, are rapidly induced by cytokines that specifically utilize the JAK/STAT regulatory machinery permitting a rapid induction of genes. However, while transcriptional activation elicited by IFNs largely relies on JAK1 and STAT1/2⁶⁴, mammary genes are predominantly activated by the pregnancy hormone prolactin, signaling through JAK2 and STAT5^8,9,11. The elaborate enhancer structure in the casein locus permits an exceptional expression of five genes accounting for 80% of milk proteins, an essential requirement for the sustained success of mammals. We propose that the compactness of enhancers and their high-density occupation with TFs and co-activators provides an optimal regulatory environment. In summary, the comprehensive dissection of promoter and enhancer elements within their native loci provided physiological insight into a complex hormone-controlled regulatory network operative in mammary and salivary tissues.

Methods

Mice

All animals were housed in an environmentally controlled room (22–24 °C, with 50 ± 5% humidity and 12 h/12 h light–dark cycle) and handled according to the Guide for the Care and Use of Laboratory Animals (8th edition) and all animal experiments were approved by the Animal Care and Use Committee (ACUC) of National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK, MD) and performed under the NIDDK animal protocol K089-LGP-17. CRISPR/Cas9 targeted mice were generated using C57BL/6N mice (Charles River) by the transgenic core of the National Heart, Lung, and Blood Institute (NHLBI). Single-guide RNAs (sgRNA) were obtained from either OriGene (Rockville, MD) or Thermo Fisher Scientific (Supplementary Table 3). Target-specific sgRNAs and in vitro transcribed Cas9 and Base editor mRNA were co-microinjected into the cytoplasm of fertilized eggs for founder mouse production^65,66,67,68. The ∆E1/2 mutant mouse was generated by injecting a sgRNA for E2 into zygotes collected from ΔE1 mutant mice. The Csn2-∆E1/2/3 and Csn2-∆P-E1/2/3 mutant mice were generated by injecting a sgRNA for E1 into zygotes collected from Csn2-ΔE2/3 mutant mice and for P into zygotes collected from Csn2-ΔE1/2/3 mutant mice, respectively. All mice were genotyped by PCR amplification and Sanger sequencing (Macrogen and Quintara Biosciences) with genomic DNA from mouse tails (Supplementary Table 4).

Two-months old mice were used in the experiments. Mammary tissue was collected from non-parous (virgin) mice, from mice at day 6 of pregnancy and at days 1 (L1) and 10 (L10) of lactation and stored at −80 °C. While most mutations did not affect the ability of dams to lactate and nurse their pups, deletion of the super-enhancer (ΔSE and ∆E1/2) and deletion of the Csn3 E2 enhancer (Csn3-ΔE2) resulted in lactation failure, which is most likely due to the absence of CSN3. The C57BL/6 mouse strain used in this study has a gestation period of 18.5 days²¹, and mammary tissues from mutant mice unable to lactate (ΔSE, SE-∆E1/2 and Csn3-ΔE2) were collected at day 18 of pregnancy (p18) or within 12 h after delivery (post-partum <12 h, pp <12 h) (Supplementary Data 10). Salivary tissues (submandibular glands) were collected from 2-month-old male mice and stored at −80 °C.

Chromatin immunoprecipitation sequencing (ChIP-seq) and data analysis

The frozen-stored tissues were ground into powder in liquid nitrogen. Chromatin was fixed with formaldehyde (1% final concentration) for 15 min at room temperature and then quenched with glycine (0.125 M final concentration). Nuclei were isolated with Farnham Lysis Buffer (5 mM PIPES pH 8.0, 85 mM KCl, 0.5% NP-40, PMSF and proteinase inhibitor cocktails). The chromatin was fragmented to 200–500 bp using sonicator 3000 (30 cycles; 20 s pulse/20 s rest, Misonix Sonicators) and further lysed in RIPA buffer. One milligram of chromatin was immunoprecipitated with Dynabeads Protein A (Novex) coated with antibodies. The following antibodies were used for ChIP-seq: STAT5A (Santa Cruz Biotechnology, sc-271542), GR (Thermo Fisher Scientific, PA1-511A), NFIB (Sigma-Aldrich, HPA003956), MED1 (Bethyl Laboratory, A300−793A), H3K27ac (Abcam, ab4729), RNA polymerase II (Abcam, ab5408), H3K4me1 (Active Motif, 39297) and H3K4me3 (Millipore, 07-473). Then, 5–10 ug of antibodies were added to 1 mg of total proteins (1 ml solution). After serial bead washes, ChIP DNA was reverse crosslinked at 65 °C overnight in the presence of 1% SDS and 1 mg/ml of Proteinase K (Invitrogen), and DNA was purified with QIAquick PCR Purification Kit (Qiagen). The DNA fragments were blunt-ended and ligated to the Illumina index using the NEBNext Ultra II DNA Library Prep kit for Illumina (New England BioLabs). Libraries for next-generation sequencing were prepared and sequenced with a HiSeq 3000 instrument (Illumina).

Quality filtering and alignment of the raw reads were done using FastQC (version 0.11.0), Trimmomatic⁶⁹ (version 0.36) and Bowtie⁷⁰ (version 1.2.2), with the parameter ‘-m 1’ to keep only uniquely mapped reads, using the reference genome mm10. Samtools⁷¹ was to convert BAM files to SAM format. Picard tools (Broad Institute. Picard, http://broadinstitute.github.io/picard/. 2016) was used to remove duplicates, and subsequently, Homer⁷² (version 4.9.1) and deepTools⁷³ (version 3.1.3) software were applied to generate bedGraph files and normalize coverage, separately. Integrative Genomics Viewer⁷⁴ (version 2.3.98) was used for visualization. Coverage plots were generated using Homer⁷² software with the bedGraph from deepTools as output. R and the packages dplyr (https://CRAN.R-project.org/package=dplyr) and ggplot2⁷⁵ were used for visualization. Each ChIP-seq experiment was conducted for two replicates, and the correlation between the replicates was computed by Spearman correlation using deepTools (Supplementary Data 10).

Although we would ideally want to study the absolute levels of binding, comparing the levels of ChIP-seq enrichments across different conditions is more difficult than one would have hoped for because particularly ChIP-seq is awfully noisy and dependent on too many experimental parameters^76,77,78. The quantification of ChIP-seq varies due to experimental variation between samples introduced by different efficiencies in nuclear extraction, DNA shearing and immunoprecipitation. However, the methods for normalization to minimize sample-to-sample variability between conditions are based on the assumption that experimental variables remain constant between datasets and assume comparable genomic binding of the protein between conditions. To overcome this issue, we used deepTools and its bamCoverage function to calculate the coverage as the number of reads per bin and normalize data by Reads Per Kilobase per Million mapped reads (RPKM) and 1x depth (reads per genome coverage, RPGC). Additionally, we normalized each peak against sequencing depth again. However, these analytical normalization methods didn’t normalize experimental variation between conditions⁷⁹, even though we generated the replicates. To overcome this issue, we presented the entire casein locus and the Cish locus as a ChIP-seq control to show the ChIP-seq pattern.

Total RNA sequencing (Total RNA-seq) and data analysis

Total RNA was extracted from frozen mammary and salivary tissue from wild-type and mutant mice and purified with RNeasy Plus Mini Kit (Qiagen, 74134). Ribosomal RNA was removed from 1 μg of total RNAs, and cDNA was synthesized using SuperScript III (Invitrogen). Libraries for sequencing were prepared according to the manufacturer’s instructions with TruSeq Stranded Total RNA Library Prep Kit with Ribo-Zero Gold (Illumina, RS-122-2301), and paired-end sequencing was done with a HiSeq 3000 instrument (Illumina).

Total RNA-seq read quality control was done using FastQC (version 0.11.0), Trimmomatic⁶⁹ (version 0.36) and STAR RNA-seq⁸⁰ (version STAR 2.5.4a) using paired-end mode was used to align the reads (mm10). Samtools⁷¹ was to convert BAM files to SAM format and HTSeq⁸¹ was to retrieve the raw counts, and subsequently, R (https://www.R-project.org/), Bioconductor⁸² and DESeq2⁷⁵ were used. Additionally, the RUVSeq⁸³ package was applied to remove confounding factors. The data were pre-filtered keeping only those genes, which have at least ten reads in total. Genes were categorized as significantly differentially expressed with an adjusted p-value below 0.05 and a fold change > 2 for upregulated genes and a fold change of <−2 for downregulated ones. The visualization was done using dplyr (https://CRAN.R-project.org/package=dplyr) and ggplot2⁸⁴.

Bisulfite-seq and data analysis

In this, 200 ng of purified genomic DNA was used to create bisulfite sequencing libraries via the large DNA insert approach using the NEBNext® Enzymatic Methyl-seq Kit (New England Biolabs). In short, purified genomic DNA was fragmented, adenylated, and ligated to EM-seq adapters. Bisulfite conversion was performed according to the manufacturer’s instructions. Internal unmethylated lambda DNA and CpG methylated pUC19 DNA controls were included with each library. Libraries were sequenced with a NovaSeq 6000 instrument (Illumina) to generate a minimum of 300 million paired-end 151 base reads per library.

Raw fastq files were quality trimmed using TrimGalore (version 0.6.7). Reads were aligned to the mouse genome (mm10) using Bismark⁸⁵ (version 0.23.0). Mapped BAM files were sorted by query name, deduplicated using Picard (version 2.27.3) and finally sorted coordinates using Samtools⁸⁶ (version 1.17). Integrative Genomics Viewer⁷⁴ (version 2.5.3) was used for visualization.

RNA isolation and quantitative real-time PCR (qRT-PCR)

Total RNA was extracted from frozen mammary tissue of wild-type and mutant mice using a homogenizer and the PureLink RNA Mini kit according to the manufacturer’s instructions (Thermo Fisher Scientific). Total RNA (1 μg) was reverse transcribed for 50 min at 50 °C using 50 μM oligo dT and 2 μl of SuperScript III (Thermo Fisher Scientific) in a 20 μl reaction. Quantitative real-time PCR (qRT-PCR) was performed using TaqMan probes (Csn1s1, Mm01160593_m1; Csn2, Mm04207885_m1; Csn1s2a, Mm00839343_m1; Csn1s2b, Mm00839674_m1; Odam, Mm02581573_m1; Csn3, Mm02581554_m1; mouse Gapdh, Mm99999915_g1, Thermo Fisher Scientific) on the CFX384 Real-Time PCR Detection System (Bio-Rad) according to the manufacturer’s instructions. PCR conditions were 95 °C for 30 s, 95 °C for 15 s, and 60 °C for 30 s for 40 cycles. All reactions were done in triplicate and normalized to the housekeeping gene Gapdh. Relative differences in PCR results were calculated using the comparative cycle threshold (C_T) method and normalized to Gapdh levels.

Identification of regulatory elements bound by STAT5 in lactation

MACS2⁸⁷ peak finding algorithm was used to identify regions of ChIP-seq enrichment over the background to get regulatory elements at L1 and L10. Peak calling was done on both STAT5A replicates and broad peak calling on H3K27ac. Only those peaks were used, which were identified in both replicates and with H3K27ac coverage underneath.

Identification of complex mammary loci

Mammary preferential genes were identified using RNA-seq data from pregnancy day six (p6), lactation day one (L1) and ten (L10). Those genes were considered, which were induced more than two-fold with an adjusted p-value below 0.05 between p6 and L1 or p6 and L10. The next step comprised the stitching of neighboring genes, by only considering protein-coding genes. Those stitched loci were subsequently compared to the contact domains (Hi-C data), and only those loci passed the validation that were not overlapping with any border of the contact domains. If a locus overlapped, they were treated the following way: (1) if the locus contains only two genes, it was discarded, as it will not pass the prerequisite that a complex locus comprises at least two genes; (2) all other loci were split up at the border, and only those were kept that comprised more than two genes; (3) the loci were shrunk to the size of the remaining genes. As possible regulatory elements (in our analysis STAT5A) are also part of complex loci, we expanded the borders of each locus to comprise STAT5A binding sites, if they were located within the adjacent intergenic region. Those new loci were finally checked again to not overlap with contact domain boundaries, otherwise, they were shrunk down to the last element not overlapping with the contact domain; (4) the final list of complex loci comprises loci with at least three genes.

The analysis was done using bedtools⁸⁸, bedops, R (https://www.R-project.org/) and Bioconductor⁸², as well as the R packages dyplr (https://CRAN.R-project.org/package=dplyr) and ggplot2⁸⁴.

Statistics and reproducibility

All samples that were used for qRT-PCR and RNA-seq were randomly selected, and blinding was not applied. No data were excluded from the analyses. Differential expression gene (DEG) identification of RNA-seq data used Bioconductor package DESeq2 in R. P-values were calculated using a paired, two-side Wilcoxon test and adjusted p-value (pAdj) corrected using the Benjamini–Hochberg method. The cut-off value for the false discovery rate was pAdj > 0.05. For comparison of samples, data were presented as the standard deviation in each group and were evaluated with a t-test and 2-way ANOVA multiple comparisons using PRISM GraphPad. Statistical significance was obtained by comparing the measures from the wild-type or control group, and each mutant group. A value of *P < 0.05, **P < 0.001, ***P < 0.0001, ****P < 0.00001 was considered statistically significant; ns, not significant.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Data availability

The ChIP-seq, RNA-seq, and Bisulfite-seq data generated in this study have been deposited in the Gene Expression Omnibus (GEO) database under accession code GSE231441. Other data were obtained from the GEO database under accession code GSE25105^22,89, GSE67386²³, GSE74826¹⁶, GSE115370²⁹, GSE127144²⁵, and GSE161620²⁵. The datasets used in the study are listed in Supplementary Data 10. The fastq files and the processed bedGraph files can be downloaded from GEO (https://www.ncbi.nlm.nih.gov/gds/?term=) and imported into the IGV browser (https://software.broadinstitute.org/software/igv/download) with a reference genome (mm10). The RNA-seq data generated in this study are provided in the Supplementary Data files. The remaining data generated in this study are provided in the Source Data file. Source data are provided with this paper.

References

Oftedal, O. T. The mammary gland and its origin during synapsid evolution. J. Mammary Gland Biol. Neoplasia 7, 225–252 (2002).
PubMed Google Scholar
Kawasaki, K., Buchanan, A. V. & Weiss, K. M. Biomineralization in humans: making the hard choices in life. Annu. Rev. Genet. 43, 119–142 (2009).
CAS PubMed Google Scholar
Kawasaki, K. The SCPP gene family and the complexity of hard tissues in vertebrates. Cells Tissues Organs 194, 108–112 (2011).
PubMed Google Scholar
Hennighausen, L. G. & Sippel, A. E. Characterization and cloning of the mRNAs specific for the lactating mouse mammary gland. Eur. J. Biochem. 125, 131–141 (1982).
CAS PubMed Google Scholar
Rijnkels, M., Elnitski, L., Miller, W. & Rosen, J. M. Multispecies comparative analysis of a mammalian-specific genomic domain encoding secretory proteins. Genomics 82, 417–432 (2003).
CAS PubMed Google Scholar
Braasch, I. et al. The spotted gar genome illuminates vertebrate evolution and facilitates human-teleost comparisons. Nat. Genet. 48, 427–437 (2016).
CAS PubMed PubMed Central Google Scholar
Zhou, J. et al. Elf5 is essential for early embryogenesis and mammary gland development during pregnancy and lactation. EMbo J. 24, 635–644 (2005).
CAS PubMed PubMed Central Google Scholar
Liu, X. et al. Stat5a is mandatory for adult mammary gland development and lactogenesis. Genes Dev. 11, 179–186 (1997).
CAS PubMed Google Scholar
Cui, Y. et al. Inactivation of Stat5 in mouse mammary epithelium during pregnancy reveals distinct functions in cell proliferation, survival, and differentiation. Mol. Cell. Biol. 24, 8037–8047 (2004).
CAS PubMed PubMed Central Google Scholar
Robinson, G. W. et al. Coregulation of genetic programs by the transcription factors NFIB and STAT5. Mol. Endocrinol. 28, 758–767 (2014).
PubMed PubMed Central Google Scholar
Shillingford, J. M. et al. Jak2 is an essential tyrosine kinase involved in pregnancy-mediated development of mammary secretory epithelium. Mol. Endocrinol. 16, 563–570 (2002).
CAS PubMed Google Scholar
Ormandy, C. J. et al. Null mutation of the prolactin receptor gene produces multiple reproductive defects in the mouse. Genes Dev. 11, 167–178 (1997).
CAS PubMed Google Scholar
Horseman, N. D. et al. Defective mammopoiesis, but normal hematopoiesis, in mice with a targeted disruption of the prolactin gene. EMBO J. 16, 6926–6935 (1997).
CAS PubMed PubMed Central Google Scholar
Rijnkels, M. et al. The epigenetic landscape of mammary gland development and functional differentiation. J. Mammary Gland Biol. Neoplasia 15, 85–100 (2010).
PubMed PubMed Central Google Scholar
Rijnkels, M., Kabotyanski, E., Shore, A. & Rosen, J. M. The chromatin landscape of the casein gene locus. Horm. Mol. Biol. Clin. Investig. 10, 201–205 (2012).
CAS PubMed PubMed Central Google Scholar
Shin, H. Y. et al. Hierarchy within the mammary STAT5-driven Wap super-enhancer. Nat. Genet. 48, 904–911 (2016).
CAS PubMed PubMed Central Google Scholar
Lee, H. K. et al. Functional assessment of CTCF sites at cytokine-sensing mammary enhancers using CRISPR/Cas9 gene editing in mice. Nucleic Acids Res. 45, 4606–4618 (2017).
CAS PubMed PubMed Central Google Scholar
Rada-Iglesias, A. Is H3K4me1 at enhancers correlative or causative? Nat. Genet. 50, 4–5 (2018).
CAS PubMed Google Scholar
Local, A. et al. Identification of H3K4me1-associated proteins at mammalian enhancers. Nat. Genet. 50, 73–82 (2018).
CAS PubMed Google Scholar
Li, M. et al. Mammary-derived signals activate programmed cell death during the first stage of mammary gland involution. Proc. Natl Acad. Sci. USA 94, 3425–3430 (1997).
CAS PubMed PubMed Central ADS Google Scholar
Murray, S. A. et al. Mouse gestation length is genetically determined. PLoS ONE 5, e12418 (2010).
PubMed PubMed Central ADS Google Scholar
Rijnkels, M. et al. Epigenetic modifications unlock the milk protein gene loci during mouse mammary gland development and differentiation. PLoS ONE 8, e53270 (2013).
CAS PubMed PubMed Central ADS Google Scholar
Dos Santos, C. O., Dolzhenko, E., Hodges, E., Smith, A. D. & Hannon, G. J. An epigenetic memory of pregnancy in the mouse mammary gland. Cell Rep. 11, 1102–1109 (2015).
PubMed PubMed Central Google Scholar
Shekar, P. C. et al. kappa-casein-deficient mice fail to lactate. Proc. Natl Acad. Sci. USA 103, 8000–8005 (2006).
CAS PubMed PubMed Central ADS Google Scholar
Lee, H. K., Willi, M., Kuhns, T., Liu, C. & Hennighausen, L. Redundant and non-redundant cytokine-activated enhancers control Csn1s2b expression in the lactating mouse mammary gland. Nat. Commun. 12, 2239 (2021).
CAS PubMed PubMed Central ADS Google Scholar
Oudelaar, A. M. & Higgs, D. R. The relationship between genome structure and function. Nat. Rev. Genet. 22, 154–168 (2021).
CAS PubMed Google Scholar
Furlong, E. E. M. & Levine, M. Developmental enhancers and chromosome topology. Science 361, 1341–1345 (2018).
CAS PubMed PubMed Central ADS Google Scholar
Hay, D. et al. Genetic dissection of the alpha-globin super-enhancer in vivo. Nat. Genet. 48, 895–903 (2016).
CAS PubMed PubMed Central Google Scholar
Lee, H. K., Willi, M., Shin, H. Y., Liu, C. & Hennighausen, L. Progressing super-enhancer landscape during mammary differentiation controls tissue-specific gene regulation. Nucleic Acids Res. 46, 10796–10809 (2018).
CAS PubMed PubMed Central Google Scholar
Santosh, A. B. & Jones, T. J. The epithelial-mesenchymal interactions: insights into physiological and pathological aspects of oral tissues. Oncol. Rev. 8, 239 (2014).
PubMed PubMed Central Google Scholar
Macias, H. & Hinck, L. Mammary gland development. Wiley Interdiscip. Rev. Dev. Biol. 1, 533–557 (2012).
CAS PubMed PubMed Central Google Scholar
Jimenez-Rojo, L., Granchi, Z., Graf, D. & Mitsiadis, T. A. Stem cell fate determination during development and regeneration of ectodermal organs. Front. Physiol. 3, 107 (2012).
PubMed PubMed Central Google Scholar
Whyte, W. A. et al. Master transcription factors and mediator establish super-enhancers at key cell identity genes. Cell 153, 307–319 (2013).
CAS PubMed PubMed Central Google Scholar
Blayney, J. et al. Super-enhancers require a combination of classical enhancers and novel facilitator elements to drive high levels of gene expression. Preprint at bioRxiv https://doi.org/10.1101/2022.06.20.496856 (2022).
Kassouf, M. T. et al. Multipartite super-enhancers function in an orientation-dependent manner. Preprint at bioRxiv https://doi.org/10.1101/2022.07.14.499999 (2022).
Hörnblad, A., Bastide, S., Langenfeld, K., Langa, F. & Spitz, F. Dissection of the Fgf8 regulatory landscape by in vivo CRISPR-editing reveals extensive intra- and inter-enhancer redundancy. Nat. Commun. 12, 439 (2021).
PubMed PubMed Central ADS Google Scholar
Thomas, H. F. et al. Temporal dissection of an enhancer cluster reveals distinct temporal and functional contributions of individual elements. Mol. Cell 81, 969–982.e913 (2021).
CAS PubMed Google Scholar
Grosveld, F., van Staalduinen, J. & Stadhouders, R. Transcriptional regulation by (super)enhancers: from discovery to mechanisms. Annu. Rev. Genomics Hum. Genet. 22, 127–146 (2021).
PubMed Google Scholar
Oudelaar, A. M., Beagrie, R. A., Kassouf, M. T. & Higgs, D. R. The mouse alpha-globin cluster: a paradigm for studying genome regulation and organization. Curr. Opin. Genet. Dev. 67, 18–24 (2021).
CAS PubMed PubMed Central Google Scholar
Liu, T. T. et al. Ablation of cDC2 development by triple mutations within the Zeb2 enhancer. Nature 607, 142–148 (2022).
CAS PubMed PubMed Central ADS Google Scholar
Honnell, V. et al. Identification of a modular super-enhancer in murine retinal development. Nat. Commun. 13, 253 (2022).
CAS PubMed PubMed Central ADS Google Scholar
Rinzema, N. J. et al. Building regulatory landscapes reveals that an enhancer can recruit cohesin to create contact domains, engage CTCF sites and activate distant genes. Nat. Struct. Mol. Biol. 29, 563–574 (2022).
CAS PubMed PubMed Central Google Scholar
Zuin, J. et al. Nonlinear control of transcription through enhancer-promoter interactions. Nature 604, 571–577 (2022).
CAS PubMed PubMed Central ADS Google Scholar
Gluck, C. et al. A global vista of the epigenomic state of the mouse submandibular gland. J. Dent. Res. 100, 1492–1500 (2021).
CAS PubMed PubMed Central Google Scholar
Nixon, B. G. et al. Cytotoxic granzyme C-expressing ILC1s contribute to antitumor immunity and neonatal autoimmunity. Sci. Immunol. 7, eabi8642 (2022).
CAS PubMed PubMed Central Google Scholar
Song, E. A. C. et al. Genetic study of Elf5 and Ehf in the mouse salivary gland. J. Dent. Res. 102, 340–348 (2023).
Lydon, J. P. et al. Mice lacking progesterone receptor exhibit pleiotropic reproductive abnormalities. Genes Dev. 9, 2266–2278 (1995).
CAS PubMed Google Scholar
Fernandez-Valdivia, R. et al. A mouse model to dissect progesterone signaling in the female reproductive tract and mammary gland. Genesis 48, 106–113 (2010).
CAS PubMed PubMed Central Google Scholar
Galouzis, C. C. & Furlong, E. E. M. Regulating specificity in enhancer-promoter communication. Curr. Opin. Cell Biol. 75, 102065 (2022).
CAS PubMed Google Scholar
Metser, G. et al. An autoregulatory enhancer controls mammary-specific STAT5 functions. Nucleic Acids Res. 44, 1052–1063 (2016).
CAS PubMed Google Scholar
Dao, L. T. M. et al. Genome-wide characterization of mammalian promoters with distal enhancer functions. Nat. Genet. 49, 1073–1081 (2017).
CAS PubMed Google Scholar
Santiago-Algarra, D. et al. Epromoters function as a hub to recruit key transcription factors required for the inflammatory response. Nat. Commun. 12, 6660 (2021).
CAS PubMed PubMed Central ADS Google Scholar
Jain, D., Baldi, S., Zabel, A., Straub, T. & Becker, P. B. Active promoters give rise to false positive ‘Phantom Peaks’ in ChIP-seq experiments. Nucleic Acids Res. 43, 6959–6968 (2015).
CAS PubMed PubMed Central Google Scholar
Kabotyanski, E. B., Huetter, M., Xian, W., Rijnkels, M. & Rosen, J. M. Integration of prolactin and glucocorticoid signaling at the beta-casein promoter and enhancer by ordered recruitment of specific transcription factors and chromatin modifiers. Mol. Endocrinol. 20, 2355–2368 (2006).
CAS PubMed Google Scholar
Kabotyanski, E. B. et al. Lactogenic hormonal induction of long distance interactions between beta-casein gene regulatory elements. J. Biol. Chem. 284, 22815–22824 (2009).
CAS PubMed PubMed Central Google Scholar
Schmidhauser, C. et al. A novel transcriptional enhancer is involved in the prolactin- and extracellular matrix-dependent regulation of beta-casein gene expression. Mol. Biol. Cell 3, 699–709 (1992).
CAS PubMed PubMed Central Google Scholar
Myers, C. A. et al. Characterization of BCE-1, a transcriptional enhancer regulated by prolactin and extracellular matrix and modulated by the state of histone acetylation. Mol. Cell. Biol. 18, 2184–2195 (1998).
CAS PubMed PubMed Central ADS Google Scholar
Winklehner-Jennewein, P. et al. A distal enhancer region in the human beta-casein gene mediates the response to prolactin and glucocorticoid hormones. Gene 217, 127–139 (1998).
CAS PubMed Google Scholar
Buser, A. C. et al. Progesterone receptor repression of prolactin/signal transducer and activator of transcription 5-mediated transcription of the beta-casein gene in mammary epithelial cells. Mol. Endocrinol. 21, 106–125 (2007).
CAS PubMed Google Scholar
Seagroves, T. N. et al. C/EBPbeta, but not C/EBPalpha, is essential for ductal morphogenesis, lobuloalveolar proliferation, and functional differentiation in the mouse mammary gland. Genes Dev. 12, 1917–1928 (1998).
CAS PubMed PubMed Central Google Scholar
Rosen, J. M., O’Neal, D. L., McHugh, J. E. & Comstock, J. P. Progesterone-mediated inhibition of casein mRNA and polysomal casein synthesis in the rat mammary gland during pregnancy. Biochemistry 17, 290–297 (1978).
CAS PubMed Google Scholar
Carelli, F. N., Liechti, A., Halbert, J., Warnefors, M. & Kaessmann, H. Repurposing of promoters and enhancers during mammalian evolution. Nat. Commun. 9, 4066 (2018).
PubMed PubMed Central ADS Google Scholar
Majic, P. & Payne, J. L. Enhancers facilitate the birth of de novo genes and gene integration into regulatory networks. Mol. Biol. Evol. 37, 1165–1178 (2020).
CAS PubMed Google Scholar
Borden, E. C. et al. Interferons at age 50: past, current and future impact on biomedicine. Nat. Rev. Drug Discov. 6, 975–990 (2007).
CAS PubMed PubMed Central Google Scholar
Shin, H. Y. et al. CRISPR/Cas9 targeting events cause complex deletions and insertions at 17 sites in the mouse genome. Nat. Commun. 8, 15464 (2017).
CAS PubMed PubMed Central ADS Google Scholar
Lee, H. K. et al. Targeting fidelity of adenine and cytosine base editors in mouse embryos. Nat. Commun. 9, 4804 (2018).
PubMed PubMed Central ADS Google Scholar
Lee, H. K. et al. Simultaneous targeting of linked loci in mouse embryos using base editing. Sci. Rep. 9, 1662 (2019).
PubMed PubMed Central ADS Google Scholar
Lee, H. K., Smith, H. E., Liu, C., Willi, M. & Hennighausen, L. Cytosine base editor 4 but not adenine base editor generates off-target mutations in mouse embryos. Commun. Biol. 3, 19 (2020).
PubMed PubMed Central Google Scholar
Bolger, A. M., Lohse, M. & Usadel, B. Trimmomatic: a flexible trimmer for Illumina sequence data. Bioinformatics 30, 2114–2120 (2014).
CAS PubMed PubMed Central Google Scholar
Langmead, B., Trapnell, C., Pop, M. & Salzberg, S. L. Ultrafast and memory-efficient alignment of short DNA sequences to the human genome. Genome Biol. 10, R25 (2009).
PubMed PubMed Central Google Scholar
Li, H. et al. The Sequence Alignment/Map format and SAMtools. Bioinformatics 25, 2078–2079 (2009).
PubMed PubMed Central Google Scholar
Heinz, S. et al. Simple combinations of lineage-determining transcription factors prime cis-regulatory elements required for macrophage and B cell identities. Mol. Cell 38, 576–589 (2010).
CAS PubMed PubMed Central Google Scholar
Ramirez, F. et al. deepTools2: a next generation web server for deep-sequencing data analysis. Nucleic Acids Res. 44, W160–W165 (2016).
CAS PubMed PubMed Central Google Scholar
Thorvaldsdottir, H., Robinson, J. T. & Mesirov, J. P. Integrative Genomics Viewer (IGV): high-performance genomics data visualization and exploration. Brief Bioinformatics 14, 178–192 (2013).
CAS PubMed Google Scholar
Love, M. I., Huber, W. & Anders, S. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biol. 15, 550 (2014).
PubMed PubMed Central Google Scholar
Guertin, M. J., Cullen, A. E., Markowetz, F. & Holding, A. N. Parallel factor ChIP provides essential internal control for quantitative differential ChIP-seq. Nucleic Acids Res. 46, e75 (2018).
PubMed PubMed Central Google Scholar
Steinhauser, S., Kurzawa, N., Eils, R. & Herrmann, C. A comprehensive comparison of tools for differential ChIP-seq analysis. Brief Bioinformatics 17, 953–966 (2016).
CAS PubMed Google Scholar
Meyer, C. A. & Liu, X. S. Identifying and mitigating bias in next-generation sequencing methods for chromatin biology. Nat. Rev. Genet. 15, 709–721 (2014).
CAS PubMed PubMed Central Google Scholar
Saleem, M. M., Mendoza-Parra, M. A., Cholley, P. E., Blum, M. & Gronemeyer, H. Epimetheus—a multi-profile normalizer for epigenomic sequencing data. BMC Bioinformatics 18, 259 (2017).
PubMed PubMed Central Google Scholar
Dobin, A. et al. STAR: ultrafast universal RNA-seq aligner. Bioinformatics 29, 15–21 (2013).
CAS PubMed Google Scholar
Anders, S., Pyl, P. T. & Huber, W. HTSeq-a Python framework to work with high-throughput sequencing data. Bioinformatics 31, 166–169 (2015).
CAS PubMed Google Scholar
Huber, W. et al. Orchestrating high-throughput genomic analysis with Bioconductor. Nat. Methods 12, 115–121 (2015).
CAS PubMed PubMed Central Google Scholar
Risso, D., Ngai, J., Speed, T. P. & Dudoit, S. Normalization of RNA-seq data using factor analysis of control genes or samples. Nat. Biotechnol. 32, 896–902 (2014).
CAS PubMed PubMed Central Google Scholar
Wickham, H. Ggplot2: Elegant Graphics for Data Analysis (Springer, 2009).
Krueger, F. & Andrews, S. R. Bismark: a flexible aligner and methylation caller for Bisulfite-Seq applications. Bioinformatics 27, 1571–1572 (2011).
CAS PubMed PubMed Central Google Scholar
Masella, A. P. et al. BAMQL: a query language for extracting reads from BAM files. BMC Bioinformatics 17, 305 (2016).
PubMed PubMed Central Google Scholar
Zhang, Y. et al. Model-based analysis of ChIP-Seq (MACS). Genome Biol. 9, R137 (2008).
PubMed PubMed Central Google Scholar
Quinlan, A. R. & Hall, I. M. BEDTools: a flexible suite of utilities for comparing genomic features. Bioinformatics 26, 841–842 (2010).
CAS PubMed PubMed Central Google Scholar
Lemay, D. G. et al. From genes to milk: genomic organization and epigenetic regulation of the mammary transcriptome. PLoS ONE 8, e75030 (2013).
CAS PubMed PubMed Central ADS Google Scholar

Download references

Acknowledgements

We thank Ilhan Akan, Sijung Yun and Harold Smith from the NIDDK genomics core for NGS and Hernan Lorenzi from the NIDDK TriLab core for advice in bioinformatic analysis. This work utilized the computational resources of the NIH HPC Biowulf cluster (http://hpc.nih.gov). H.K.L., M.W. and L.H. were supported by the Intramural Research Programs (IRPs) of the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK), and C.L. was supported by Intramural Research Programs (IRPs) of National Heart, Lung, and Blood Institute (NHLBI).

Funding

Open Access funding provided by the National Institutes of Health (NIH).

Author information

Authors and Affiliations

Section of Genetics and Physiology, Laboratory of Cellular and Molecular Biology, National Institute of Diabetes and Digestive and Kidney Diseases, US National Institutes of Health, Bethesda, Maryland, 20892, USA
Hye Kyung Lee, Michaela Willi & Lothar Hennighausen
Transgenic Core, National Heart, Lung, and Blood Institute, US National Institutes of Health, Bethesda, Maryland, 20892, USA
Chengyu Liu

Authors

Hye Kyung Lee
View author publications
You can also search for this author in PubMed Google Scholar
Michaela Willi
View author publications
You can also search for this author in PubMed Google Scholar
Chengyu Liu
View author publications
You can also search for this author in PubMed Google Scholar
Lothar Hennighausen
View author publications
You can also search for this author in PubMed Google Scholar

Contributions

H.K.L. and L.H. designed the study and analyzed data. C.L. generated mutant mice. H.K.L. established mutant mouse line. H.K.L. performed experiments and data analysis. M.W. identified multigene loci. H.K.L. and L.H. supervised the study. H.K.L. and L.H. wrote the manuscript, and all authors approved the final version.

Corresponding authors

Correspondence to Hye Kyung Lee or Lothar Hennighausen.

Ethics declarations

Competing interests

The authors declare no competing interests.

Peer review

Peer review information

Nature Communications thanks Christa Buecker, Monique Rijnkels and the other anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

Supplementary Information

Peer Review File

Description of Additional Supplementary Files

Supplementary Data 1

Supplementary Data 2

Supplementary Data 3

Supplementary Data 4

Supplementary Data 5

Supplementary Data 6

Supplementary Data 7

Supplementary Data 8

Supplementary Data 9

Supplementary Data 10

Reporting Summary

Source data

Source Data

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.

Reprints and permissions

About this article

Cite this article

Lee, H.K., Willi, M., Liu, C. et al. Cell-specific and shared regulatory elements control a multigene locus active in mammary and salivary glands. Nat Commun 14, 4992 (2023). https://doi.org/10.1038/s41467-023-40712-0

Download citation

Received: 31 January 2023
Accepted: 08 August 2023
Published: 17 August 2023
DOI: https://doi.org/10.1038/s41467-023-40712-0

Comments

By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.