Blimp-1/PRDM1 is a critical regulator of Type III Interferon responses in mammary epithelial cells

The transcriptional repressor Blimp-1 originally cloned as a silencer of type I interferon (IFN)-β gene expression controls cell fate decisions in multiple tissue contexts. Conditional inactivation in the mammary gland was recently shown to disrupt epithelial cell architecture. Here we report that Blimp-1 regulates expression of viral defense, IFN signaling and MHC class I pathways, and directly targets the transcriptional activator Stat1. Blimp-1 functional loss in 3D cultures of mammary epithelial cells (MECs) results in accumulation of dsRNA and expression of type III IFN-λ. Cultures treated with IFN lambda similarly display defective lumen formation. These results demonstrate that type III IFN-λ profoundly influences the behavior of MECs and identify Blimp-1 as a critical regulator of IFN signaling cascades.

Survival at these early stages depends on the highly specialized mammary glands that produce milk and enable the mother to feed her newborn offspring. The mammary epithelium contains two structurally and functionally distinct cell subpopulations: the outer myoepithelial/basal cells and the inner luminal cell population. Considerable progress has been directed towards understanding cell fate decisions during ductal morphogenesis, lumen formation, and alveologenesis. The functional contributions made by distinct mammary stem cell subpopulations, including both unipotent and bipotent progenitors have been extensively described in lineage tracing experiments 22 . Recent experiments demonstrate that Blimp-1 expression within a rare subset of luminal progenitors is up-regulated in response to pregnancy hormones and conditional inactivation results in defective mammary gland morphogenesis 23 . Strikingly Blimp-1 functional loss disrupts epithelial architecture and lumen formation both in vivo and in three-dimensional (3D) primary cell cultures 23 .
To further investigate the underlying causes of these tissue disturbances, here we performed transcriptional profiling experiments. To avoid studying possible contributions made by strongly Blimp-1-positive endothelial cells within this highly complex vascularized tissue in vivo, we exploited the inducible ROSA26:CreERT2 allele to activate Cre-mediated Blimp-1 deletion via tamoxifen treatment of 3D mammary epithelial cell (MEC) cultures. As expected Blimp-1 deficient (cKO) MEC cultures display defective lumen formation and fail to establish apical-basal polarity. Surprisingly however functional annotation analysis of up-regulated genes identified highest enrichment scores for categories associated with innate immunity and IFN signaling pathways. This gene list substantially overlaps with those recently described as conserved Blimp-1 targets via ChIP-Seq analysis, including key components of the MHC class I peptide-loading pathway 20 . Additionally IFN-stimulated genes and pathway regulators including Usp18, Oligoadenylate synthetase (OAS) family members, and the key transcriptional activator Stat1, were identified here as direct Blimp-1 targets. Finally the present experiments demonstrate that Blimp-1 loss of function results in up-regulated expression of double stranded (ds)RNA and type III IFN lambda (IFN-λ). Moreover type III IFN-λ treatment of wild type MEC cultures causes defective lumen formation and maturation. These results demonstrate for the first time that the zinc finger transcriptional repressor Blimp-1 silences type III IFN expression in mammary epithelial cells and establish its actions as a global regulator of the IFN signaling cascade.

Transcriptional profiling experiments demonstrate up-regulated expression of IFN signaling components.
To learn more about Blimp-1 functional contributions during mammary gland morphogenesis, we used the Illumina array platform to compare transcripts in wild type versus Blimp-1 cKO MEC cultures harvested at Day 3 (D3) and Day 4 (D4), when tissue disturbances become readily visible. The greatest differences in gene expression profiles were detectable at Day 4. The relatively few down-regulated genes (n = 138) in D4 Blimp-1 cKO MECs showed no significant enrichment for any GO annotated pathways or biological processes and were not investigated further (see Supplementary Table S1). In contrast, functional annotation clustering analysis of significantly up-regulated genes surprisingly revealed a striking enrichment for innate immunity, stress and viral response categories (Fig. 1a). Results summarized in Fig. 1b strongly suggest that Blimp-1 regulates expression of anti-viral effectors such as oligoadenylate synthetase (OAS) family members, IFN-stimulated genes (ISG's) such as Usp18 and Isg15, IFN-inducible components of the MHC class I peptide-loading pathway, as well as the transcriptional activators Irf7, Stat1 and Stat2.
Additionally a subset of up-regulated transcripts overlapped with those previously identified in Blimp-1 cKO embryonic (E18.5) small intestine 19 . These expression changes were validated by quantitative PCR (qPCR, Fig. 1c), and additionally for Stat1 and Stat2 in cell staining experiments (Fig. 1d). Representative images shown in Fig. 1d demonstrate dramatically up-regulated and sustained predominantly nuclear Stat1, pY-Stat1 and Stat2 expression.
Identification of overlapping Blimp-1 and Stat1 target genes. Comparisons of our list of up-regulated transcripts with the direct Blimp-1 target genes previously identified in ChIP-Seq experiments revealed substantial overlap (Fig. 2a). Moreover, there are Blimp-1 ChIP peaks present in the promoter regions of several IFN signaling pathway regulators -i.e., Stat1, Usp18, Oas3, and Oasl2 (Fig. 2b). Stat1 plays a central role in transcriptional activation of both type I and type II IFN responses 24,25 . Results above demonstrate increased Stat1 expression readily detectable at Day 3 the earliest time point examined (Fig. 1c). As shown in Fig. 2c, the vast majority of up-regulated genes identified here have been previously identified as Stat1 target genes, containing ChIP-Seq peaks constitutively occupied by Stat1 as well as those activated in response to type I IFN-β recognized by Stat1-Stat2 heterodimers together with IRF-9 -i.e., the so-called ISGF3 complex 24 . These observations suggest that dramatically increased and sustained Stat1 expression in the absence of Blimp-1 could potentially be responsible for activation of IFN-signaling genes.
Blimp-1 expression in the mammary gland is restricted to a rare subset of luminal progenitors, representing less than 1% of the total luminal cell population recovered from pregnant females 23 . The eGFP-tagged knock-in allele used previously for ChIP-Seq analysis of embryonic day (E) 18.5 small intestine is severely hypomorphic elsewhere 20 and FACS sorting is problematic due to relatively weak eGFP fluorescence intensity even in plasma cells. For these reasons, we elected to perform qPCR validation of Blimp-1 occupancy at candidate target sites using transiently transfected CommaDβ cells, an established cell line that shares many functional and behavioral characteristics with bona fide mammary stem cells 26 , in combination with our expression construct encoding the C-terminal eGFP-tagged Blimp-1 fusion protein and the proven GFP mAb 27 . Results shown in Supplementary Figure S1 confirm Blimp-1 occupancy at key target sites in mammary epithelial cells. transcriptional profiling experiments shown above reveal that Blimp-1 conditional loss causes up-regulated expression of several dsRNA binding proteins including the DEXD/H box helicases Ddx60, Dhx58 (LGP2), and RIG-1 (Ddx58) (Fig. 1a). However, these were not identified as direct target genes. Interestingly, recent studies demonstrate increased expression of dsRNA is associated with global epigenetic re-programming in human colorectal cancer cells exposed to DNA-demethylating agents 30 . We wondered whether Blimp-1 functional loss might similarly activate viral mimicry pathways upstream of IFN signaling responses. To test this possibility, next we used the well characterized J2 monoclonal antibody that specifically recognizes dsRNA greater than 40 bp in

Expression of type III IFN-λ by Blimp-1 cKO mammary epithelial cells. Recent experiments
demonstrate that type III IFN-λ via its unique receptor complex selectively activates anti-viral defense pathways in epithelial cells 32,33 . Considerable evidence demonstrates that type III IFN-λ plays a crucial protective role at mucosal barrier surfaces lining the respiratory, gastro-intestinal, and reproductive tracts 32,33 . Blimp-1 was previously shown to silence IFN-λ1 expression in human lung epithelial cells 34 . However in mice IFN-λ1 is a pseudogene. To test whether Blimp-1 regulates type III IFN-λ expression in mouse mammary epithelial cells, here we performed RT-PCR experiments using primers designed to detect IFN-λ2/3 transcripts. As shown in Fig. 4a and b, expression of type III IFN-λ2/3 and type I IFN-β was dramatically up-regulated in Blimp-1 cKO MEC cultures. In contrast, there was no evidence for expression of type II IFN-γ by these highly organized epithelial cell cultures.
Overlapping functional roles played by individual IRF family members have been intensely investigated 21 . Irf-3 and Irf-7 have been shown to induce Type I and Type III IFN expression in virally infected cells 35 . Interestingly we found here that Irf-1, 2, 3, and 6 are constitutively expressed by mammary epithelial cells whereas Blimp-1 inactivation leads to increased expression of Irf-7 and Irf-9 (Fig. 4c). Additionally IFN-λ receptor complexes were robustly expressed by MEC 3D cultures (Fig. 4c) suggesting that Type III IFN-λ could potentially influence the behavior of mammary epithelial cells.

Treatment with Type III IFN lambda disrupts epithelial cell polarity.
To directly test this possibility, recombinant IFNs were added at Day 0 to MEC 3D cultures. As shown in Fig. 5a, IFN-λ treatment resulted in dramatic morphological changes including the collapsed acini phenotype, as well as dramatically up-regulated Stat1, pY-Stat1, and Stat2 expression (Fig. 5b). Moreover treatment with IFN-λ resulted in up-regulated gene expression profiles as described above for Blimp-1 cKO MEC 3D cultures ( Supplementary Fig. S3). In contrast we failed to observe any noticeable effect in the presence of type II IFN-γ, whereas treatment with type I IFN-β was associated with an intermediate phenotype ( Fig. 5a and Supplementary Fig. S4). These results demonstrate for the first time that as for diseases associated with excessive type I IFN activities 36 , that type III IFN-λ also has potentially detrimental effects. Thus it appears that Blimp-1 normally functions to restrain type III IFN-λ signaling and maintains the ability of mammary epithelial cells to differentiate normally and undergo ductal morphogenesis.

Discussion
The ability of Blimp-1 to bind specifically to the PRD1 element of the human IFN-β gene promoter and silence expression was characterized nearly three decades ago 1 . We now understand that Blimp-1 plays essential roles governing cell fate decisions in a wide variety of tissue contexts dependent not only on target site selection but also its ability to recruit co-repressor complexes required to silence gene expression 3 . Recent experiments exploiting powerful high throughput methodologies have provided detailed insights into its functional activities guiding lineage commitment and terminal differentiation in specialized cell populations 6,9,13 . Early studies revealed that the Blimp-1 consensus binding motif closely resembles the IRF-E sequence element recognized during activation of IFN responses. Ironically at present, relatively little is known about Blimp-1-dependent regulation of innate immunity or IFN signaling pathways.
The transcriptional networks and signaling pathways controlling dynamic developmental changes observed in the mammary gland during puberty, pregnancy, lactation, and involution, have been intensely investigated 22 .
Our recent experiments demonstrate that Blimp-1 expression in a rare subset of luminal progenitors plays an essential role in mammary gland development 23 . Blimp-1 regulates proliferation and luminal cell maturation during puberty and pregnancy. Conditional inactivation causes defective epithelial cell polarity and disrupts milk secretion 23 . To learn more about Blimp-1 functional activities in the mammary gland, here we performed transcriptional profiling experiments.
Surprisingly up-regulated genes in Blimp-1 cKO MEC 3D cultures were greatly enriched for key regulators of viral defense, innate immunity, and IFN signaling pathways. This result was entirely unexpected since only healthy animals were used in our experiments. Besides protective roles played by IFN signaling activated during viral infections, constitutive expression of type I IFN at low levels in the absence of viral infection is also known to be required for the maintenance of tissue homeostasis under diverse circumstances. The importance of so-called tonic signaling in hematopoietic stem cells 37 , thymic epithelial cell development 38 and immune cell function 39 has been previously characterized. Recent experiments suggest that Type I IFN could potentially regulate functional properties of mammary stem cells 40 . Adaptive immune responses including CD4 + T cell production of type II IFN-γ have been shown to contribute to mammary gland organogenesis 41 . In contrast here to avoid potential contributions made by lymphocytes, endothelial cells, macrophages, and local stromal elements, we exploited 3D mammary epithelial cell cultures where the formation of self-organized acini structures recapitulates those normally observed during mammary gland organogenesis. Under these circumstances, we found that Blimp-1 inactivation resulted in up-regulated expression of type I and type III IFNs, but there was no evidence for expression of Type II IFN-γ.
The present experiments demonstrate for the first time in the mammary epithelium that Blimp-1 directly silences expression of IFN signaling pathway regulators including Oas family members, the ubiquitin-specific peptidase Usp18, and importantly the transcriptional activator Stat1 that universally drives gene expression changes upstream of Type I, Type II, as well as Type III IFN signaling cascades. The balance between Type I vs  Table S1). For comparison, microarray probe signals from embryonic (E) day 18.5 and postnatal (P) day 7 wild type and Blimp-1 cKO small intestine datasets corresponding to NCBI GEO accession numbers GSE29658 and GSE30556 respectively, are shown. Type II target site selection by Stat1 reflects its phosphorylation status and distinct transcriptional partnerships 24 . Additionally unphosphorylated Stat1 has been shown to selectively regulate target gene expression 25,42 . For example, its association with IRF-1 activates constitutive LMP2 gene expression 43 . Here we report in the absence of Blimp-1 that up-regulated genes substantially overlap with those activated by Stat1. Thus, by virtue of its ability to directly antagonize Stat1 expression, Blimp-1 potentially functions as a global regulator of IFN signaling pathways.
Since its discovery in 2003, Type III IFN-λ has been shown to selectively provide anti-viral protection in epithelial cells, particularly those present at the mucosal surfaces lining the respiratory and gastrointestinal tracts 32,33 . Recent experiments suggested that Blimp-1 is a transcriptional repressor of IFN-λ1 expression in virally infected human airway epithelial cells 34 . However, the type III IFN genes display considerable divergence across evolution and human IFN-λ1 corresponds to a pseudogene in mice 44 . Interestingly however ChIP on chip experiments identified highly conserved Blimp-1 binding sites adjacent to the human IFN-λ2 promoter 45 and Blimp-1 knockdown increased levels of IFN-λ2/3 transcripts in cultured human airway epithelium 34 . Consistent with the idea that Blimp-1 recognizes these highly conserved ISRE sites, here we observed up-regulated IFN-λ2/3 expression in Blimp-1 cKO MEC cultures. Thus, combinatorial regulatory motifs controlling type III IFN gene expression closely resemble those governing type I IFN-β expression 34,35 .
The simplest scenario is that Blimp-1 occupancy at ISRE/PRD1 sites directly silences IFN-λ2/3 gene expression. Unfortunately however the eGFP-tagged Blimp-1 knock-in allele successfully used for ChIP-Seq analysis in E18.5 intestinal epithelium is only weakly expressed in the mammary gland in a rare subpopulation of luminal progenitors. Occupancy at selected target sites was validated in qPCR experiments using transiently transfected and FACS-sorted CommaDβ cells enriched for expression of the eGFP-tagged Blimp-1 fusion protein. However we found no evidence for significant binding at predicted sites upstream of the IFN-λ2/3 gene promoters. Unfortunately ectopic Blimp-1 over-expression is toxic excluding analysis of stably transfected cell lines and limiting our ability to explore Blimp-1 occupancy at multiple time points. The present results clearly demonstrate that Blimp-1 function is required to prevent type III IFN-λ expression in mammary epithelial cells. However, additional work will be needed to characterize its role in assembly of higher order complexes governing chromatin accessibility and transcriptional output at this locus.
Recent studies implicate a functional relationship between epithelial cell polarity and IFN-λ expression from peroxisomes 46 . Cytoplasmic sensors such as RIG-1 (Ddx-58) that recognize viral RNA trigger membrane bound signaling proteins such as MAVS that in turn activate assembly of transcriptional complexes driving IFN-stimulated gene expression. In contrast to these disturbances downstream of viral entry and replication, the striking morphological changes described here associated with up-regulated IFN-λ expression are simply caused by Blimp-1 inactivation. Interestingly Blimp-1 cKO also resulted in accumulation of dsRNA recognized by the J2 monoclonal antibody. Thus, activation of IFN signaling responses could simply be due to the presence of dsRNA.
The J2 mAb recognizes dsRNA greater than 40 bp in length but its fine specificity remains ill defined. An extensive literature has described diverse host mechanisms responsible for controlling expression of endogenous retroviral sequences scattered across the genome. It is tempting to speculate that as for many other zinc finger proteins, Blimp-1 may directly silence ERV transcription in a sequence specific manner. However mapping ChIP seq peaks to repetitive elements requires the application of sophisticated bioinformatics strategies. An important long term goal is to characterize dsRNAs that accumulate in the absence of Blimp-1 and test this attractive model.
The present results help to resolve the paradoxical finding previously described in our recent studies, namely how could Blimp-1 inactivation in rare progenitor cell population cause such dramatic widespread phenotypic changes throughout the entire epithelial cell population 23 . We now understand in the absence of Blimp-1 that type III IFN-λ secretion by rare luminal progenitors acting in a paracrine fashion has the ability to influence the behavior of neighboring cells. Our results demonstrate for the first time that type III IFN-λ not only activates anti-viral responses in mucosal epithelial cells, but also in mammary epithelial cells. While the role played by type III IFNs in innate immune protection at mucosal barrier surfaces has been extensively investigated, relatively little is known about viral infections in the mammary epithelium or how this impacts on neonatal health. It will be important to learn more about protective and potentially negative effects mediated by type III IFN-λ.

Methods
Animals. Female C57BL/6 mice (8-10 weeks of age) were used as the wild type strain. For inducible Blimp-1 gene deletion in mammary epithelial cell (MEC) cultures, Prdm1 BEH/+ and Prdm1 CA/CA mice 23 were crossed with the ROSA26:Cre ERT2 line 47 to generate Prdm1 BEH/CA ;ROSA26:Cre ERT2 females. Mice were genotyped by PCR as described in the original reports. All animal experiments were performed in accordance with Home Office regulations and were approved by the University of Oxford Local Ethical Committee.
Immunofluorescence. For immunofluorescence staining, acini were fixed with 4% PFA in PBS and permeabilized with 0.5% Triton X-100 in PBS. Fixed cells were blocked with 10% normal goat serum/2% BSA in PBS for 2 hrs at RT, and incubated with Blimp-1, Stat1, pStat1 (Tyr701), Stat2, or dsRNA (J2) antibodies (see Supplementary Table S2 for antibody dilutions and source) in block overnight at 4 °C. Primary antibodies were detected by incubation with an appropriate secondary Alexa Fluor-conjugated antibody (Molecular probes). F-Actin was detected using Phalloidin-Alexa Fluor 633 (Invitrogen). Acini with F-actin staining at the apical surface of cells surrounding a single lumen were identified as acini with normal lumens. For all immunostaining, slides were mounted in DAPI-containing Fluoroshield. The pictures were captured with an inverted Olympus FV1200 laser scanning confocal microscope equipped with X40 and X60 oil-immersion objective. Z stack steps were of 1 μm. Images were analyzed with ImageJ.
RNA extraction, qRT-PCR and Microarray profiling. Total RNA from five independent MEC 3D cultures each from Prdm1 BEH/CA ;ROSA26:Cre ERT2 and wild type treated with tamoxifen on day 0 and harvested on Day 3 and 4 was isolated using an RNeasy Mini kit (Qiagen) with on column DNase treatment according manufacturer's protocol. Microarray transcriptional profiling experiments were performed using triplicate RNA samples of each genotype at both Day 3 and 4 as previously described using Illumina Mouse WG-6 v2 Expression BeadChips 8 . Differential probe expression was determined following rank-invariant normalization using the Illumina custom error model option of Gene Expression Analysis Module V1.6.0 of GenomeStudio V2009 (Illumina) with Benjamini and Hochberg false discovery rate. Probes with significant different expression (1.5 fold and differential score ≥ 13, equivalent to P ≤ 0.05) were analyzed by WebGestalt using default parameters (http://www.webgestalt.org/option.php). qPCR analysis was performed on 5 RNA samples for each genotype at day 3 and 4. First-strand cDNA was reverse transcribed from RNA (1 μg) using Superscript III (Invitrogen, UK) according to manufacturer's protocol and qPCR performed using QuantiTech SYBR Green master mix (Q2040143; Qiagen) on a Rotor-Gene Q (Qiagen). Relative gene expression was calculated using the ΔΔCt method in comparison with Hprt as the reference. Reverse transcription-PCR (RT-PCR) was performed using the OneStep RT-PCR kit (Qiagen) according to manufacturer's protocol. Primer sequences are provided in Supplementary Table S3.
Published ChIP-Seq dataset analysis. ChIP-Seq datasets corresponding to NCBI GEO accession numbers GSE66069, GSE60204 and GSE33913 were used to identify Blimp-1 6,20 and Stat1 24 genomic binding sites. ChIP-peaks were functionally annotated with GREAT version 3.0.0 49 , using the basal extension rule linking peaks to the nearest transcription start site (TSS) ± 100 kb. qPCR-ChIP. CommaDβ cells were transiently transfected with a plasmid expression construct encoding C-terminal eGFP-tagged Blimp-1 fusion protein 27 using Lipofectamine 2000 (Invitrogen) according to the manufacturer's protocol. GFP positive cells, sorted by FACS at 24-48 hours post transfection were cross-linked with 1% formaldehyde in growth media for 20 min at RT and subsequently processed for ChIP using 10 μg of mouse anti-GFP IgG2a (clone 3E6, A11120; Invitrogen), as described previously 50 . Non-transfected CommDβ cells were used as negative controls. Eluted DNA samples were recovered using ChIP DNA Clean and Concentrator column kit (Zymo Research) and qPCR analysis of quadruplicate ChIP and input samples performed using QuantiTech SYBR Green master mix (Q2040143; Qiagen) on a Rotor-Gene Q (Qiagen). Primers were designed to amplify 100-200 bp regions central to Blimp-1 binding sites identified in ChIP-Seq dataset GSE66069. Selected genes included c-Myc, Prdm1, Stat1 (both TSS and 5' of TSS ChIP-Seq peak regions -see Fig. 2b), Oasl2 and Usp18. A non-enriched ChIP-Seq region in the 3'UTR of the Prdm1 gene was used as a negative control 20 . Primer sequences are provided in Supplementary Table S3. Fold enrichment of ChIP over input was determined relative to a standard curve generated from log-diluted sheared genomic DNA. Statistical analysis. Statistical analyses were performed using two-tailed unpaired t-test for wild type versus Blimp-1 cKO comparisons and ANOVA followed by Tukey post-hoc test to compare between treatment conditions (Microsoft Excel and GraphPad Prism 7.0.). P values ≤0.05 were considered significant. Data availability. The microarray data has been deposited in NCBI GEO with the accession number GSE100747.