Activation of ERα Signaling Differentially Modulates IFN-γ Induced HLA-Class II Expression in Breast Cancer Cells

The coordinate regulation of HLA class II (HLA-II) is controlled by the class II transactivator, CIITA, and is crucial for the development of anti-tumor immunity. HLA-II in breast carcinoma is associated with increased IFN-γ levels, reduced expression of the estrogen receptor (ER) and reduced age at diagnosis. Here, we tested the hypothesis that estradiol (E2) and ERα signaling contribute to the regulation of IFN-γ inducible HLA-II in breast cancer cells. Using a panel of established ER− and ER+ breast cancer cell lines, we showed that E2 attenuated HLA-DR in two ER+ lines (MCF-7 and BT-474), but not in T47D, while it augmented expression in ER− lines, SK-BR-3 and MDA-MB-231. To further study the mechanism(s), we used paired transfectants: ERα+ MC2 (MDA-MB-231 c10A transfected with the wild type ERα gene) and ERα− VC5 (MDA-MB-231 c10A transfected with the empty vector), treated or not with E2 and IFN-γ. HLA-II and CIITA were severely reduced in MC2 compared to VC5 and were further exacerbated by E2 treatment. Reduced expression occurred at the level of the IFN-γ inducible CIITA promoter IV. The anti-estrogen ICI 182,780 and gene silencing with ESR1 siRNA reversed the E2 inhibitory effects, signifying an antagonistic role for activated ERα on CIITA pIV activity. Moreover, STAT1 signaling, necessary for CIITA pIV activation, and selected STAT1 regulated genes were variably downregulated by E2 in transfected and endogenous ERα positive breast cancer cells, whereas STAT1 signaling was noticeably augmented in ERα− breast cancer cells. Collectively, these results imply immune escape mechanisms in ERα+ breast cancer may be facilitated through an ERα suppressive mechanism on IFN-γ signaling.

HLA-II expression is controlled at the transcription level by a highly conserved regulatory module, located in the promoter of genes encoding the aand b-chains of all HLA-II molecules and in the gene encoding the Ii co-chaperone [21][22][23][24][25][26]. This regulatory module forms a platform for the class II transactivator (CIITA), a non-DNA binding protein, which acts as a transcriptional integrator by connecting transcription factors, bound to the MHC-II promoter with components of the general transcriptional machinery [23,[27][28][29][30]. The central role of CIITA is evident from lack of constitutive or IFN-c inducible HLA-II in bare lymphocyte syndrome [31,32].
CIITA expression is controlled by three distinct promoters: promoter I (pI) for constitutive expression in dendritic cells; promoter III (pIII), for constitutive expression in B cells; promoter IV (pIV) for IFN-c inducible expression [21,26,33]. This promoter system is crucial for controlling CIITA messenger RNA (mRNA) and protein levels, and they, in turn, regulate HLA-II expression. The molecular regulation of CIITA pIV is intricately linked to the classical IFN-c signaling pathway. IFN-c, binds to IFN-c receptors (IFNGR) on the cell surface, resulting in autophosphorylation of Janus kinase 2 (JAK2) and JAK1, followed by phosphorylation, dimerization and nuclear translocation of signal transducer and activator of transcription 1 (STAT1) [34,35]. Phosphorylated STAT1 (pSTAT1) binds to IFN-activated sites (GAS) in the promoter of target genes including the IFN-regulatory factor 1 (IRF1), thus stimulating its expression. IRF1 binds cooperatively with IRF2 to its associated IRF element (IRF-E) in CIITA pIV, and concomitant pSTAT1 binding to GAS in CIITA pIV results in transcriptional activation of CIITA [33,36]. Moreover, signaling pathways such as mitogen activated protein kinases (MAPK) and PI3K/Akt that are frequently activated in breast cancer cells [37] modulate expression of IRF1 and STAT1 [38][39][40], further impacting the levels of IFN-c inducible CIITA and subsequent HLA-II expression on tumor cells.
Previously, we showed that HLA-II (HLA-DR, HLA-DM and Ii) was discordantly expressed on tumor cells in human breast cancer tissues [12]. Furthermore, tumor cell expression of HLA-DR and Ii, but not HLA-DM, correlated with reduced expression of estrogen receptors (ER) and reduced age at diagnosis. Importantly, tumors with coordinate expression of HLA-DR, Ii and HLA-DM had the highest IFN-c mRNA levels and correlated with increased patient survival [12]. Undoubtedly, the mechanisms governing tumor cell expression of HLA-II in breast carcinoma are likely multifaceted, involving IFN-c secreted by infiltrating immune cells [12], circulating and tumor-associated estrogens [41] and activation of growth factor and hormone receptor pathways in the tumor cells [42,43]. Estradiol and antiestrogens, tamoxifen and fulvestrant or ICI 180,720 (ICI), were shown to modulate IFN-c inducible MHC-II in various cell types [17,19,44,45] through mechanisms not involving ligand activation of the estrogen receptor (ER) pathway.
In this study, using established human ER 2 and ER + breast cancer cell lines (BCCL) and an ERa-transfected BCCL, we investigated the specific and combined effects of estradiol (E 2 ) and ERa on HLA-II regulation. We found IFN-c inducible HLA-II expression was modulated by E 2 -ER activation at the level of the CIITA pIV. Furthermore, E 2 -treatment of ERa + BCCL and ERa 2 BCCL differentially affected various components of the IFN-c signaling pathway that are required for transactivation of CIITA pIV.

Results
Estradiol differentially modulates HLA-DR expression in breast cancer cell lines Stemming from our previous finding that HLA-II expression in breast carcinoma tissues correlates with increased IFN-c mRNA, reduced age at diagnosis and reduced ER levels [12] we questioned whether E 2 , in the absence or presence of its cognate receptor ERa, modulates HLA-DR expression in established ER 2 and ER + BCCL, treated or not with IFN-c for 96 hours. Analysis of ER 2 BCCL using flow cytometry ( Figure 1A & 1B) revealed low basal expression of HLA-DR in MDA-MB-231, but not in SK-BR-3 while IFN-c induced strong expression in both cell lines. E 2treatment augmented IFN-c inducible HLA-DR, although this was significant for only SK-BR-3 ( Figure 1B). These results, confirmed by Western blot analysis of cell lysates ( Figure 1C & 1D), suggest E 2 may modulate HLA-DR expression in ER 2 breast cancer through an ERa independent mechanism [46].
Since the least HLA-DR in human breast carcinoma tissues occurred in ER + tumors [12] we hypothesized that E 2 -activation of the ERa pathway inhibits HLA-DR expression. Analysis of ER + BCCL, treated as described above, revealed a variable pattern of IFN-c inducible HLA-DR expression with amounts that were barely detectable, moderate and abundant in BT-474, MCF-7, and T47D, respectively ( Figure 1E & 1F). Constitutive HLA-DR was detected at the cell surface in only T47D ( Figure 1E). Furthermore, E 2 treatment significantly reduced HLA-DR in MCF-7 and BT-474, but not in T47D ( Figure 1E). Similar results were obtained from Western blot analysis of cell lysates ( Figure 1G  & 1H). Notably, ERa levels were not altered by IFN-c but E 2 treatment increased the amount in the nucleus, indicating ligand activation of the ERa pathway ( Figure. 1G). Taken together these data suggest that E 2 -inhibition of HLA-II expression in ERa + BCCL is mediated through activation of ligand-dependent ERa pathway.

Transfection of ESR1 in an ERcell line diminishes IFN-c inducible HLA-II proteins
To further explore the role of ERa on IFN-c inducible HLA-DR, we used two stably transfected cell lines, derived from MDA-MB-231 clone 10A [47,48]: MC2 expresses wild type ERa and VC5 expresses the empty vector. Since MDA-MB-231 clone 10A was selected for negative expression of ERa and ERb [47], the transfected pair is a suitable model to assess ERa mediated effects on HLA-II without interference from other ERs including GPR30, reported to be deficient in MDA-MB-231 [48,49]. The cells, treated and analyzed for HLA-DR expression as described above, revealed significantly reduced cell surface HLA-DR in MC2, as compared to VC5 and MDA-MB-231 clone 10A (Figure 2A & 2B). Moreover, E 2 -treatment greatly diminished HLA-DR in MC2 but not in VC5 and MDA-MB-231 clone 10A. These results were confirmed by Western blot analysis of cell extracts ( Figure 2C). Again, HLA-DR protein in the ERa + MC2 was severely reduced and exacerbated by E 2 , whereas MDA-MB-231 clone10A and VC5 expressed abundant HLA-DR in the presence and absence of E 2 . As the only known difference between MC2 and VC5 is the expression of ERa, these results further implicate ERa in negatively regulating HLA-DR expression.
Although HLA-II genes are coordinately regulated [25], we found most breast cancer lesions with HLA-DR + tumor cells do not have detectable HLA-DM expression [12]. We reasoned that if ERa and its activation by E 2 coordinately down regulates HLA-II, then blocking ER signaling with ICI, a selective anti-estrogen that degrades ER, should reverse the inhibition. To test this hypothesis, MC2 and VC5 were pretreated with 10 26 M ICI in the presence or absence of 10 29 M E 2 . Following stimulation with IFN-c for 96 hours, HLA-DR, -DM and Ii were analyzed by flow cytometry and Western blot. HLA-DR, -DM and Ii expression levels were significantly reduced in MC2 compared to VC5 ( Figure 3A-3C), while E 2 -treatment further diminished HLA-II expression in MC2, but not in VC5. Although ICI-treatment, alone or with E 2 , did not restore HLA-II in MC2 to VC5 levels, it clearly reversed the E 2 -inhibitory effect on HLA-II expression. Western blot analysis ( Figure 3D-3G) and immunocytochemistry (data not shown) confirmed the reduced expression of HLA-DR, -DM and Ii in MC2 and the involvement of ERa signaling in the inhibitory effect of E 2 on HLA-II expression.

Activation of the ERa signaling pathway impedes CIITA expression
Since HLA-II expression is coordinately regulated by CIITA, we predicted that ERa interfered with CIITA expression in ERaexpressing MC2. MC2 and VC5 were pretreated with E 2 and/or ICI, as described above, followed by addition of IFN-c for 24 hours. Western blot analysis of nuclear and cytoplasmic extracts showed inducible CIITA expression in MC2 was about 70% of VC5 levels ( Figure 4A & 4B). E 2 -treatment further reduced CIITA in MC2 while increasing the amount of nuclear ERa; in contrast, ICI reversed the inhibitory effect of E 2 on CIITA expression, coincident with ICI-mediated reduced ER levels ( Fig 4A Lanes 7 and 8). These results indicated that E 2 inhibits HLA-II expression by downregulating CIITA expression.
To further determine the inhibitory effect of E 2 on CTIIA gene expression, VC5 and MC2 cells were pretreated with E 2 and/or ICI for 1 hour and then stimulated with and without IFN-c for 4 hours, an optimal time for CIITA mRNA expression [50]. CIITA transcription was induced in both VC5 and MC2, but the induction of CIITA mRNA in MC2 was about half in VC5 ( Figure 4C). E 2 further decreased CIITA mRNA in MC2, while ICI reversed the E 2 -mediated effect on CIITA.
To confirm the above results, we silenced the ERa transgene in MC2 using ESR1 siRNA and then treated with E 2 or vehicle control followed by IFN-c stimulation for 24 hours. VC5, treated in the same way, was used as a control. Western blot analysis of cell lysates showed ERa was greatly reduced in MC2 transfected with ESR1 siRNA, but not with scrambled siRNA ( Figure 5A). Similar to the ICI-mediated effects, ESR1 siRNA clearly reversed the E 2 -mediated inhibition observed in the scrambled siRNA transfectants. E 2 increased CIITA in the ER 2 VC5, whether transfected with scrambled or ESR1 siRNA. Analysis of CIITA transcripts using real time PCR on siRNA-treated cells ( Figure 5B), revealed equivalent levels of CIITA transcripts in ESR1 and scrambled siRNA transfectants; again, ESR1-siRNA abolished the inhibitory effect of E 2 on constitutive and induced CIITA transcripts. These results suggest a mechanism whereby E 2activated ER interferes with CIITA transcription induced by IFNc in breast cancer cells.

E 2 activated ERa inhibits CIITA promoter IV activity
Since IFN-c inducible HLA-II expression requires activation of CIITA pIV [33], we hypothesized that E 2 activation of ERa interferes with CIITA pIV activity. We transfected VC5 and MC2 with a CIITA pIV luciferase construct and treated the cells with E 2 and/or ICI, followed by stimulation or not with IFN-c for 12 hours. E 2 -treatment further reduced both basal and IFN-c induced CIITA pIV activity in MC2, while ICI reversed the inhibitory effect of E 2 in MC2 cells ( Figure 6). Treatment with ICI and/or E 2 did not significantly affect constitutive or IFN-c inducible CIITA pIV activity in VC5.
To determine whether E 2 directly regulates CIITA pIV activity, we searched for presence of ERE sites using three different computer software programs (http://tfbind.hgc.jp/, http:// alggen.1si.upc.es/ and http://www.cbrc.jp/index.eng.html) and identified four putative ERE sites in CIITA pIV ( Figure 7A, bold letters in boxes). Sites 1 to 3 are upstream of the STAT1 and IRF1 binding sites. Site 4 is downstream of these sites and precedes the start codon. To determine if either of these sites serves as an ERa repressor of CIITA transcription, three deletion mutant constructs (Site 1/2 deletion mutant, Site 3/4 deletion mutant and Site 1-4 deletion mutant) were created ( Figure 7A, open boxes). VC5 and MC2, transfected with one of the mutant CIITA pIV constructs, were pretreated with E 2 or vehicle control and then stimulated with IFN-c for 12 hours, followed by measurement of luciferase activity ( Figure 7B, left panel). All three deletion constructs demonstrated significantly reduced IFN-c stimulated CIITA pIV activity in E 2 -treated MC2, similar to that observed in MC2 transfected with the wild type CIITA pIV plasmid. By comparison CIITA pIV activity was similar in E 2 or vehicle treated VC5 cells whether transfected with wild type or deletion constructs. Intriguingly, constructs Del 3 & 4 and Del 1-4 resulted in dramatic and significant loss of CIITA pIV activity in both cell lines, suggesting there may be other or overlapping sites in CIITA pIV that interact with currently unknown transcription factors for a fully active promoter. Alternatively, the deletion of these sites may have led to the creation of a novel site that has an inhibitory effect on CIITA pIV activity. Importantly, these results do not support the hypothesis that diminished CIITA pIV activity in MC2 treated with E 2 occurs via ERE sites in the proximal region of CIITA pIV.

E 2 -ERa interferes with STAT1 signaling in ERa transfected MC2 cells
To explore whether STAT1 signaling, necessary for activation of CIITA pIV, is adversely affected by ERa activation, we transfected the 8 X GAS luciferase plasmid in VC5 and MC2, followed by treatment, or not, with E 2 and/or IFN-c for 6 hours. Compared to VC5, STAT1 signally was clearly reduced in MC2 ( Figure 8A & 8B); moreover, E 2 significantly reduced basal and induced GAS promoter activity by about 44% and 40%, respectively, in MC2 ( Figure 8B). Although E 2 increased basal GAS promoter activity by about 28% in VC5, this was not significant; E 2 had no effect on induced activity ( Figure 8A).
To test whether reduced GAS activity in MC2 was the result of reduced pSTAT1, we performed Western blot analysis on lysates from cells treated or not with IFN-c for 15 minutes. As shown in Figure 8C, total STAT1 and pSTAT1 at tyrosine (Y) 701 and serine (S) 727 were reduced in MC2, compared to VC5. Similar results were observed in an experiment in which cells were also treated with E 2 for 4 hours, followed by IFN-c treatment for 15 minutes; moreover, E 2 did not alter levels of phosphorylated or total STAT1 in MC2 or in VC5 ( Figure 8D). We next examined IRF1 expression, also essential for CIITA pIV activation, in MC2 and VC5, treated with E 2 and stimulated with IFN-c for 96 hours ( Figure 8E). We found IRF1 levels were significantly decreased in MC2, compared to VC5, that E 2 -treatment had only a trivial effect on IRF1 in MC2, whereas it significantly increased the levels in VC5. Collectively, these results show that ectopic expression of ERa and, moreover, its activation by E 2 attenuates STAT1 signaling, however, E 2 has only a marginal inhibitory effect on IRF1 levels in MC2. These findings imply that attenuation of CIITA pIV and subsequent reduced HLA-II expression in ERa positive breast cancer may be due to defects in STAT1 regulation. E 2 differentially affects IFN-c signaling in established ERa + and ERa 2 breast cancer cells To ensure that attenuated STAT1 signaling in MC2 was not merely a peculiarity of the transfected model, we further analyzed These included STAT1, IRF1, IRF9, a member of the IRF family of transcription factors that is not implicated in CIITA expression [51], and gamma-interferon-inducible lysosomal thiol reductase (GILT), a STAT1 regulated but CIITA-independent protein, that is important for antigen processing [52] Basal and IFN-c inducible STAT1 levels were not substantially altered by E 2 in either cell line ( Figure 9D-9F); however, STAT1 regulated proteins, IRF1, IRF9 and GILT were differentially modulated in E 2 -treated MCF-7 and BT-474 (Fig 9D & 9E).
In contrast to the E 2 -inhibitory effect on GAS promoter activity in the ERa + lines, E 2 noticeably enhanced GAS promoter activity in ERa 2 BCCL, MDA-MB-231 and SK-BR-3 ( Figure 9G & 9H). Furthermore, E 2 -treatment augmented expression of IRF1 and GILT in MDA-MB-231 cells, and of STAT1 in SK-BR-3 ( Figure 9I & 9J). Taken together, the results suggest that E 2 differentially modulates the IFN-c and HLA-II pathways in ERa + and ERa 2 BCCL.

Discussion
We previously reported the frequency of HLA-II positive tumor cells in ER + breast carcinomas is decreased, compared to ER 2 tumors from younger women [12]. As estrogen levels are high in breast carcinoma tissues, irrespective of age and menopausal status [41], we hypothesized a negative role for estrogen-activated ERa in HLA-II regulation in breast cancer cells. Herein, we provided experimental evidence that ERa and E 2 -activated ERa attenuate HLA-II expression in BCCL. Using paired ERa (MC2) and vector (VC5) transfected MDA-MB-231 clone 10A cells we showed: i) E 2 -treatment coordinately decreased IFN-c inducible HLA-II and CIITA in ERa + MC2 but not in ERa 2 VC5; ii) reduction of ERa by ICI or siRNA reversed the E 2 -inhibitory effect on HLA-II expression, CIITA pIV activity and transcrip- (B) Cytoplasmic CIITA and nuclear CIITA were normalized to GAPDH and P84 respectively; bar graphs represent the mean 6 SEM ratio of three independent experiments (**p,0.01). (C) CIITA mRNA was relatively quantified by real time PCR using Taqman gene expression assay. GAPDH was used as an endogenous control and the data were expressed relative to a control B cell line (RAJI). Bar graphs represent the mean 6 SEM of three replicate assays (**p,0.01). doi:10.1371/journal.pone.0087377.g004 tional activation of CIITA in MC2; iii) E 2 -activated ERa adversely affected IFN-c induced transcription as shown by GAS reporter assay and expression levels of IFN-c inducible proteins. Importantly, similar results were observed in the ERa + BCCL, MCF-7 and BT-474, in which GAS activity, STAT1 regulated genes and HLA-DR were down regulated by E 2 ; by contrast, E 2 augmented GAS activity and expression of STAT1 regulated genes in the ERa 2 BCCL, MDA-MB-231 and SK-BR-3.
Overall our data support a negative role for E 2 -ERa signaling in the regulation of HLA-II in breast cancer cells, but cell-specific differences are evident. For example, E 2 treatment attenuated HLA-DR in MCF-7 and BT-474, but not in T47D. This finding is compatible with an older study in which BCCL, cultured in E 2sufficient medium, exhibited a hierarchy of IFN-c inducible HLA-DR levels with T47D.MCF-7.BT-474 [6]. Differential HLA-II in these cells is not surprising, given that ER + BCCL, although expressing many of the same genes associated with a luminal subtype, will differ in expression of many other genes [53], which may or may not be regulated by E 2 . Multiple factors including the ratio and localization of ERa and ERb receptors, levels of coactivators and corepressors, cell surface receptors such as GPR30 and EGFR and cross-talk with other signaling pathways determine which genes are up or down regulated [54]. E 2activated ERb inhibits recruitment of ERa to ERE in target genes, thus, suppressing ERa regulated gene expression [55]. Furthermore, activation of the ERb2 isoform results in ERb2/ERa heterodimers that are targeted for proteasomal degradation [56]. It is noteworthy, then, that E 2 increases ERb in T47D but not in MCF-7 or BT-474 [57] and the ER b:a ratio in T47D is reported to be greater than in MCF-7 [53,58] thus, suggesting that cellspecific differences in ER subtypes and other receptors may underlie differential HLA expression in breast cancer.
The most convincing evidence that activated ERa modulates HLA-II and CIITA expression came from our experiments using the transfected ERa + line, MC2. Since MC2 and its ERa 2 vector control, VC5, are derived from MDA-MB-231 clone 10A, which is negative for both ERa and ERb [47], it should be a valid model to directly assess the effect of activated ERa on the HLA-II pathway. Our finding, that E 2 attenuation of HLA-II and CIITA in MC2 could be reversed by knockdown of ERa in MC2 with ICI ( Figures 3D and 4A) or siRNA (Figures 5A), provides compelling evidence that the classical ERa signaling pathway interferes with CIITA regulation. However, we were puzzled that even without adding E 2 , HLA-II and CIITA were reduced in MC2 and that knockdown of ERa by ICI and siRNA did not restore CIITA activity in MC2 to VC5 levels. Although we used phenol red free medium and E 2 -depleted FBS, there might still be a minimum level of E 2 in the culture medium, which is sufficient to activate ERa and suppress CIITA activity. Furthermore, the incomplete depletion of ERa by ICI or siRNA ( Figures 3D, 4A & 5A), may also explain why HLA-II and CIITA expression were not completely restored.
Identification of putative ERE binding sites in the proximal region of CIITA pIV ( Figure 7A) led us to explore a direct role for   to +50 with the GAS and IRF1 binding sites (shaded hexagon) and the predicted ERE (clear rectangles) were identified using online transcription factor prediction software, (http://tfbind.hgc.jp/, http://alggen.lsi.upc.es/ and http://www.cbrc.jp/index.eng.html). Site directed mutagenesis was used to perform deletion of the predicted ERE. (B) VC5 and MC2 were transfected with CIITA pIV constructs, then treated with vehicle (ethanol) or E 2 (10 29 M) and stimulated with IFN-c (100 U/ml) for 12 hours, followed by determination of luciferase activity. Bar graphs represent the mean 6 SEM of three independent experiments (**p,0.01, ***p,0.001). doi:10.1371/journal.pone.0087377.g007  ERa as a suppressor of CIITA pIV activation. Although mutagenesis of these sites did not reverse the inhibitory effect of ERa or E 2 -activated ERa on CIITA pIV activity ( Figure 7B), the experiments do not completely exclude direct ERa suppression of CIITA activity as there may be other unidentified ERE sites in either the proximal or distal region of CIITA pIV through which this effect is mediated. Alternatively, ERa may indirectly suppress CIITA pIV activation through interacting with another factor such as AP1 or NFKb that may bind CIITA pIV [21], or by interacting with factors such as CREB, SRC-1 and CBP/p300 [59] that interact with the regulatory elements of CIITA pIV and HLA-II promoters [23,60,61]. This remains to be further studied.
Although others have shown an E 2 inhibitory effect on MHC class II expression [17,19,44,45], the described mechanisms were not CIITA dependent. Tzortzakaki et al (2003) reported E 2inhibition of IFN-c inducible HLA-DR in both MCF-7 and T47D, whereby the mechanism involved sequestering the steroid receptor co-activator 1 (SRC-1) away from the HLA-DRA promoter by the E 2 -activated ER [17]. Our study did not assess cofactors, but similarly, we found E 2 -inhibition of DR expression and DRA promoter activity with only slightly reduced CIITA in MCF-7 ( Figure 1 and data not shown). However, our results for T47D conflict with theirs, as we found no E 2 inhibition of HLA-DR in this cell line. This could be due to differences in the amounts of E 2 , as their study used 3-4 log fold more than ours. Higher than physiological concentrations of E 2 were also used to show an E 2 inhibitory effect on murine MHC-II that did not involve reduced CIITA [45]. Here the E 2 inhibitory effect was mediated through reduced association of the histone acetylation transferase, CBP, with the MHC-II promoter. Since CBP is required for acetylation of histones 3 and 4 in the MHC-II promoter, this resulted in decreased transcription of MHC-II. Intriguingly, the cell lines in this study expressed both ER subtypes, which bound to the MHC I-Eb promoter, but as neither ICI nor tamoxifen reversed the E 2 inhibitory effect on MHC-II promoter, they concluded the mechanism was ER-independent. Subsequently, they showed the E 2 inhibitory effect on CBP was mediated through E 2 activation of JNK MAPK pathway [45]. Although these studies are not directly comparable to ours, they do suggest additional factors may have contributed to E 2 -inhibition of HLA-DR. However, the underlying mechanisms for E 2 -ERa inhibition of CIITA transactivation and STAT1 signaling in breast cancer are likely to be more diverse and complex.
Studies investigating deficient CIITA and MHC class II expression in various cancer cell lines have identified epigenetic modifications that result in transcriptional silencing [61,62]. These include histone deacetylation of the CIITA pIV in squamous cell carcinomas [63] and rhabdomyosarcomas [64], and hypermethylation of the CpG islands in CIITA pIV colon and gastric carcinoma lines. Hypermethylation and recruitment of dysregulated methyltransferases were hypothesized as mechanisms for defective CIITA and HLA-II expression in metastatic breast cancer [65,66], but these studies were based on a presumed breast cancer cell line MDA-MB-435. This cell line and its metastatic variants have a controversial history [67], as there is strong evidence that they originated from a melanoma cell line [68]. However, it is conceivable that epigenetic modifications are implicated in the E 2 -liganded ERa deleterious effect on CIITA pIV, as numerous epigenetic modifications have been described in breast cancer that include silencing of ERa in the MDA-MB-231 cell line and downregulation of tumor suppressor genes [69][70][71][72][73].
In our study the E 2 mediated downregulation of CIITA pIV and HLA-II expression in the ERa + BCCL appears likely due to aberrant STAT1 signaling with reduced expression of IRF1 or reduced ability to bind the CIITA promoter. Others have shown that STAT1 and IRF1 are aberrantly expressed in some ER + breast cancer tissues and cell lines [74][75][76][77] and both have tumor suppressor properties. Chan et al (2012) reported significantly decreased STAT1 in human neoplastic tissue of ER + breast tumors and showed that knocking out STAT1 in a mouse model correlated with the development of ER + PR + luminal A adenocarcinoma [77]. Intriguingly, the reduced phosphorylation of STAT1 and reduced levels of total STAT1 in MC2, compared to VC5 ( Figure 8C), whether treated or not with E2 ( Figure 8D) implies that ERa somehow negatively regulates STAT1 activation and signaling. We speculate this could occur via direct interaction of ERa with STAT1, possibly interfering with dimerization and nuclear translocation or indirectly by interfering with STAT1 promoter activation. Whatever the mechanism, aberrant STAT1 signaling is likely to result in reduced IRF1 levels and subsequently reduced CIITA activation. However, as ICI treatment of MC2 did not substantially increase STAT1 levels (data not shown), nor completely degrade ERa, more studies are required to test this concept.
A potential explanation for the dramatic reduction of CIITA pIV activity in MC2 is decreased IRF1 (Figure 8D), which is essential for IFN-c inducible CIITA transcriptional activation and HLA-II expression [50,78,79]. Furthermore, E 2 diminished IRF1 in MCF-7 and dramatically reduced its expression in BT-474, a cell line that expresses insignificant amounts of HLA-DR in the presence and absence of E 2 (Figures 1 & 9). In contrast, ERa 2 lines appear to have an intact IFN-c signaling pathway that is not inhibited by E 2 . We did not investigate mechanisms underlying E 2 -mediated increase in GAS and STAT1 activity, but others have shown a dependency on SRC kinase activity [80]. Furthermore, E 2 also activates other pathways such as MAPK and PI3K pathways that interact with the JAK-STAT1 pathway [40,81,82].
In conclusion, our results show that HLA-II expression is regulated differently by estrogen in ER 2 and ER + breast cancer cells. To our knowledge this report is the first to show that activation of ERa by its ligand E 2 , results in downregulation of CIITA pIV activity. Although the mechanism is not fully elucidated, the data suggest that the dysregulation occurs at the level of STAT1 activation. Such a mechanism would explain the HLA-DR negative tumor cells in breast carcinomas despite infiltrating T-cells and high levels of IFN-c and has further implications for tumor immune escape.

Real-time RT-PCR
Total RNA, extracted using TRIzol Reagent (Invitrogen) and treated with AmbionH TURBO TM DNase to remove contaminating DNA, was quantified using NanoDrop (Thermo Scientific). The High Capacity cDNA Reverse Transcription kit (Applied Biosystems) was used for cDNA synthesis according to the manufacturer's protocol. Real time PCR was performed using TaqManH Probe-Based Gene Expression Analysis kit for CIITA (Hs00172106_m1) and GAPDH (Hs99999905_m1) following the manufacturer's recommendations. Quantification was performed by the comparative threshold cycle (DD CT ) method and normalized to GAPDH using StepOnePlus TM (Applied Biosystems). A control sample without RNA and a reference sample (RAJI, B cell line) were included in each experiment.

siRNA Transfection
Cells, plated in a 6-well plate at 3610 5 cells/well for 24 hours, were transfected with either 25 nM ON-TARGET plus SMART pool siRNA for ESR1 or non-targeting siRNA (Dharmacon, USA) using 4 ml DharmaFECT4 transfection reagent (Dharmacon, USA) per well according to the manufacturer's protocol. Fortyeight hours later, the cells were treated with E 2 10 29 M or vehicle control (ethanol) and stimulated with IFN-c, 100 units/ml, for 4 or 24 hours for mRNA and protein expression, respectively.

Reporter gene assays
The CIITA promoter IV firefly luciferase construct [79] and the 8 X GAS firefly luciferase construct [86] were kind gifts from Dr. Jenny Ting and Dr. Eleanor N. Fish, respectively. Transfection conditions were optimized using Fugene HD (Roche) transfection reagent according to the manufacturer's protocol: briefly a master mix was prepared by diluting the appropriate plasmid with Opti-MEM (Gibco) to a concentration of 0.02 mg/ml; Fugene HD was added to the same mixture in the ratio of 7:2 (Fugene HD in ml:Plasmid DNA in mg) and left for 20 minutes at ambient temperature. Cells, plated in a 96-well plate at 2610 4 cells/well for 24 hours, at 37uC were transfected with 5 ml of this mixture and incubated for an additional 24 hours. The medium was then replaced with medium containing the appropriate treatments and incubated for 12 hours for CIITA pIV or 6 hours for 8 X GAS constructs. Transfection efficiency was estimated by co-transfecting the cells with SV-40 Renilla luciferase or green fluorescent protein (GFP). Luciferase activity was measured using the dual luciferase assay system (Promega) and a 96-well luminometer (Fluoroskan Ascent Fl, Labsystems).

Generation of CIITA pIV deletion constructs
Different sets of deletion mutants of P-346/+50 CIITA pIV were generated by site-directed mutagenesis using QuikChange Lightning Site-Directed Mutagenesis Kit (Stratagene) according to the manufacturer's instructions. Mutagen primers (Table 1) were designed with Agilent's web-based QuikChange Primer Design Program Sequences. Sequences, deleted from the original template, are bolded and underlined. All deletions were confirmed by sequencing.

Statistics
Statistical analysis was performed using Microsoft excel 2010 software. One-way analysis of variance (ANOVA) and Tukey post hoc tests were used for comparisons within a group. The student t-test was used for comparing two different treatments for one cell. All tests were two-sided and p,0.05 was considered significant.