High-Fat Diet Induces Unexpected Fatal Uterine Infections in Mice with aP2-Cre-mediated Deletion of Estrogen Receptor Alpha

Estrogen receptor alpha (ERα) is a major regulator of metabolic processes in obesity. In this study we aimed to define the relevance of adipose tissue ERα during high-fat diet (HFD)-induced obesity using female aP2-Cre−/+/ERαfl/fl mice (atERαKO). HFD did not affect body weight or glucose metabolism in atERαKO- compared to control mice. Surprisingly, HFD feeding markedly increased mortality in atERαKO mice associated with a destructive bacterial infection of the uterus driven by commensal microbes, an alteration likely explaining the absence of a metabolic phenotype in HFD-fed atERαKO mice. In order to identify a mechanism of the exaggerated uterine infection in HFD-fed atERαKO mice, a marked reduction of uterine M2-macrophages was detected, a cell type relevant for anti-microbial defence. In parallel, atERαKO mice exhibited elevated circulating estradiol (E2) acting on E2-responsive tissue/cells such as macrophages. Accompanying cell culture experiments showed that despite E2 co-administration stearic acid (C18:0), a fatty acid elevated in plasma from HFD-fed atERαKO mice, blocks M2-polarization, a process known to be enhanced by E2. In this study we demonstrate an unexpected phenotype in HFD-fed atERαKO involving severe uterine bacterial infections likely resulting from a previously unknown negative interference between dietary FAs and ERα-signaling during anti-microbial defence.


Results
No metabolic phenotype but increased mortality in HFD-fed atERαKO mice. Metabolic baseline characterization of 6 weeks old female atERα KO mice resulted in the expected metabolic phenotype with increased body weight (BW) and decreased energy expenditure (Table 1). However, in 15 weeks old wt-and atERα KO mice on control diet (CD) BW-differences disappeared (Fig. 1A). More importantly, HFD-induced increase of BW after 14 weeks feeding did not differ between wt-and atERα KO mice (Fig. 1A,B). Along this line, no differences could be detected for glucose tolerance and insulin sensitivity between wt-and atERα KO mice after HFD-feeding (Fig. 1C,D). Neither energy expenditure nor locomotor activity, assessed by metabolic cage experiments, showed any alteration in the HFD-fed atERα KO group when compared to control (Fig. 1E,F). However, mutant mice displayed an increase in food intake (Fig. 1G). These data appeared highly controversial to recently published results about the role of ERα in adipose tissue pointing towards a protective action of ERα against BW gain and HFD-mediated glucose intolerance 7 . Expression analysis of metabolically relevant genes in white adipose tissue showed only modest effects of ERα , reaching statistical significance only for the reduction of ATGL expression in the absence of ERα (Fig. 1H). These data corroborate our previous findings pointing to a pivotal role of ERα in lipolysis 12 . Surprisingly, after the onset of HFD-feeding we observed a markedly increased mortality in HFD-fed atERα KO mice compared to atERα KO mice on CD and to wt-mice (Fig. 1I). Together these data suggest that HFD-feeding induces a fatal pathology in atERα KO which likely impacts on physiological metabolic processes.
HFD mediates fatal uterine infections in atERαKO mice. Autopsies of wt-and atERα KO mice revealed massive uterine fluid accumulation in CD-fed atERα KO mice compared to wt-mice as previously described ( Fig. 2A) 13 . More importantly, HFD-feeding aggravated these uterine processes in atERα KO mice leading to a destructive, pus-filled swelling of the uterus and uterine appendages in line with a severe bacterial infection in all HFD-fed atERα KO mice ( Fig. 2A, right panels). In consonance, microbiological analysis of the uterine fluid of HFD-fed atERα KO mice showed the presence of commensal microbes including Enterococcus sp. and bacterial DNA for E.coli (data not shown). These results were confirmed by histological analysis (Fig. 2B) demonstrating massive cellular infiltration of the uterine wall in HFD-fed atERα KO accompanied by destruction of the intramural glandular and epithelial structure (Fig. 2B, right panels). The grade of uterine inflammation as determined by microscopic analysis (for detailed protocol see methods) was significantly higher in atERα KO mice fed HFD vs. CD (Fig. 2C). CD-or HFD-fed wt-mice showed no evidence for uterine inflammation (data not shown). Bacterial uterine infections in HFD-fed atERα KO mice were characterized by pronounced neutrophil wt atERαKO accumulation associated with a low number of macrophages in the uterine wall (Fig. 2D,E). Neutrophils initially recruited to the site of bacterial infection are usually cleared by macrophages to initiate physiological resolution and to prevent exaggeration of inflammation, a process called efferocytosis 14 . Regular macrophage-mediated Intraperitoneal glucose-(C) and insulin-(D) tolerance test (n = 9-10). No differences occurred in energy expenditure (E) and in locomotor activity (F) between the genotypes, while atERα KO mice showed an increase in food intake (G) (n = 10). (H) Gene expression analyses in white adipose tissue of wt and atERα KO mice (n = 5-8). (I) Survival curves of atERα KO-and wt-mice on CD and HFD. Onset of HFD-feeding at 42 days of age is indicated in the graph. n.s. = non-significant (P > 0.05, 2-way-ANOVA (Bonferroni-posttest) and 2-way-ANOVA with repeated measures (Bonferroni-posttest)). * P < 0.05 vs. wt (two-tailed t-test). neutrophil clearance depends on specific neutrophil marks to engage particularly M2-macrophages 14,15 . To understand the neutrophil-macrophage interaction resulting in prominent uterine neutrophil accumulation in HFD-fed atERα KO mice in our study, the regulation of neutrophil signals in uteri was first studied. So called "Don't eat me" signals" on viable neutrophils such as CD47 and PAI-1 14 did not differ in uteri from CD-and HFD-fed atERα KO mice (Fig. 2F). However, mRNA expression analysis of M1-and M2 macrophage markers in uteri from atERα KO mice revealed a statistical non-significant increase of the M1-marker Nos2, but more importantly a highly significant reduction of the M2-marker Arg1 by HFD (Fig. 2G). These data suggest that HFD does not directly affect signals on neutrophils but more likely augments neutrophil presence at the site of infection by an indirect shift in macrophage polarization towards pro-inflammatory conditions and resulting in a significant repression of M2 macrophages. A similar trend could be registered in the expression of Arg1 in white adipose tissue, however, no alteration of Nos2 expression could be detected (Fig. 2H). These data suggest that the macrophage phenotype may be regulated in a tissue-specific manner in our model, and that the bacterial environment is, at least in part, required for the enhanced presence of M1 macrophages in the uteri.
To sum up, we show that HFD-feeding promotes a severe bacterial uterine infection in atERα KO mice likely causing higher mortality. Uterine bacterial infection in HFD-fed atERα KO mice is characterized by an excess of uterine neutrophil and M1/M2-macrophage imbalance.
HFD-feeding impairs amplified ERα signaling in atERαKO mice. In order to understand the underlying mechanism leading to fatal bacterial infections in atERα KO mice through "simple" HFD-feeding, additional plasma and expression analysis in wt-and atERα KO mice were performed. Antonson and colleagues recently found that aP2-Cre-mediated deletion of ERα not only results in ERα depletion in adipose tissue but also in the hypothalamus 13 . In agreement with these findings, ERα mRNA expression was significantly reduced in white adipose tissue and hypothalamus but not in the uterus (Fig. 3A). According to the study of Antonson and colleagues, we found that atERα KO female mice are infertile and have no proper estrous cycle compared to wt, as demonstrated by evaluation of vaginal smear (Fig. S1). E2 levels were also significantly higher in atERα KO-compared to wt-mice (CD: 40.2 ± 4.6 vs. 6.1 ± 2.0 pg/ml; HFD: 45.9 ± 5.7 vs. 19.7 ± 3.4 pg/ml; P < 0.001) suggesting enhanced estrogen signaling in atERα KO mice on CD and HFD (Fig. 3B). Accordingly, E2/ERα -target genes were markedly upregulated in CD-fed atERα KO mice, in organs still expressing ERα such as uterus and spleen (Fig. 3C, white bars). Of note was that under HFD ERα -target gene induction in uterus and spleen did no longer reach statistical significance (Fig. 3C, black bars) despite high E2-levels implicating impairment of ERα -signaling by HFD. To identify potential mediators in HFD responsible for disturbed ERα -signaling plasma lipid analysis using HPLC/ triplequad mass spectrometry was performed. As shown in Fig. 3D, HFD feeding resulted in the regulation of distinct plasma fatty acids (FA) among which C18:0 (stearic acid), C18:2n6 (linoleic acid), and C20:4n6 (arachidonic acid) were significantly upregulated, and only C18:1n9 (oleic acid) was downregulated. Taken together, atERα KO mice demonstrated reduced hypothalamic ERα expression levels associated with increased circulating E2-levels leading to enhanced ERα signaling in ERα expressing tissue. HFD-feeding induced a distinct plasma FA-profile accompanied by a blockade of ERα signaling.
C18:0 (stearic acid) impairs E2-mediated macrophage polarization and phagocytotic activity. C18:0 (stearic acid) induces pro-inflammatory processes in macrophages 16 whereas E2 exerts anti-inflammatory actions in these cells 17 . Plasma level of C18:0 showed the strongest up-regulation by HFD feeding in our model. To answer whether the HFD-induced rise in plasma FAs accounts for M1/M2-macrophage imbalance, disturbed neutrophil depletion and exaggerated bacterial inflammation in atERα KO mice, the effect of C18:0 on macrophage M1/M2 polarization in the presence of high E2 was characterized. As previously mentioned, depletion of CD206-positive macrophages (alternatively activated/M2 polarized) can lead to aggravation of inflammatory processes characterized by accumulation of neutrophils at the inflammatory sites. Stimulation of THP-1 macrophages with C18:0 on an E2-background significantly induced CCR7 mRNA levels, a marker for M1 macrophages, and significantly suppressed the expression of CD206, independently from the presence of lipopolysaccharide (LPS) ( and from 99.26 to 78.02 (only E2 co-treatment) due to C18:0 action. CD11b, an activation marker and an important integrin for immune cell adhesion, was not significantly regulated (Fig. 4C,D), but shows a slight tendency towards a decrease of CD11b induced by C18:0. Stimulation with the FA leads to a decrease of the average MFI from 1076.33 to 719.36 (E2 and LPS co-treatment) and from 976.56 to 724.82 (only E2 co-treatment). These data show that C18:0 is capable of blocking E2-mediated M2-macrophage polarization, a process important for E2's anti-inflammatory capacity. Furthermore the functionality of macrophages in the presence of C18:0 was tested. The addition of the fatty acid negatively influenced phagocytotic activity of THP-1 macrophages (Fig. 4F). Similar effects on the functionality of macrophages were observed in mice lacking ERα in myeloid cells 9 .

Discussion
ERα exerts multiple metabolic actions in adipose tissue 7 . To prove that adipose ERα plays a crucial role for whole body metabolism, we investigated the role of ERα in a tissue specific knock-out model using aP2-Cre −/+ /ERα fl/fl mice. A similar approach has previously been conducted by Davis and colleagues. This demonstrated that, in adiponectin promoter driven-Cre transgenic mice crossed with floxed ERα mice, a lack of ERα in adipose tissue results in an increase of gonadal adipose tissue mass, enhanced adipose tissue fibrosis and inflammation, and in full-length blot was provided as supplemental data). # P < 0.05, ## P < 0.01, ### P < 0.001 vs. E2+ BSA; ** P < 0.01, *** glucose intolerance in male mice 8 . Due to an unexpected mortality increase in HFD-fed atERα KO mice, based on a severe bacterial infection, interpretation of the metabolic data in our study are limited which prompted us to focus on the etiology of the uterine infection in our model. The observed discrepancies between Davis and our data likely result from differences in the utilized Cre lines.
In line with Mullican and colleagues and other investigators, we observed that the aP2-Cre model lacks specificity for adipose tissue, reducing ERα mRNA levels also in the hypothalamus 22 . Hypothalamic ERα appears to be important for regular uterine development. Antonson and colleagues recently showed that female aP2-Cre −/+ / ERα fl/fl fed a CD display a reproductive phenotype involving infertility, hyperplasia and hydrometra of the uterus 13 . Moreover, other knock-out models of ERα in the brain such as the CamKIIa-Cre-mediated model, also display an enlarged and fluid-filled uterus 23 . Uterine pathologies observed in these models, including this, presumably involve a lack of a central ERα -dependent negative feedback resulting in elevated E2-levels subsequently affecting uterine development. Furthermore, elevated E2-levels likely conceal some of the metabolic effects usually expected in mice lacking ERα exclusively in adipose tissue.
The present study, however, is the first report of increased mortality in aP2-Cre −/+ /ERα fl/fl mice. In contrast to previous studies, a severe bacterial, E. coli/Enterococcus sp. positive, uterine infection was observed. Since HFD-feeding is the obvious difference between Antonson's 13 and this study, analysis was conducted to see whether HFD content, in particular FAs, act as pathogenic mediators of disturbed bacterial defense. HFD is well known to induce pro-inflammatory processes on the tissue-and systems level 24,25 . Indeed, in contrast to atERα KO mice on CD, a severe inflammatory response was identified in uterine tissue from mice fed a HFD characterized by massive neutrophil accumulation and a reduced number of uterine M2-macrophages. Defects of neutrophil clearance have been linked to severe inflammation including sepsis 26 . Neutrophil clearance from the infectious site is a complex process predominantly orchestrated by neutrophils themselves and macrophages 14,26,27 . Anti-inflammatory (M2 polarized) macrophages seem to especially be required for neutrophil clearance and play a crucial role in the resolution of inflammation 15,28 . It was hypothesized that the HFD-induced atERα KO mice phenotype is mainly a result of HFD-mediated blockade of anti-inflammatory E2-ERα -actions. E2-dependent ERα activation usually induces M2-macrophage polarization, and thus attenuating inflammatory responses 9,17 . Under a normal diet, elevated E2 levels in atERα KO mice lead to enhanced E2-signaling and likely to containment of the inflammatory response. In contrast under HFD, it was shown that C18:0, most strongly up-regulated in plasma from HFD-fed mice, inhibits E2-induced ERα -activation significantly blocking E2-stimulated M2-macrophage polarization, and probably exaggerating inflammation. In consonance, C18:0 has been already described to possess pro-inflammatory effects on macrophages 16 . In summary these data suggest that the inflammatory control by E2 is abolished by C18:0 triggering the multifactorial process that induces fatal bacterial infections in this model. Depending on the used ap2-Cre strain, it is known that the targeted gene can be also affected in macrophages 29,30 opening the possibility of reduced ERα expression. However, this seems not to be the case in the aP2-Cre strain used in our study. Using the identical strain, Mayoral et al. recently showed that macrophages in this strain do not express Cre-recombinase 31 . In line with this findings, in CD-fed aP2-Cre −/+ /ERα fl/fl mice we still observed a robust induction of the ERα target gene CTSD by E2 in the spleen, a gene highly expressed in macrophages. These data suggest an existing E2 response in macrophages in CD-fed aP2-Cre −/+ /ERα fl/fl mice rendering it unlikely that macrophages lack ERα .
Finally, the molecular mechanism of C18:0's repressive action on E2-ERα -signaling was investigated. Stearoylation of proteins has been recently identified as a new determinant of protein function 32 . Senyilmaz and colleagues demonstrated that C18:0 stearoylates the human transferrin receptor 1 (TFR1), a protein important for mitochondrial function 32 . Other reports had previously shown the existence of an interaction between ERα and long-chain fatty acids 33 . Stearic acid was described to inhibit E2-effects by downregulation of ERα in hypothalamic tissue 33 , while palmitic acid (C16:0) promotes extra-nuclear action of ERα and prevents its degradation through acylation 21 . In the work of La Rosa and colleagues, the lack of ERα -palmitoylation at cysteine (C) residue 447 was shown to repress its transcriptional activity in HEK293 cells 21 . Contrastingly, we observed a significant decrease in E2-induced ERE-activity with C18:0 in differentiated THP-1 macrophages, also likely mediated via C447. HPLC/triplequad MS-based profiling of FAs bound to immunoprecipitated ERα demonstrated a significant protein-FA interaction between ERα , its protein co-factors and C18:1n9 (oleic acid), the main metabolite of C18:0 after enzymatic transformation through the stearoyl-CoA desaturase-1 (SCD-1) 34 . Predominance of cellular C18:1n9 after C18:0 stimulation is not surprising, since it has long been known that desaturation of C18:0 is greater than compared to other saturated FAs 35 . Up until now, the following mechanism can be proposed: challenge of macrophages with C18:0 leads to repression of ERα -transcriptional activity potentially involving the direct interaction between FAs and the of ERα transcriptional protein complex. The relative contribution of the direct FA -ERα -complex interaction to the observed phenotype in HFD-fed atERα KO mice requires further experiments.
The importance of these data is further supported by clinical observations in female dogs. In sexually mature female dogs, pyometra is a common disease entity 36 . Up until now, the pathogenesis has been linked to increased estradiol levels and prolonged progesterone stimulation 36 . In light of our findings, it might be worth to investigate the relevance of dietary FAs in conjunction with high E2/progesterone levels for the pathogenesis of canine pyometra. Along this line, these findings might also be relevant for bacterial uterine infections in humans when conditions occur in which frequent HFD consumption is combined with high plasma E2-levels, e.g. during pharmacological application.
This study identifies an unexpected phenotype of atERα KO mice fed HFD characterized by increased mortality likely due to fatal bacterial uterine infections driven by commensal microbes. This phenotype points towards a previously unknown interaction of ERα signaling and FAs in macrophages. In particular, it was shown that C18:0 provided by HFD impairs E2-ERα action in macrophages with a concomitant perturbation of M2-macrophage polarization usually required for regular neutrophil-mediated bacterial defense. In accompanying in-vitro experiments we identified a direct interaction between FAs and the intracellular ERα transcriptional complex associated with a C18:0-mediated inhibition of ERα transcriptional activity and E2-dependent regulation of genes involved in M2 macrophage polarization. These data suggests a new mechanism of how FAs are capable of disturbing macrophage E2-ERα -signaling thereby aggravating commensal bacterial infections at predilection sites.
Therefore, it is suggested a new mechanism of dietary FA interference with ERα resulting in the impairment of E2-mediated anti-inflammatory actions and the promotion of severe bacterial uterine infections.

Materials and Methods
Mice. Female adipose tissue-specific (aP2) -mice were generated by crossing B6.Cg-Tg(Fabp4-cre)1Rev/J mice with B6.129X1-Esr1tm1Gust (ERα fl/fl ) mice, kindly provided by J.-A. Gustafsson (University of Houston, TX, USA). ERα fl/fl /aP2-Cre −/+ (atERα KO) and control littermates ERα fl/fl /aP2-Cre −/− (wt) were used for all experiments. Animals were maintained in a temperature-controlled facility with a 12 h dark/light cycle. At the age of 6 weeks, a group of atERα KO and wt-mice (n = 24), were challenged for 14 weeks with a HFD (60% kcal from fat, Brogaarden, Lynge, Denmark). The HFD groups and the CD were both fed ad-libitum. Metabolic phenotyping was performed with a metabolic cage system (TSE-Systems GmbH, Bad Homburg, Germany), based on indirect calorimetry as described before 12 . Body composition was assessed by nuclear magnetic resonance imaging (Echo MRI mouse, Echo Medical Systems, Houston, USA). Body temperature was measured rectally 3 to 5 days in row and mean values were calculated. Wt-mice were sacrificed during estrous phase of the cycle. atERα KO mice did not present a physiological estrous cycle, and were sacrificed independent of (non-) cycle phases; their vaginal smear was collected and analyzed to exclude a proestrous analogous stage. Organs were collected and frozen in liquid nitrogen. All animal experiments were approved by the Landesamt für Gesundheit und Soziales (LaGeSo, Berlin, Germany) and were conducted in alliance with the German Law on the Protection of Animals.
Bone-marrow derived macrophages (BMDM) were isolated as described before 37 , using 10% L929-conditioned medium for macrophage differentiation. RPMI with 10% cs-FBS and 1% Pen-Strep was used to differentiate and cultivate cells. For stimulation, phenol red-free medium was used with addition of E2, BSA and C18:0 in the same concentrations as done with THP-1 cells.
Gene expression analysis. RNA from cultured THP-1 cells, and BMDM, from adipose tissue, uteri and hypothalamus was isolated with RNeasy Mini Kit (Qiagen) following the manufacturer's instructions. For real-time PCR analysis, RNA was reverse transcribed and relative mRNA expression was normalized to 18 S (animal tissues), β 2-microglobulin (BMDM) or to β -Actin (THP-1 cells). Primer sequences will be provided on demand.
17β-estradiol measurement in plasma. Plasma samples of atERα KO and wt mice were obtained after final blood collection. The measurement of 17β -estradiol (E2) concentrations was performed by a radioimmunoassay set up as a sequential assay as previously described 38 . Prior to radioimmunoassay, blood plasma was extracted twice with toluene and the pooled extracts were evaporated to dryness and re-dissolved. The antiserum used was directed against E2-6-carboximethyloxim (CMO)-BSA and exhibited the following cross-reactions: estrone, 1.3%; estriol, 0.7%; all tested non-phenolic steroids < 0.01%. The minimum detectable concentration was 5 pg/ml; intra-and interassay coefficient of variation (CV) were 7.1 and 17.6%, respectively. Fatty acid profiling in plasma samples and in protein precipitates. Fatty acids were analysed in plasma samples from atERα KO mice on CD and HFD with a triplequad mass spectrometer coupled HLPC method, as described before 39 . Briefly, after hydrolysis, samples were diluted 1:10 in methanol containing internal standards. Samples were injected and separated by a reverse-phased column (Phenomenex Kinetex-C18 column 2.6 μ m, 2.1 × 150 mm) with a solvent gradient containing aqueous formic acid (0.1%) and acetonitrile. After separation of components, fatty acid's identity was analysed by an Agilent 6460 triplequad mass spectrometer with electrospray ionisation.
Histology. After sacrificing mice, uteri were partly fixed with 4% formalin, and paraffin-embedded. Sections were deparaffinised with xylene and rehydrated with a gradient of alcoholic solutions. Antigen heat retrieval with citric acid and incubation with goat serum was performed with slides prior to hematoxylin and eosin (H&E), monoclonal rat anti-mouse Mac3 antibody (clone M3/84, BD Pharmingen, USA), and monoclonal rat anti-mouse Ly6G antibody (clone 1A8, Biolegend, USA) staining. For evaluation of the grade of inflammation, and for Mac3 and Ly6G expression a semi-quantitative scoring system was applied, using a standardized procedure under observation of the equivalent anatomical structures of the uteri. Flow cytometry. For FACS analysis, THP-1 macrophages were stained with PE conjugated anti-CD11b (BioLegend Inc. San Diego, USA) or PE conjugated anti-CD209 (BioLegend), and 7-AAD (Cell Viability Solution, BioLegend). FACS analysis was performed on a FACS Calibur flow cytometer (BD Biosciences, Heidelberg, Germany). As cells were only single stained, auto fluorescence controls were used to set negatives. The expression of the constitutively expressed adhesion protein CD11b and the M2-marker CD209 was quantified as mean fluorescence intensity (MFI). Relative up-or downregulation of CD11b-and CD209-expression was calculated by the following equation: MFI (CD11b (or CD209) of cells)/MFI (CD11b (or CD209) of E2+ LPS treated cells). 7-AAD-staining was used to determine and exclude dead cells.
Phagocytosis assay and immunofluorescence staining. For testing of phagocytosis activity, phagocytosis Assay Kit (Cayman) was used and performed in accordance to the manufacturer's instructions. Briefly, FITC-labelled beads were added to the supernatant and THP-1 macrophages were processed analogously as for immunostaining experiments. For immunostaining cells were fixed with formalin and blocked with goat serum. Afterwards, cells were incubated with PE conjugated anti-CD209 antibody (Biolegend), washed, stained with DAPI (1:1000, Thermo Scientific, blue) and mounted with mounting medium (Dako). Three fluorescent pictures were captured randomly from every sample (n = 6 per group) with an all-in-one fluorescence microscope (BZ-9000E, Keyence). Finally, representative pictures were selected.
Luciferase reporter assay. After plating and differentiating THP-1 cells as described above, cells were transiently transfected with hERα -pSG5 (kindly provided by P. Chambon, Institut Clinique de la Souris, Illkirch Cedex, France), or with pcDNA flag ERα wt or pcDNA flag ERα C447A (kindly provided by F. Acconcia, University Roma Tre, Rome, Italy), and with pERE-TkGL3 (kindly provided by P.J. Kushner, Metabolic Research Unit and Diabetes Center, University of California, San Francisco, USA), and pRL-CMV (Promega), a renilla luciferase vector. Transfection was performed with jetPEI ® -Macrophage transfection reagent (Polyplus-transfection) according to the manufacturer's instructions. Prior to stimulation, cells were starved for a maximum of 10 h. After harvesting, cell lysates were used to measure luciferase activity with the dual-luciferase reporter assay system (Promega).
Statistical analysis. All experiments were repeated at least three times and n-numbers are indicated for each experiment. Statistical analysis was performed with GraphPad Prism Software and statistical significance was assumed at p < 0.05. Comparison of mean values was evaluated by two-way ANOVA (Bonferroni posttest), two-way ANOVA with repeated measures (Bonferroni posttest), or unpaired t-tests, as appropriate.