Identification of PARP-1, Histone H1 and SIRT-1 as New Regulators of Breast Cancer-Related Aromatase Promoter I.3/II

Paracrine interactions between malignant estrogen receptor positive (ER+) breast cancer cells and breast adipose fibroblasts (BAFs) stimulate estrogen biosynthesis by aromatase in BAFs. In breast cancer, mainly the cAMP-responsive promoter I.3/II-region mediates excessive aromatase expression. A rare single nucleotide variant (SNV) in this promoter region, which caused 70% reduction in promoter activity, was utilized for the identification of novel regulators of aromatase expression. To this end, normal and mutant promoter activities were measured in luciferase reporter gene assays. DNA-binding proteins were captured by DNA-affinity and identified by mass spectrometry. The DNA binding of proteins was analyzed using electrophoretic mobility shift assays, immunoprecipitation-based in vitro binding assays and by chromatin immunoprecipitation in BAFs in vivo. Protein expression and parylation were analyzed by western blotting. Aromatase activities and RNA-expression were measured in BAFs. Functional consequences of poly (ADP-ribose) polymerase-1 (PARP-1) knock-out, rescue or overexpression, respectively, were analyzed in murine embryonic fibroblasts (MEFs) and the 3T3-L1 cell model. In summary, PARP-1 and histone H1 (H1) were identified as critical regulators of aromatase expression. PARP-1-binding to the SNV-region was crucial for aromatase promoter activation. PARP-1 parylated H1 and competed with H1 for DNA-binding, thereby inhibiting its gene silencing action. In MEFs (PARP-1 knock-out and wild-type) and BAFs, PARP-1-mediated induction of the aromatase promoter showed bi-phasic dose responses in overexpression and inhibitor experiments, respectively. The HDAC-inhibitors butyrate, panobinostat and selisistat enhanced promoter I.3/II-mediated gene expression dependent on PARP-1-activity. Forskolin stimulation of BAFs increased promoter I.3/II-occupancy by PARP-1, whereas SIRT-1 competed with PARP-1 for DNA binding but independently activated the promoter I.3/II. Consistently, the inhibition of both PARP-1 and SIRT-1 increased the NAD+/NADH-ratio in BAFs. This suggests that cellular NAD+/NADH ratios control the complex interactions of PARP-1, H1 and SIRT-1 and regulate the interplay of parylation and acetylation/de-acetylation events with low NAD+/NADH ratios (reverse Warburg effect), promoting PARP-1 activation and estrogen synthesis in BAFs. Therefore, PARP-1 inhibitors could be useful in the treatment of estrogen-dependent breast cancers.


Aromatase Activity Testing
The in vivo evaluation of aromatase function in BAFs was performed by the tritium water release assay in 24-well plates as described previously [16,17]. All conditions were tested in triplicate per experiment.

NAD + /NADH Quantification
NAD + /NADH in BAFs was quantified using the colorimetric EZScreen TM NAD + /NADH Assay Kit (BioVision; Milpitas, CA, USA) according to manufacturer's protocol. The NAD + and NADH extraction protocol was adapted from a published method [18].

Preparation of Soluble Nuclear Extracts
Preparation of soluble nuclear extracts was based on a method published by Wilde et al. [19]. The protein concentration was quantified by the Bradford method [20].

Electrophoretic Mobility Shift Assays
For electrophoretic mobility shift assays (EMSA), 10 µg soluble nuclear extract protein per condition was incubated in the presence of binding buffer (50 mM Tris/HCl pH 7.5, 0.1 M NaCl, 0.1 mM EDTA, 5 mM 2-mercaptoethanol) for 30 min at 37 • C with various double-stranded probes (Appendix A, Table A1)-25 pmol of a Cy5-labeled normal sequence probe (either alone or in the presence of a 20-fold molar excess of an unlabeled normal sequence probe (competitor)), or 25 pmol of a Cy5-labeled SNV-containing probe or Cy5-labeled quadruple mutation probe (complete destruction of putative binding-sites). For antibody competition, 2 µL of anti-PARP-1 antibodies (Appendix A, Table A2) were incubated for 30 min before the addition of probes. Separations were carried out on a 6% non-denaturing acrylamide gel at 4 • C (18 cm, 300 V, and 70 min; [26]). The wet gels were directly scanned on a Fuji FLA-3000 imaging system and quantified using the AIDA Software (Raytest, Straubenhardt, Germany).

Immunoprecipitation-Based DNA-Binding Assay
An immunoprecipitation-based DNA-binding assay protocol was developed for PARP-1 and histone H1, respectively. Soluble nuclear extract proteins (50 µg) were pre-incubated with 2 µL pre-cleared (in soluble nuclear extract buffer) Protein G-Sepharose 4 Fast Flow (GE Healthcare, Freiburg, Germany) at 4 • C in a rotator to eliminate proteins binding non-specifically to protein G. After centrifugation of the pre-incubated samples (20 s, 12,000× g, 4 • C), the supernatants were transferred into new tubes and incubated with either 1 µL anti-histone H1 or 1 µL anti-PARP-1 antibodies (Appendix A, Table A2) for 24 h at 4 • C under constant rotation. After the antibody incubation, 20 µL pre-cleared Protein G-Sepharose 4 Fast Flow was added and incubated for 4 h at 4 • C under constant rotation. For the final isolation of anti-histone H1 or anti-PARP-1 immunoprecipitates, respectively, the samples were washed three times in a three-fold volume of DNA-binding buffer C (20 mM HEPES pH 7.9, 1 mM EDTA, 1 mM EGTA, 1 mM DTT, 1 mM PMSF) with centrifugation after each step (20 s, 12,000× g, 4 • C). The final immunoprecipitates were resuspended in 8 µL buffer C. The DNA-binding reaction was a modification of the sample preparation protocol for electrophoretic mobility shift assays, as described by Taylor et al. [26]. Immunoprecipitates in buffer C (8 µL) were mixed with 1.7 µL 10-fold binding buffer (500 mM Tris/HCl pH 7.5, 1 M NaCl, 1 mM EDTA, 50 mM β-mercaptoethanol) and were incubated with 50 pmol fluorescence-labeled oligonucleotides with normal sequence (either alone or in the presence of a 26-fold molar excess of an unlabeled normal sequence competitor) or sequence with SNV (Appendix A, Table A1). For normal binding reactions, the premix was added to 2 µL Cy5-labeled double-stranded oligonucleotides (25 pmol/µL) in 5.3 µL water. For binding reactions in the presence of a competitor, 8 µL immunoprecipitate, 2 µL fluorescence-labeled double-stranded oligonucleotides and 5.3 µL unlabeled competitor double-stranded oligonucleotides (250 pmol/µL; same sequences as fluorescence-labeled oligonucleotides) were mixed. After incubation for 30 min with rotation at room temperature in the dark, the samples were washed three times in a three-fold sample volume of wash buffer (50% buffer C, 10% 10-fold binding buffer, 40% water), followed by a 20 s centrifugation at 12,000× g at room temperature. Finally, the oligonucleotide-bound immunoprecipitates were resuspended in 17 µL wash buffer and transferred to a well of a 96-well plate for fluorescence measurement (excitation 600 nm; emission 670 nm, cut off 630 nm). As a control, the unspecific binding of fluorescent oligonucleotides to Protein G-Sepharose 4 Fast Flow beads treated as described above in the absence of antibodies was analyzed, resulting in negligible fluorescence signals. All conditions were tested in triplicate per experiment.

Transfection and Luciferase Reporter Gene Assays
Half-confluent 3T3-L1 preadipocytes or PARP-1 knock-out or wt MEFs in 24-well plates were transfected using Roti-Fect Plus (Carl Roth, Karlsruhe, Germany) according to the manufacturer's instructions and stimulated 24 h later. To quantify promoter activities, pGL3-basic firefly luciferase plasmid containing aromatase promoter I.3/II with either a normal or SNV sequence were used (Appendix A, Table A3). For PARP-1 overexpression, pSG9M-PARP-1 plasmid was used [28]. Furthermore, we utilized pRL-SV40 plasmid-expressing Renilla luciferase for normalization. All conditions were tested in triplicate per experiment. The luciferase activity measurement was described by Hampf and Gossen [29].

Chromatin Immunoprecipitation (ChIP)
The chromatin immunoprecipitation protocol is a modified version of that published by Weiske and Huber [30]. BAFs from four 10 cm dishes per condition were used per experiment. For each Protein G-based immunoprecipitation, 1 µg anti-histone H1, anti-PARP-1 or anti-SIRT-1 antibody was used per 5 µg of cross-linked DNA, respectively (Appendix A, Table A2). PCR was carried out using Paq5000 DNA-polymerase (Agilent Technologies, Waldbronn, Germany) in a Veriti ® 96-Well Thermal Cycler (Applied Biosystems, Darmstadt, Germany). Two SNV region-spanning primer sets were used (Supplementary Materials, Table S1). The products were analyzed on 12% polyacrylamide gels stained with ethidium bromide, as described previously [19].

Quantification of Aromatase mRNA-Expression in BAFs
The quantification of full-length aromatase mRNA-expression and utilization of promoters I. 3 and II, respectively, was performed by quantitative real-time PCR (qRT-PCR), as described in detail by Wilde et al. [19]. Primer sequences are given in the Supplementary Materials, Table S1. All conditions were tested in triplicate per experiment. The evaluation of the PCR results was done by the ∆∆C T -method [31].

Quantification of Aromatase Promoter-Utilization in Transfected PARP-1 Wild-Type and Knock-Out MEFs
Murine PARP-1 wild-type and knock-out MEFs were transfected using pGL3-PII-522 wt plasmid, as described (Appendix A, Table A3). After 24 h of forskolin stimulation, DNA and mRNA were isolated using the AllPrep DNA/RNA Mini Kit (Qiagen; Hilden, Germany) with on-column DNA-digestion during the RNA-isolation. cDNA was synthesized as described [19]. The ABsolute SYBR Green Rox Mix (Thermo Scientific; Schwerte, Germany) was used for qRT-PCR in a StepOnePlus™ real-time PCR system (Applied Biosystems). The transfected aromatase promoter-dependent expression of luciferase or the expected fusion sequences containing 5 -ends from aromatase exon I.3 (derived from the 5or 3 -region of exon I.3, respectively) and a 3 -ends derived from luciferase were measured using the primers described in the Supplementary Materials, Table S1, and cDNA as templates. Furthermore, the amounts of firefly luciferase in relation to murine glyceraldehyde-3-phosphate dehydrogenase (GAPDH) were measured using isolated DNA from the same cells as the templates, enabling a correction for transfection efficiencies. All conditions were tested in triplicate per experiment. The evaluation of the PCR results was performed by an accordingly modified ∆∆C T -method [31].

Statistical Analyses
Statistical analyses of all experiments and creation of diagrams were carried out using SigmaPlot 13 software (Systat Software GmbH, Erkrath, Germany). The data are presented as means ± sem or using box plots, were appropriate. Initial normal distribution of values was tested by the method of Shapiro-Wilk. Normally distributed values were compared to another group by Student´s t-test or by one-way ANOVA for multiple comparisons. In the case of non-normally distributed values, two groups were compared by the Mann-Whitney U-test. For all tests, the significance criterion p < 0.05 was used.

SNV-Dependent Protein Complex Formation in the Aromatase Promoter I.3/II Region
We identified a new, extremely rare single nucleotide variant (SNV) in the aromatase promoter I.3/II-region of a healthy DNA-donor (SNV(T-241C); NC_000015.10:n.51243270T>C; GRCh38.p7 human genome reference; Supplementary Materials, Figure S1). This SNV decreased aromatase promoter I.3/II activity in luciferase-reporter gene assays in 3T3-L1 cells by up to 70%, when the cells were stimulated with the cAMP-elevating agonists di-butyryl-cAMP or forskolin ( Figure 1A). This indicates a crucial role for the base-pair at position −241 in relation to the transcriptional start site (TSS) of aromatase promoter II. Two specific protein-oligonucleotide complexes could be identified in soluble nuclear extracts from 3T3-L1 preadipocytes in EMSAs using normal and SNV-containing oligonucleotides, respectively ( Figure 1B). Complex formation was independent of forskolin, which induces an increase in cAMP and thereby mimics cancer-related aromatase promoter I.3/II activation. Furthermore, complex 1 was only formed with wild-type oligonucleotide, but not on the SNV-containing oligonucleotide or an extended quadruple mutation thereof. Complex 2 was not affected by the SNV. This indicates that the formation of complex 1 might be necessary for the full induction of aromatase transcription. The data are means of the three experiments done with triplicate replicates of each. Differences caused by SNV(T-241C) versus wild-type were analyzed with Students t-test (n = 3; * p < 0.05; *** p < 0.001). (B) Soluble nuclear extracts from 3T3-L1 cells were used for electrophoretic mobility shift assays (EMSAs). Complex 1 formation was inhibited by competitor oligonucleotides (comp). SNV (1m)-or quadruple mutation (4m)-containing probes did not form complex 1, as seen with normal sequence probe (n). Complex 2 was solely inhibited by competitor (n = 6; one representative experiment shown; * p < 0.05 for comparison with normal sequence probe only). (C) Flow chart highlighting subsequent steps of investigation (for a more detailed overview see Figure 7).

PARP-1 and Histone H1-Isoforms Bind to the SNV-Region
We subsequently identified the DNA-binding protein(s) of the SNV-dependent complex 1. Firstly, DNA-binding proteins from nuclear extracts of 3T3-L1 cells were purified with SNV sequence-containing oligonucleotides coupled to magnetic beads. After SDS-PAGE, a 110 kDa protein was detectable and identified by mass spectrometry as PARP-1 ( Figure 2A and Supplementary Materials, Table S2). The formation of complex 1 was blocked by anti-PARP-1 antibody in EMSA with soluble nuclear extracts of 3T3-L1 cells, confirming PARP-1 as a component of complex 1 ( Figure 2B). Similar experiments were performed with nuclear extracts from PARP-1 Differences caused by SNV(T-241C) versus wild-type were analyzed with Students t-test (n = 3; * p < 0.05; *** p < 0.001). (B) Soluble nuclear extracts from 3T3-L1 cells were used for electrophoretic mobility shift assays (EMSAs). Complex 1 formation was inhibited by competitor oligonucleotides (comp). SNV (1m)-or quadruple mutation (4m)-containing probes did not form complex 1, as seen with normal sequence probe (n). Complex 2 was solely inhibited by competitor (n = 6; one representative experiment shown; * p < 0.05 for comparison with normal sequence probe only). (C) Flow chart highlighting subsequent steps of investigation (for a more detailed overview see Figure 7).

PARP-1 and Histone H1-Isoforms Bind to the SNV-Region
We subsequently identified the DNA-binding protein(s) of the SNV-dependent complex 1. Firstly, DNA-binding proteins from nuclear extracts of 3T3-L1 cells were purified with SNV sequence-containing oligonucleotides coupled to magnetic beads. After SDS-PAGE, a 110 kDa protein was detectable and identified by mass spectrometry as PARP-1 ( Figure 2A and Supplementary Materials, Table S2). The formation of complex 1 was blocked by anti-PARP-1 antibody in EMSA with soluble nuclear extracts of 3T3-L1 cells, confirming PARP-1 as a component of complex 1 ( Figure 2B). Similar experiments were performed with nuclear extracts from PARP-1 wild-type and knock-out MEFs to confirm this observation ( Figure 2C,D). In PARP-1 wild-type MEFs, the formation of complex 1 on wild-type oligonucleotide was inhibited by anti-PARP-1 antibody or when a mutated probe was used, confirming the results from the 3T3-L1 cells described above. Furthermore, PARP-1-containing complex 1 (but not complex 2) was massively reduced in PARP-1 knock-out MEFs.
wild-type and knock-out MEFs to confirm this observation ( Figure 2C,D). In PARP-1 wild-type MEFs, the formation of complex 1 on wild-type oligonucleotide was inhibited by anti-PARP-1 antibody or when a mutated probe was used, confirming the results from the 3T3-L1 cells described above. Furthermore, PARP-1-containing complex 1 (but not complex 2) was massively reduced in PARP-1 knock-out MEFs. In addition to PARP-1, a second protein band of 32 kDa, specifically binding to the aromatase promoter I.3/II-region, was detectable and identified by mass spectrometry as histone H1-isoforms ( Figure 2A and Supplementary Materials, Table S2).

PARP-1 and Histone H1 Compete for Binding to the Aromatase Promoter I.3/II-Region
In competition experiments using biotin-labeled oligonucleotides for pull-down of associated proteins from 3T3-L1-cell nuclear extracts with magnetic beads, PARP-1 was detectable as the dominant binding protein in ruthenium-stained gels and on western blots ( Figure 3A). The addition of a 10-fold molar excess of non-biotinylated competitor oligonucleotide displaced PARP-1 and caused a prominent histone H1-binding. This asymmetrical competition suggests that PARP-1 is a high-affinity/low-concentration binder of the aromatase promoter I.3/II-region, whereas histone H1 represents a low-affinity/high-concentration binder.
Immunoprecipitation-coupled DNA-binding assays using 3T3-L1 nuclear extracts accessorily revealed the involvement of PARP-1 ( Figure 3B). Immunoprecipitated PARP-1 revealed a similarly weak binding to an SNV-containing probe (1m), as it did to a wild-type probe in the presence of competitor oligonucleotides. In contrast, immunoprecipitated histone H1 revealed an even greater binding to the SNV-containing probe (1m).

Histone H1 Is Parylated
The molecular weight of unmodified murine histone H1 is 21.79 kDa (uniProtKb database). Nonetheless, histone H1 pulled-down with oligonucleotide-coupled magnetic beads from 3T3-L1 nuclear extracts revealed an apparent molecular mass of 32 kDa ( Figure 3A). Western blot analysis of immunoprecipitated histone H1 with a parylation-specific antibody detected a parylated 32 kDa-band ( Figure 3C). This observation suggests that PARP-1-catalyzed parylation caused the observed mass shift of histone H1. The addition of a competitor displaced PARP-1 and enabled histone H1 binding. (B) PARP-1 or histone H1 isolated from soluble nuclear extracts of 3T3-L1 cells by immunoprecipitation bound to wild-type sequence probe (n). Binding to the fluorescent probe was inhibited by the competitor (comp). The SNV-containing fluorescent probe (1m) revealed reduced binding to PARP-1 but increased binding to histone H1 (n = 3, * p < 0.05 for comp or 1m versus n). (C) Western blot using immunoprecipitated histone H1 from 3T3-L1 cell soluble nuclear extract. Parylation was detected by parylation-specific antibodies. In lane 1, anti-histone H1 ab indicates the loading of an excess of the pure antibody used for immunoprecipitation. The asterisk marks the 32 kDa-band of parylated histone H1.

A Dual Role for PARP-1 on the Aromatase Promoter I.3/II
Having identified that PARP-1 preferentially binds to the wild-type sequence in the SNV-region, we subsequently endeavored to elucidate the functional consequence of this interaction on the aromatase promoter I.3/II activity. The overexpression of PARP-1 in 3T3-L1 cells resulted in increased firefly luciferase activity, which was even more prominent after forskolin stimulation ( Figure 4A). Similar results were observed in PARP-1 knock-out MEFs, where PARP-1 overexpression led to biphasic dose responses ( Figure 4B). The maxima of luciferase activities were measured when 150 ng PARP-1 expression plasmid per well was used for transfection and analysis was performed with the reporter gene plasmid containing the wild-type aromatase promoter I.3/II sequence and 75 ng/well using the SNV-sequence. A further increase in PARP-1 expression plasmid resulted in significantly lower reporter gene activity.

A Dual Role for PARP-1 on the Aromatase Promoter I.3/II
Having identified that PARP-1 preferentially binds to the wild-type sequence in the SNV-region, we subsequently endeavored to elucidate the functional consequence of this interaction on the aromatase promoter I.3/II activity. The overexpression of PARP-1 in 3T3-L1 cells resulted in increased firefly luciferase activity, which was even more prominent after forskolin stimulation ( Figure 4A). Similar results were observed in PARP-1 knock-out MEFs, where PARP-1 overexpression led to biphasic dose responses ( Figure 4B). The maxima of luciferase activities were measured when 150 ng PARP-1 expression plasmid per well was used for transfection and analysis was performed with the reporter gene plasmid containing the wild-type aromatase promoter I.3/II sequence and 75 ng/well using the SNV-sequence. A further increase in PARP-1 expression plasmid resulted in significantly lower reporter gene activity. Of note, however, is that in the dual-reporter gene assays, PARP-1 overexpression also massively induced the Renilla luciferase vector (Appendix B, Figure A1A). Therefore, we validated and confirmed the inducing effect of PARP-1 on the aromatase promoter region by another approach. PARP-1 knock-out or wild-type MEFs were transfected with the reporter plasmid containing the normal aromatase promoter I.3/II sequence (pGL3-PII-522 wt) and analyzed by RT-qPCR for transcribed mRNAs, which could be normalized exactly via qPCR for the transfected plasmid-DNA. Three amplicons (firefly luciferase-specific, exon I.3-5 -region-specific and exon I.3-3 -region-specific) were tested and revealed a massive 15-40-fold inducing effect of PARP-1 on promoter I.3/II-dependent aromatase transcription ( Figure 4C). Taken together, PARP-1 binding clearly affects the aromatase promoter I.3/II-region.
Based on these observations, we hypothesized that the inhibition of PARP-1-activity should inhibit aromatase promoter activity. Surprisingly, we observed that treatment with the PARP-1-inhibitor PJ34 resulted in a dose-dependent increase in reporter gene activities of the firefly luciferase promoter I.3/II wild-type-construct ( Figure 4D) and of the Renilla-vector (Appendix A, Figure A1B) in 3T3-L1 cells. This suggests that PARP-1 may also possess inhibitory potential. In contrast, total aromatase expression regulated by promoters I.3 and II as well as aromatase enzyme activity were inhibited by PJ34 in biphasic dose-responses in BAFs ( Figure 4E-H). The strongest inhibition was observed with 5-7 µM PJ34, whereas higher concentrations almost restored these values to those of untreated cells. Taken together, these data suggested a dose-and cell type-dependent dual role of PARP-1 in the regulation of aromatase promoter I.3/II activity. Moreover, both of the transcriptional start sites within the promoter I.3/II-region are affected [7].

HDACs Modify Aromatase Promoter I.3/II Activity
Deacetylation is known to affect PARP-1 and histone H1 function [32,33]. Therefore, HDAC inhibitors were screened in aromatase promoter I.3/II reporter gene assays in 3T3-L1 cells. In forskolin-stimulated cells, the HDAC class I/IIa inhibitor N-butyrate increased luciferase activity, which was significantly reduced by PJ34-mediated PARP-1 inhibition ( Figure 5A), contrary to the effect of PJ34-treatment in the absence of HDAC inhibitor (see Figure 4D), pointing to an interdependent regulation. The class I/II/IV inhibitor panobinostat increased the aromatase promoter I.3/II driven luciferase activity in both the wild-type and SNV-promoter genotypes ( Figure 5B). The SIRT-1 (HDAC class III) inhibitor selisistat, like n-butyrate, augmented luciferase activity, which was significantly reduced by PARP-1 inhibition ( Figure 5C). In contrast to panobinostat, the effect of selisistat could only be observed with the normal sequence reporter, which is indicative of an antagonistic participation of SIRT-1 in the control of the aromatase promoter I.3/II SNV-region by PARP-1. Having identified selisistat/SIRT-1 as the most likely partner for PARP-1, we went back to BAFs. Selisistat significantly inhibited total and promoter I.3/II-specific aromatase mRNA-expression ( Figure 5D) as well as aromatase activity in forskolin-treated BAFs ( Figure 5E). Interestingly, we observed a statistically not significant tendency for a biphasic dose response to PARP-1 inhibition in the presence of selisistat in the RNA expression measurements, which was diametrically opposite to the U-shaped PJ34 dose response of the controls. The large variances are due to widely differing sensitivities of BAFs to selisistat. In the aromatase activity assays, these effects were not detectable, probably due to interference with post-transcriptional effects of the inhibitors. In summary, SIRT-1 and PARP-1 functionally interact in the control of aromatase expression.
Enzymatic activities of both PARP-1 and SIRT-1 depend on NAD + -consumption [13]. Despite the presence of multiple other NAD + -metabolizing enzymes, the inhibition of PARP-1 or SIRT-1 or both together resulted in increased NAD + /NADH ratios ( Figure 5F), indicating that both enzymes consume quite large amounts of, and compete at least locally for, NAD + .

Aromatase Expression Is Regulated by an Antagonizing System Including PARP-1, Histone H1 and SIRT-1
To conclusively verify the in vivo binding of PARP-1, histone H1 and SIRT-1 to the SNV-region in the aromatase promoter I.3/II, we performed ChIP assays utilizing control and forskolin-stimulated BAFs. Two overlapping primer sets covering the SNV-containing region of promoter I.3/II were used ( Figure 6A). In accordance with the results of the in vitro studies as described above, forskolin treatment lead to a significantly increased PARP-1 binding, when analyzed with primer set 1 ( Figure 6B,C). The switch in promoter occupancy was even more obvious when ratios of PARP-1/SIRT-1 and of PARP-1/histone H1 were analyzed ( Figure 6D). Moreover, primer set 2 suggested that the partial displacement of histone H1 and SIRT-1 by PARP-1 takes place in the immediate neighborhood of the SNV-region ( Figure 6E,F). Here too, the ratios of PARP-1/SIRT-1 and of PARP-1/histone H1 significantly increased in forskolin-stimulated BAFs ( Figure 6G), indicating the partial displacement of SIRT-1 and histone H1 by PARP-1. Taken together, these results indicate that PARP-1, H1, and SIRT-1 directly interact in an aromatase promoter I.3/II-region crucial for the induction in BAFs in the vicinity of breast cancer cells in vivo.  Figure S2; the smaller band is unspecific). Forskolin stimulation increased PARP-1 binding and increased ratios of PARP-1/SIRT-1 and of PARP-1/histone H1 (n = 4; * p < 0.05 versus control).   Figure S2; the smaller band is unspecific). Forskolin stimulation increased PARP-1 binding and increased ratios of PARP-1/SIRT-1 and of PARP-1/histone H1 (n = 4; * p < 0.05 versus control).

Discussion
Besides its role in DNA damage response and various other processes, the involvement of PARP-1 in transcriptional control has been established [34,35]. The identification of PARP-1 as a transcriptional regulator of aromatase expression signifies the promoter I.3/II-region of aromatase as one of PARP-1´s many targets [36]. More recently, PARP-1 received attention as a therapeutic target in TNBC and is currently the subject of a series of clinical trials of specific inhibitors [14,37,38], based on its function in DNA damage control. In addition, some authors have concluded that PARP-1 inhibition could be useful in a broader range of breast cancers, including ER + breast cancers (not necessarily mutated in the BRCA DNA repair associated genes) [14,39].
Our study provides evidence that the inhibition of PARP-1 might be a pharmacological possibility for treating ER + breast cancers by targeting transcription. By happenstance, we have identified a new rare SNV, located immediately downstream of the TATA-box of aromatase promoter I.3 (see Figure 7), which strongly reduced reporter gene activity driven by the promoter I.3/II-region. Promoter I.3 is intimately connected with promoter II, both functionally and by proximity (the transcriptional start-site (TSS) of promoter II is only 226 bp downstream of the promoter I.3-TSS) [7,8]. PARP-1 preferentially binds to the normal sequence of the SNV. The reduction in binding to the SNV variant in in vitro binding assays was almost equal to a reduction in promotor activity in reporter gene assays, which strongly suggests that PARP-1 is a critical component for the full induction of the breast cancer-associated promoter I.3/II-region.
In principle, it seems counterintuitive that PARP-1 binding to a DNA site that at least overlaps with the TATA-box core-promoter element should enhance transcription. However, several lines of evidence clarify this supposed discrepancy. First, the binding location is typical for PARP-1, as Krishnakumar et al. [36] identified the most probable PARP-1 accumulation to be about 250 bp upstream of TSSs in their large-scale analysis of promoters, almost perfectly fitting the SNV-spacing of 241 bp upstream of the promoter II TSS. Secondly, the overexpression of PARP-1 induces promoter I.3/II controlled reporter gene activity in cell lines, whereas the inhibition of PARP-1 inhibits aromatase expression and activity in BAFs. Thirdly, Wang et al. [40] found that the 5´-region of the non-coding exon I. 3 inhibits the translation of the aromatase protein. The results of our qPCR-validation of the reporter gene experiments in MEFs (see Figure 4C) reveal that the presence of PARP-1 leads to preferential transcription of a 5´-truncated exon I.3, resulting in the omission of this inhibitory region. Taken together, these findings indicate a central role of the SNV-region and PARP-1 for induction of aromatase.
We observed biphasic dose responses of the aromatase expression and enzyme activity of BAFs treated with the PARP-1-inhibitor PJ34. This cannot be explained by a stand-alone action of PARP-1 at the aromatase promoter. Furthermore, the overexpression of PARP-1 in PARP-1 knock-out MEFs also resulted in a biphasic dose response, which might be caused by autoparylation at high concentrations of PARP-1 and concomitant autoinhibition [41]. Taken together, all of the results suggest that PARP-1 must be part of a multifactorial system, in which interactions of different partners modulate the aromatase promoter I.3/II activity (Figure 7).
One important interaction partner is histone H1 (H1), which co-purified with PARP-1 in our assays. H1 even exhibited an increased binding to the mutant SNV sequence in line with the functional association of H1 and PARP-1 reported in the literature [34]. Aubin et al. described euchromatin formation by PARP-1-mediated H1 parylation and its displacement, resulting in gene activation [42]. PARP-1 inhibition reduced gene activity via H1-mediated heterochromatin formation. In this context, such dynamic interactions of H1 and PARP-1 in large-scale analyses were reported by Krishnakumar et al. [36]. Our competition experiments and ChIP analyses confirmed this dynamic regulation for the aromatase promoter I.3/II-region as well. Interestingly, most of the H1 from 3T3-L1 cells was parylated. In Western blots of immunoprecipitated H1, as well as in the DNA-coupled bead assays, we found a size shift in H1 to 32 kDa, which is typical for parylation, as described by Huletsky et al. [43]. The involvement of H1 in aromatase promoter I.3/II regulation also indicates the involvement of HDACs. Besides the modulation of the nucleosome configuration by histone deacetylation, HDACs modify transcription factors [44]. HDAC class I/II/IV inhibition in 3T3-L1 cells increased the aromatase promoter I.3/II activity, while concurrent PARP-1-inhibition reduced its activity, indicating that acetylation events necessary for aromatase transcription need PARP-1 for a maximum effect. Contrarily, PARP-1 is activated through acetylation by p300/CBP, which can be reversed by class I HDACs [32]. The involvement of H1 in aromatase promoter I.3/II regulation also indicates the involvement of HDACs. Besides the modulation of the nucleosome configuration by histone deacetylation, HDACs modify transcription factors [44]. HDAC class I/II/IV inhibition in 3T3-L1 cells increased the aromatase promoter I.3/II activity, while concurrent PARP-1-inhibition reduced its activity, indicating that acetylation events necessary for aromatase transcription need PARP-1 for a maximum effect. Contrarily, PARP-1 is activated through acetylation by p300/CBP, which can be reversed by class I HDACs [32].
Nonetheless, a special role in the functional relationship between HDACs and PARP-1 is assigned to class III HDACs (sirtuins). Their deacetylase function involves a NAD + -dependent catalytic mechanism [45] and, in particular, SIRT-1 mediates deacetylation of PARP-1 [46]. In our experiments, SIRT-1 inhibition in 3T3-L1 cells resulted in similar effects as HDAC class I/II/IV inhibition. In contrast, in BAFs SIRT-1 inhibition caused a reduction in aromatase expression and activity. This inhibition tended to be alleviated in the presence of moderate concentrations of a PARP-1 inhibitor. In BAFs, SIRT-1 seems to act as an aromatase promoter I.3/II activator. Taken together, experimental evidence suggests a strong functional relationship between SIRT-1 and PARP-1, which may be modulated in a cell type-specific manner. This functional link appears to be the NAD + -requirement of both enzymes [34,47,48]. Furthermore, SIRT-1 and PARP-1 have different kinetic features. The Michaelis constant (K M ) of PARP-1 is 20-60 µM NAD + [49]. K M of SIRT-1 is 150-200 µM NAD + [50]. Therefore, PARP-1 is more active than SIRT-1 at low NAD + -levels, as can be observed, for example, in cardiomyocytes [51]. Conversely, high NAD + -levels favor increased SIRT-1 activation, leading to PARP-1 inhibition by deacetylation (and inhibition of gene expression) [46]. In fact, in our ChIP experiments, a shift towards stronger promoter I.3/II occupancy by PARP-1 upon forskolin stimulation could be observed. Furthermore, with PARP-1 and SIRT-1 being involved in the regulation of aromatase expression, the cellular context becomes important. BAFs in the neighborhood of tumor cells may exhibit a reverse Warburg effect like CAFs [52]. This effect is well described for CAFs in breast cancers [53,54]. The reverse Warburg effect is triggered by metabolites like lactate, released from tumor cells. In stromal cells, their utilization can reduce the NAD + /NADH ratio [55]. The subsequently low NAD + -level favors PARP-1-activity as compared to SIRT-1-activity.
The U-shaped dose responses to PARP-1 inhibition in BAFs result from an interaction with a partner, which itself is functionally coupled to PARP-1 and stimulates aromatase expression independently. Our observations using SIRT-1 inhibition in BAFs suggest that SIRT-1 is this partner. Active PARP-1 deprives SIRT-1 of NAD + , thus preventing its own inactivation by SIRT-1. Furthermore, PARP-1 supports euchromatin formation by H1-parylation (in addition to acetylation by acetyltransferases) contributing to the activation of aromatase promoter I.3/II (Figure 7). Only under conditions of PARP-1 inhibition does the cellular NAD + -concentration increase to a sufficiently high level to allow SIRT-1 to stimulate aromatase expression via, as of now, unidentified mechanisms.

Conclusions
In summary, PARP-1 is a key regulator of the aromatase promoter I.3/II activity, which can be activated in BAFs by metabolic coupling to breast cancer cells. Lack of PARP-1 binding to the promoter I.3/II region or PARP-1 inhibition, respectively, reduces aromatase expression as well as estrogen synthesis in BAFs. Thus, low doses of PARP-1 inhibitors might be potentially useful for estrogen-deprivation of ER + breast cancers.
Supplementary Materials: The following are available online at http://www.mdpi.com/2073-4409/9/2/427/s1, Figure S1: SNV in the aromatase promoter I.3/II DNA sequence, Figure S2: Original gel image for Figure 6E, Table S1: Primers and conditions for PCR amplifications, Table S2:  Funding: A.K. was supported by a stipend from the Richard-Winter-Stiftung, Stuttgart. The founding sponsors had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, and in the decision to publish the results.  Table A1. Oligonucleotide sequences for protein identification, EMSA, DNA-affinity purification and immunoprecipitation-based DNA-binding assay. Mutated positions/bases differing from the GenBank-sequence are given in bold letters. Labels are at the 5 -ends.