Activated Transcription Factor 3 in Association with Histone Deacetylase 6 Negatively Regulates MicroRNA 199a2 Transcription by Chromatin Remodeling and Reduces Endothelin-1 Expression

Abstract

Previous studies showed that high levels of placenta growth factor (PlGF) correlated with increased plasma levels of endothelin-1 (ET-1), a potent vasoconstrictor, in sickle cell disease (SCD). PlGF-mediated transcription of the ET-1 gene occurs by activation of hypoxia inducible factor 1α (HIF-1α) and posttranscriptionally by microRNA 199a2 (miR-199a2), which targets the 3′ untranslated region (UTR) of HIF-1α mRNA. However, relatively less is known about how PlGF represses the expression of miR-199a2 located in the DNM3 opposite strand (DNM3os) transcription unit. Here, we show that PlGF induces the expression of activated transcription factor 3 (ATF3), which, in association with accessory proteins (c-Jun dimerization protein 2 [JDP2], ATF2, and histone deacetylase 6 [HDAC6]), as determined by proteomic analysis, binds to the DNM3os promoter. Furthermore, we show that association of HDAC6 with ATF3 at its binding site in this promoter was correlated with repression of miR-199a2 transcription, as shown by DNM3os transcription reporter and chromatin immunoprecipitation (ChIP) assays. Tubacin, an inhibitor of HDAC6, antagonized PlGF-mediated repression of DNM3os/pre-miR-199a2 transcription with a concomitant reduction in ET-1 levels in cultured endothelial cells. Analysis of lung tissues from Berkeley sickle (BK-SS) mice showed increased levels of ATF3 and increased expression of ET-1. Delivery of tubacin to BK-SS mice significantly attenuated plasma ET-1 and PlGF levels. Our studies demonstrated that ATF3 in conjunction with HDAC6 acts as a transcriptional repressor of the DNM3os/miR-199a2 locus.

INTRODUCTION

Placenta growth factor (PlGF), an angiogenic growth factor of the vascular endothelial growth factor (VEGF) family, has pleiotropic and redundant roles in development but features prominently in various pathological conditions such as ischemia, angiogenesis, inflammation, atherosclerosis, preeclampsia, and diabetic wound healing (1–9). PlGF expression is induced by hypoxia, erythropoietin, and iron; PlGF effects are manifested via binding to its sole cognate receptor, VEGF receptor 1 (VEGFR-1) (10–13).

Recent studies have shown that plasma levels of PlGF are abnormally high in patients with hemolytic anemias, such as in sickle cell disease (SCD) (12, 14). Moreover, high plasma PlGF levels correlate with increased incidence of vaso-occlusive events in SCD subjects (12). Consistent with these findings, plasma levels of plasminogen activator inhibitor 1 (PAI-1) and endothelin-1 (ET-1) are high in a mouse model of sickle cell disease, namely, Berkeley sickle (BK-SS) mice, which also show high plasma PlGF levels (15, 16). Conversely, PlGF knockout mice exhibit low plasma PlGF levels and low plasma levels of PAI-1 and ET-1 (15, 16). Furthermore, augmentation of erythroid PlGF expression in normal mice to the levels seen in sickle mice results in increased production of endothelin-1 with associated pulmonary changes, as seen in pulmonary hypertension (PH) in SCD (16); these results were corroborated in a study of 123 patients with SCD (16). Thus, studies in vitro and in vivo support the crucial role of PlGF in upregulating the expression of ET-1 in endothelial cells and associated pulmonary changes characteristic of pulmonary hypertension in SCD.

We show that PlGF activates endothelial cells to upregulate the expression of genes such as the ET-1 and PAI-1 genes via activation of hypoxia-inducible factor 1 (HIF-1α), independently of hypoxia (15, 17). Moreover, we have shown that posttranscriptional regulation of ET-1 and PAI-1 is achieved by microRNA 199a2 (miR-199a2), which targets the 3′ untranslated region (UTR) of HIF-1α mRNA (18).

The DNM3 opposite strand (DNM3os) gene produces a noncoding RNA (ncRNA) that serves as the precursor of miR-199a2 and miR-214 (18). This locus is located within an intron of DNM3 and is transcribed from the opposite strand, hence, its designation. Transcription of DNM3os and of cotranscribed miR-199a2/miR-214 is greatly attenuated by PlGF, thus allowing unhindered expansion of HIF-1α activity and expression of genes, i.e., the ET-1 and PAI-1 genes, requiring this transcription factor, (18). However, the molecular mechanism of PlGF-mediated downregulation of miR-199a2/miR-214 and DNM3os ncRNA is not fully understood.

In the present study, we show that repression of DNM3os in response to PlGF required the participation of activated transcription factor 3 (ATF3), which binds to ATF3 response elements in the DNM3os promoter. The ATF3 gene, a stress-inducible gene, has been shown to play important roles in several pathological conditions, including host immunity and cancer (19–21). To delineate the mechanism by which ATF3 represses DNM3os transcription, we performed a whole-cell proteomic analysis of ATF3-interacting proteins, identified by protein mass spectroscopy of multiprotein complexes coimmunoprecipitated with antibody to ATF3. This analysis revealed that ATF2, c-Jun dimerization protein 2 (JDP2), and histone deacetylase 6 (HDAC6) associated with ATF3 as candidate accessory factors, suggesting that chromatin remodeling of the DNM3os locus was the mechanism by which PlGF repressed the gene. Moreover, our studies showed that HDAC6 activity contributed to DNM3os repression in vitro and in the BK-SS mouse model of SCD.

MATERIALS AND METHODS

Reagents.

Recombinant human PlGF was purchased from Peprotech (Rocky Hill, NJ); primary antibodies to ATF3, JDP2, ATF2, c-Jun, HDAC6, and HDAC7 were obtained from Santa Cruz Biotechnology (Santa Cruz, CA); anti-histone H3 acetylated at K9 (anti-H3K9Ac) and anti-H3K27Ac were obtained from Abcam (Cambridge, MA); antibodies against β-actin and secondary antibodies conjugated to horseradish peroxidase (HRP) were purchased from Sigma-Aldrich Chemical Company (St. Louis, MO). The DNA primers used for PCR amplification and mutagenesis of the ATF3 promoter were purchased from Valuegene (San Diego, CA). ATF3 (wild-type [wt] transcript variant 1; full-length, 181 amino acids [aa]) was obtained from Origene (Rockville, MD). A truncated ATF3 clone (ΔATF3) lacking the leucine zipper domain was created by site-directed mutagenesis, creating a stop codon after that for amino acid 91. Trichostatin A (TSA), tubacin, and mocetinostat (HDAC inhibitors) were purchased from Selleckchem (Houston, TX). Bulk quantities of tubacin and niltubacin for animal studies were obtained from AbMole Bioscience, Inc. (Houston, TX). Unless specified, all other reagents were purchased from Sigma-Aldrich Chemical Company (St. Louis, MO).

Cell culture.

The immortalized human dermal microvascular endothelial cell line 1 (HMEC-1) was cultured as previously described (18).

Mice.

Berkeley sickle (BK-SS) mice originally obtained from Jackson Laboratories were bred up to the sixth generation against a C57BL/6NJ background (15, 22). In-house-bred C57BL/6NJ mice served as controls. Six to 10 animals per group, 2 to 3 months old were used. Animal protocols were approved by the Institutional Animal Care and Use Committee at Cincinnati Children's Hospital Medical Center. At the end of the study mice were exsanguinated, and lungs were removed; tissues were stored at −80°C for later assay of microRNAs (miRNAs) and ET-1.

RNA interference.

The ATF3, Jun N-terminal protein kinase 1 (JNK-1), JNK-2, phosphatidylinositol 3-kinase (PI3K), mitogen-activated protein kinase (MAPK), HDAC1, HDAC3, HDAC4, HDAC6, HDAC7, HDAC8, HDAC9, and HDAC11 short hairpin RNA (shRNA) clones were a generous gift from Jae Jung (University of Southern California [USC]). The human shRNA library was purchased from Open Biosystems (now Thermo Scientific, Grand Island, NY). All synthetic shRNAs were supplied in pGIPZ plasmids for transfection.

Isolation of RNA and qRT-PCR.

Total RNA was isolated, and mRNA expression was quantified using specific primers (Table 1) by quantitative real-time PCR (qRT-PCR) (18).

TABLE 1.

Oligonucleotide primers used in this study

Gene or target	Method	Forward sequence (5′–3′)	Reverse sequence (5′–3′)
DNM3os	qRT-PCR	GTCAGCGCAGCAGAATTCAG	CGGCAGTCTTTTCTCAGCAG
HIF-1α	qRT-PCR	CCATTAGAAAGCAGTTCCGC	TGGGTAGGAGATGGAGATGC
ET-1	qRT-PCR	GAAACCCACTCCCAGTCCAC	CGGGAGTGTTGACCCAAATG
ATF3	qRT-PCR	TAGCATTACGTCAGCCTGGG	AGCGTTGCATCACCCCTTTT
GAPDHa	qRT-PCR	AACCTGCCAAGTACGATGACATC	GTAGCCCAGGATGCCCTTGA
Pre-miR-199a2	qRT-PCR	AGGAAGCTTCTGGAGATCC	TGCTCTCCCTTGCCCAG
ATF3mt	PCR	TCTTACGCGTGCTAGCAATTACATACATAGCCTGGG	GATCGCAGATCTCGAGCTCACTTCCGAGGCAGAG
miR-199a-5p	Northern blotting	GAACAGGTAGTCTGAACACTGGG
5S rRNA	Northern blotting	CAGGCCCGACCCTGCTTAGCTTCCGAGATCAGACG
ATF3mt1	PCR	CCCAGGGTGAGCCCATCCCATATATGGACTCTCC	GAGGTTTGGGTATGCGCT
ATF3mt2	PCT	CGTTTCTTTAACTCAAACTATAGCACAGCATTC	TTACAGTTTTTGTTTGAAGCATC
ATF3 binding site 1 of DNM3os	ChIP-PCR	CAGGCGATTCTAGCGGTCTC	GCTGGGCTGGAGAGTCCATA
ATF3 binding site 2 of DNM3os	ChIP-PCR	TCGTTGTTTCGTAGTGGATGC	TGTGGCCTTCCTAGTTTCCC
Distal binding site (negative control) of DNM3os	ChIP-PCR	TGCCTTGGAGGCCACATAAA	ATCATGCCCCAAACCCATTA
ATF3 (mouse)	qRT-PCR	CTCACTCAGCGAGACGCC	TTGTTTCGACACTTGGCAGC
HIF-1α (mouse)	qRT-PCR	CCTGTAAGCAAGGAGCCAGA	TGGGGAAGTGGCAACTGATG
ET-1 (mouse)	qRT-PCR	TGCCTCTGAAGTTAGCCGTG	AGTTCTCCGCCGCCTTTTTA
GAPDH (mouse)	qRT-PCR	TTGCAGTGGCAAAGTGGAGA	GTCTCGCTCCTGGAAGATGG

^aGAPDH, glyceraldehyde-3-phosphate dehydrogenase.

Isolation and Northern blotting of miRNAs.

MicroRNAs (miRNAs) were isolated as described above and quantified by Northern blotting utilizing biotinylated probes for miR-199a2 and 5S rRNA, purchased from Valuegene (San Diego, CA), as described previously (18).

Plasmids and plasmid construction.

A 2-kb DNA segment containing the DNM3os promoter was ligated into a pGL3-basic luciferase reporter plasmid (18) and also used as the template for site-specific mutations. The ATF3 site mutations were generated with a Q5 site-directed mutagenesis kit (New England BioLabs, Ipswich, MA) utilizing primers listed in Table 1. The resulting single ATF3 mutations in the DNM3os promoter are designated ATF3mt1 (nucleotides [nt] −216/−206) and ATF3mt2 (nt −789/−779). A promoter with both ATF3mt1 and ATF3mt2 (double mutant; ATF3DM) sites was generated from the ATF3mt1 construct as the template using primers listed in Table 1. Michael Kilberg (University of Florida) kindly provided the ATF3 promoter (bases −100/+35)-luciferase reporter in pGL3 and an ATF3 promoter mutant with a mutated cyclic AMP (cAMP) response element (CRE) (located at nt −107/+35) (23). A truncated ATF3 expression gene (ΔATF3) plasmid, lacking the leucine zipper domain, was generated by creating a stop codon after that for Arg-91 utilizing a Q5 site-directed mutagenesis kit (New England BioLabs, Ipswich, MA) with the wt ATF3 expression plasmid as the template. All constructs and sequence mutations were verified by DNA sequencing (Retrogen, San Diego, CA).

Transient transfections.

Endothelial cells (10⁶ cells) were transiently transfected by nucleofection (24) with the indicated shRNA vector, expression constructs (0.5 μg), and luciferase reporter plasmids (1.0 μg) (24). The Renilla luciferase plasmid (pRLSV40; 1.0 μg) was cotransfected with firefly luciferase reporter constructs to monitor transfection efficiency. Following nucleofection, the cells were incubated in growth medium overnight, followed by serum deprivation for 3 h, and then treated with PlGF (250 ng/ml) for the indicated time periods. Luciferase activity was normalized to Renilla luciferase activity and expressed relative to the activity of the pGL3 control vector, as appropriate.

Western blot analysis.

Cells were lysed in radioimmunoprecipitation assay (RIPA) buffer, and ∼25 μg of protein extracts was subjected to SDS-PAGE, followed by Western blotting (18). Membranes were probed with the indicated dilutions of anti-ATF3 (1:1,000) (sc-22798, lot L1412; Santa Cruz Biotechnology, Santa Cruz, CA), anti-HDAC6 (1:1,000) (sc-11420, lot J0713; Santa Cruz Biotechnology), anti-JDP2 (1:500) (sc-367695, lot J3013; Santa Cruz Biotechnology), and anti-ATF2 (sc-187, lot B03115; Santa Cruz Biotechnology). HRP-conjugated anti-rabbit secondary antibodies were used (1:5,000) (A-0545, lot 083M4752; Sigma-Aldrich, St. Louis, MO). Membranes were stripped and reprobed with anti-β-actin (1:50,000) (A-5316, lot R124317; Sigma-Aldrich). Quantitative analysis was performed using ImageJ analysis software.

Immunoprecipitation, mass spectroscopy, and proteomic analysis.

HMEC-1 cells (10⁶ cells) suspended in RPMI 1640 medium (0.1 ml) were transfected with an ATF3 expression plasmid (1 μg) by nucleofection followed by incubation for 24 h (25) or treated with PlGF (250 ng/ml) for 6 h. The cells were washed with phosphate-buffered saline (PBS) and lysed with RIPA buffer. Cell lysates from 2.5 × 10⁷ cells were incubated for 4 h with either ATF3 antibody (Santa Cruz Biotechnology) or rabbit IgG in binding buffer (Tris-buffered saline [TBS] containing 350 mM NaCl and 0.3% Nonidet P40), followed by the addition of protein A/G-agarose beads (Pierce) and further incubated overnight at 4°C on a rocking platform. The beads were washed six times with binding buffer, and antibody-bound beads were subjected to reduction/alkylation (5 mM dithiothreitol [DTT] at 56°C for 30 min, followed by treatment with 25 mM iodoacetamide in the dark for 20 min). The alkylated protein was incubated with 10 ng of trypsin and Lys-C (modified sequencing grade; Roche) in 50 mM sodium carbonate buffer overnight at 37°C with shaking. After digestion the mixture was acidified with 10 μl of 10% formic acid (FA), and peptides were separated from the beads by filtering through a C₁₈ Stage Tip column (Proxeon), followed by elution with 20 μl of 50% methanol–5% FA. The eluate was subjected to liquid chromatography-tandem mass spectrometry (LC-MS/MS) sequencing using an LC/MS system consisting of an Eksigent NanoLC Ultra 2D system (Dublin, CA) and a Thermo Fisher Scientific LTQ Orbitrap XL (San Jose, CA) (26). Protein identifications were made using the commercially available Proteome Discoverer, version 1.4 (Thermo Fisher Scientific), search engine (27).

Identification of ATF3 accessory proteins following coimmunoprecipitation.

Briefly, the lysates from ATF3-expressing cells were immunoprecipitated with different antibodies against candidate accessory factors, followed by immunoblotting (IB) to identify proteins associated with ATF3. Coimmunoprecipitation reactions were conducted with the indicated antibody or control IgG as described above. Binding was performed at 4°C overnight on a rocking platform, followed by six washes in binding solution. Bound material was eluted from beads by boiling for 5 min in Laemmli buffer, fractionated on a 10% SDS-polyacrylamide gel, and subjected to a Western blotting procedure.

FAIRE.

HMEC-1 cells (5 × 10⁶ cells) in complete medium were treated with PlGF for 18 h, followed by the addition of formaldehyde to a final concentration of 1%, and quenched with glycine as described previously (28). Cells were scraped, and DNA was subjected to sonication for six 18-s pulses, resulting in fragment lengths of ∼300 to 400 bp; subsequent procedural steps were as described in the protocol of Giresi et al. (28). The genomic (input) DNA and DNA from formaldehyde-assisted isolation of regulatory elements (FAIRE) were isolated and purified and subjected to quantitative PCR (qPCR) as described in the protocol (28).

ChIP assay.

HMEC-1 cells (5 × 10⁶ cells) were kept in serum-free medium (SFM) for 3 h and treated with PlGF (250 ng/ml) for 6 h. Chromatin immunoprecipitation (ChIP) analysis was performed using antibodies to ATF3, c-Jun, HDAC6, HDAC7 (Santa Cruz Biotechnology, Santa Cruz, CA), H3K9Ac, H3K27Ac (Abcam, Cambridge, MA), and nonspecific IgG (Santa Cruz Biotechnology, Santa Cruz, CA) (29). Briefly, immunoprecipitated DNA was air dried and suspended in nuclease-free water. DNA samples were subjected to PCR amplification utilizing primers for the DNM3os promoter regions of interest corresponding to ATF3 binding sites proximal and distal to the transcription start site (TSS) (Table 1). PCR was performed for 35 cycles under the following conditions: denaturation at 95°C for 30 s, annealing at 55°C for 60 s, and extension at 72°C for 2 min. The PCR products were subjected to electrophoresis on a 2% agarose gel, visualized by ethidium bromide staining, and quantified using the ImageJ analysis software (18).

ELISA.

HMEC-1 cells (10⁶ cells) were transfected with the indicated HDAC6 or ATF3 shRNA in complete medium and incubated at 37°C for 24 h. Cells were washed with serum-free medium (SFM) and incubated for 3 h in SFM (2 ml). Treatments were begun by replacement of SFM with fresh SFM (1 ml). Cells were treated with either PlGF (250 ng/ml), shRNAs, or tubacin overnight, after which the culture supernatant was collected, and an aliquot (0.1 ml) was assayed for ET-1 using an enzyme-linked immunosorbent assay (ELISA) kit (Assay Design/Enzo Life Sciences, Farmingdale, NY) (17). Cell pellets were assayed for protein content utilizing the Bradford method. Plasma specimens were assayed for ET-1 levels using a DuoSet ELISA kit from R&D Systems (Minneapolis, MN).

Tubacin treatment of Berkeley sickle mice.

Berkeley sickle mice (BK-SS) bred in-house as described above were subjected to two different drug delivery protocols. In the first protocol, 18 BK-SS mice (4 to 6 months old), equally distributed for gender between treatment groups, were injected daily with tubacin (0.5 mg/kg intraperitoneally [i.p.]) (M1742; AbMole) or the inactive tubacin derivative (niltubacin; 0.5 mg/kg i.p.) (M1738; AbMole). Three mice in each treatment arm were exsanguinated on day 5, at which time lung tissues and blood were harvested. The remaining mice were dosed daily with tubacin or niltubacin, as appropriate. On the 10th day, mice were exsanguinated for harvesting of blood and lung tissues. Since the anemic and relatively fragile BK-SS mice poorly tolerate long-term i.p. injections, in the second treatment protocol drug delivery was accomplished by continuous infusion employing subcutaneous osmotic pumps (catalogue number 1004; Alzet) for drug delivery over 30 days. In these studies, a total of 19 BK-SS mice were distributed between treatment groups for delivery of tubacin (n = 10) and niltubacin (n = 9). The drugs were dissolved in dimethyl sulfoxide (DMSO)-PBS (1:1) to a concentration of 1.56 mg/liter for pump loading. Two minipumps were inserted in the dorsal flank of each mouse. Mice were dosed with 0.0083 mg/day (equivalent to a 0.33-mg/kg daily dose, assuming a body weight of 25 g). The drug was infused at a rate of 0.11 μl/h, and each pump delivered 100 μl. After 4 weeks, mice were sacrificed, and lungs and blood were collected. Lung tissue samples were immediately snap-frozen and stored at −80°C; blood was centrifuged to obtain plasma, which was stored at −80°C for later use in assays.

Statistical analysis.

The significance between two data groups was ascertained using an unpaired Student t test, and results are presented as means ± standard errors of the means (SEM).

RESULTS

PlGF upregulates expression of ATF3, and transcription of the ATF3 gene involves an ATF/CRE site in its promoter.

Since ATF3, a member of the basic region-leucine zipper (bZIP) family of transcription factors, has been shown to function as a transcriptional repressor and since the DNM3os promoter-proximal region contains ATF3 consensus DNA binding sites, we examined this role of ATF3. PlGF treatment of HMEC-1 cells resulted in a time-dependent increase in ATF3 mRNA expression, with the maximum increase of ∼3.5-fold observed at 2 h, while expression of DNM3os declined by ∼50% at 6 h post-PlGF induction (Fig. 1A). At 6 h there was also a ∼2.4-fold increase in ATF3 protein expression (Fig. 1B). Next, we examined whether PlGF increased transcription of the ATF3 gene, utilizing an ATF3 promoter-luciferase reporter plasmid containing the −107/+35 promoter fragment of the ATF3 gene (23). This ATF3 promoter fragment was previously shown to have full transcriptional activity compared to a larger promoter segment (−1850/+35). It was responsive to homocysteine and contains the ATF/cAMP response element (CRE) (30). Treatment with PlGF increased reporter expression by ∼3-fold (Fig. 1C, lane 2 versus lane 1), while mutation of the ATF/CRE site (located at nt −93/−86) in this promoter rendered the reporter unresponsive to stimulation by PlGF (Fig. 1D, lane 2 versus lane 1). These data showed that the ATF/CRE site is required for ATF3 gene induction in response to PlGF.

FIG 1.

PlGF upregulates expression of ATF3 and represses the downstream target, the DNM3os gene, in HMEC-1 cells. (A) Time course of PlGF-mediated expression of ATF3 and DNM3os mRNA. (B) PlGF-mediated expression of ATF3 protein at 6 h. The mRNA data are means ± SEM (n = 3), whereas the Western blots are representative of three independent experiments. (C and D) The regulation of ATF3 promoter by PlGF involves ATF/cAMP response element (CRE). HMEC-1 cells were transfected with indicated wild-type ATF3 promoter (nt −107/+35) reporter plasmid or mutant promoter construct (mutation of ATF/CRE site in ATF3 promoter; ATF3mt), followed by treatment with PlGF for 6 h and measurement of luciferase activity. Data are means ± SEM (n = 3). ***, P < 0.001; ns, not significant (P > 0.05).

PlGF-mediated upregulation of ATF3 represses transcription of pre-miR-199a2 and its host, the DNM3os gene.

Next, we examined whether ATF3 repressed DNM3os transcription. As previously reported, the DNM3os transcription unit includes miR-199a2 and miR-214 (18), implying that DNM3os transcription leads to synthesis of a pre-miR-199a2/pre-miR-214 precursor that is further processed into discrete miRNA molecules. Under basal conditions, in the absence of PlGF treatment, transfection with ATF3 shRNA had no effect on DNM3os transcription, similar to the result observed with a scrambled shRNA (scRNA) in HMEC-1 cells (Fig. 2A, lane 2 versus lane 3). Transcription of pre-miR-199a2 also was not affected by ATF3 shRNA (Fig. 2B, lanes 2 versus lane 3). However, upon treatment with PlGF there was an ∼70% reduction in DNM3os RNA (Fig. 2A, lane 4 versus lane 1) (P < 0.001) and ∼60% reduction of pre-miR-199a2 (Fig. 2B, lane 4 versus lane 1) (P < 0.001). The repression observed by PlGF treatment was reversed upon transfection with ATF3 shRNA but not with scRNA, as indicated by increased DNM3os RNA synthesis (Fig. 2A, lane 5 versus lane 6) and expression of pre-miR-199a2 (Fig. 2B, lane 5 versus lane 6).

FIG 2.

ATF3 acts as a repressor of DNM3os and miR-199a2 transcription. (A and B) Effect of ATF3 shRNA on PlGF-mediated expression of DNM3os RNA and pre-miR-199a2. (C and D) Effects of exogenous ATF3 (variant 1) or truncated ATF3 on DNM3os and pre-miR-199a2 expression. The mRNA data are means ± SEM (n = 3). (E) Northern blot of miR-199a2 expression in HMEC-1 cells in response to treatment with PlGF alone for 6 h, transfection with ATF3 shRNA followed by PlGF treatment, and transfection with the ATF3 expression plasmid without PlGF treatment. The values indicated are from densitometric scans after normalization to 5S rRNA as an internal control and are representative of three independent experiments. ***, P < 0.001; ns, not significant (P > 0.05).

As further validation of ATF3 acting as a repressor of DNM3os transcription, HMEC-1 cells were transfected with a wt ATF3 expression plasmid (ATF3 variant 1) and a truncated ATF3 expression plasmid (ΔATF3) lacking the leucine zipper domain. In the absence of PlGF, expression of wt ATF3 reduced basal levels of DNM3os (Fig. 2C, lane 3 versus lane 2) and pre-miR-199a2 (Fig. 2D) by ∼70%, whereas the truncated ATF3 (ΔATF3) was ineffective (Fig. 2C and D). Next, we measured miR-199a2 levels under these conditions by Northern blotting. As shown in Fig. 2E, PlGF treatment resulted in a ∼70% reduction of mature miR-199a (Fig. 2E, lane 2 versus lane 1), while transfection with ATF3 shRNA followed by PlGF treatment partially antagonized (∼70%) the expected repressive effect on the miR-199a2 level (Fig. 2E, lane 3 versus lane 2). Transfection of HMEC-1 cells with wt ATF3 plasmid resulted in a 60% reduction of miR-199a (Fig. 2E, lane 4 versus lane 1). The control scRNA did not affect miR-199a expression (Fig. 2E, lane 5 versus lane 1). Taken together these data showed that augmented expression of ATF3 either by PlGF treatment or by the ATF3 expression plasmid negatively regulated DNM3os and miR-199a2 transcription.

ATF3 acts as a repressor of DNM3os as demonstrated by DNM3os promoter analysis and chromatin immunoprecipitation assay.

In silico analysis of the DNM3os 5′-flanking region (∼2.0 kb) revealed the presence of seven potential ATF3 cis-binding elements; the two sites proximal to the transcription start site are depicted in the schematic of Fig. 3A. To establish the role of ATF3 in the regulation of DNM3os, HMEC-1 cells were transiently transfected with a wt DNM3os-luciferase (DNM3os-luc) reporter plasmid. PlGF treatment resulted in a ∼40% reduction of luciferase activity (Fig. 3B, lane 2 versus lane 1). Cotransfection of the reporter with ATF3 shRNA reversed the reduction of luciferase activity (Fig. 3B, lane 3 versus lane 2), while the effect of scRNA was not significant (Fig. 3B, lane 4 versus lane 2).

FIG 3.

ATF3 acts as a repressor of DNM3os, as demonstrated by DNM3os promoter analysis and chromatin immunoprecipitation assay. (A) Schematic of DNM3os promoter showing the locations and sequences of ATF3 binding sites; base substitutions are shown by asterisks corresponding to sites 1 and 2. (B) PlGF-mediated DNM3os promoter luciferase activity of full-length promoter (wt; 2.0 kb) and constructs mutated at ATF3 sites in the DNM3os promoter. HMEC-1 cells were transfected with the indicated wt DNM3os luciferase plasmid, mutant reporter plasmid, or the indicated shRNA along with the Renilla luciferase plasmid. Transfected cells were treated as indicated with PlGF for 24 h. The cell lysates were assayed for luciferase activity, and activity was normalized to Renilla activity. (C) Effect of ATF3 overexpression via an exogenous ATF3 expression plasmid on wt DNM3os luciferase reporter activity. Luciferase activity data are means ± SEM (n = 3). (D and E) ChIP analysis of ATF3 binding to native chromatin. HMEC-1 cells were transfected overnight with either an ATF3 shRNA or ATF3 expression plasmid. The soluble chromatin was isolated and immunoprecipitated with antibody to either ATF3 or control rabbit IgG. The primers employed (listed in Table 1) flank ATF3 binding sites (1 and 2), as indicated in schematic shown in panel A. Immunoprecipitated chromatin was processed for PCR analysis, and expected product sizes are indicated. The bottom panel shows amplification of the input DNA before immunoprecipitation. Data are representative of two independent experiments run in duplicate, and the indicated values are from densitometry values normalized to input DNA. *, P < 0.05; ***, P < 0.001; ns, not significant (P > 0.05).

Next, we generated base substitution mutations in the wt DNM3os reporter luciferase plasmid, wherein ATF3 binding sites proximal to the transcription start site of the DNM3os promoter were mutated as depicted in the schematic of Fig. 3A. In HMEC-1 cells transfected with mutated reporters with base substitutions at site 1 (ATF3mt1) (Fig. 3B, lanes 5 and 6 versus lane 4) or site 2 (ATF3mt2) (Fig. 3B, lanes 9 and 10 versus lane 8), neither PlGF treatment nor cotransfection with ATF3 shRNA affected luciferase activity. Moreover, transfection of HMEC-1 cells with the double ATF3 site mutations of the DNM3os promoter (ATF3DM) resulted in no change in luciferase activity in response to either PlGF or ATF3 shRNA (Fig. 3B, lanes 12 and 13 versus lane 11). Taken together, these results showed that ATF3 binding to the DNM3os promoter participated in PlGF-dependent repression of DNM3os transcription.

Additionally, we examined the effect of exogenous ATF3 on DNM3os reporter activity. As shown in Fig. 3C, cotransfection with an ATF3 expression plasmid reduced wt DNM3os reporter activity by ∼60% (Fig. 3C, lane 2 versus lane 1), while scRNA had no significant effect (Fig. 3C). To determine whether ATF3 bound to the endogenous DNM3os promoter of HMEC-1 cells, we performed ChIP (chromatin immunoprecipitation) assays on chromatin obtained from control (untreated) cells and cells transfected with either an ATF3 expression plasmid or ATF3 shRNA. Chromatin samples were immunoprecipitated with antibody to either ATF3 or a nonspecific rabbit IgG (control). HMEC-1 cells transfected with the ATF3 expression plasmid showed a ∼1.7-fold increase in the expected PCR product size for site 1 (91 bp), corresponding to the TSS-proximal site of the DNM3os promoter (nt −98 to −8) (Fig. 3D, ATF3 versus the control). Conversely, transfection with ATF3 shRNA modestly (∼20%) reduced basal ATF3 binding to site 1 (Fig. 3D, ATF3 shRNA versus the control). Similarly, chromatin obtained from HMEC-1 cells transfected with ATF3 shRNA showed reduced (∼60%) binding of endogenous ATF3 to site 2 corresponding to nt −912 to −798 of the DNM3os promoter (refer to the schematic of Fig. 3A), as shown by reduced recovery of the expected PCR product (115 bp) (Fig. 3E, ATF3 shRNA versus the control). The amplification of input DNA before immunoprecipitation of the chromatin samples (Fig. 3D and E, ATF3 lanes) was equivalent in all samples. Immunoprecipitation of chromatin samples with control rabbit IgG did not recover any DNA with the expected PCR product sizes (Fig. 3D and E, bottom rows). These data showed preferential binding of ATF3 to the proximal site (nt −98 to −8) of the DNM3os promoter in vivo, consistent with ATF3-mediated repression of DNM3os transcription.

Characterization of accessory proteins of ATF3 involved in transcriptional repression of DNM3os by proteome analysis.

Previous studies have shown that ATF3 can act either as a repressor or activator of transcription by recruitment of partner proteins to a gene promoter of interest (31). Thus, HMEC-1 cells were analyzed after exogenous enrichment of ATF3 by transfection with an ATF3 expression plasmid after treatment with PlGF or without treatment. For this analysis, the cell lysate was immunoprecipitated with antibody to either ATF3 or normal rabbit IgG (control). The immune complexes were purified utilizing protein A/G-agarose beads, and protein-bound beads were subjected to proteolytic digestion followed by mass spectroscopy and proteomic analysis. Analysis of the anti-ATF3 immune complexes showed the presence of >80 proteins/polypeptides in the coimmunoprecipitates (data not shown); the highest scoring candidates are listed in Tables 2 and 3. These data showed significant similarity between the highest scoring candidates for ATF3 coimmunoprecipitates obtained from cells either supplemented with the exogenous ATF3 expression plasmid or treated with PlGF. Based on this information, HMEC-1 cells transfected with the ATF3 expression plasmid could be used as a surrogate system for PlGF treatment. Based on a gene ontology bias, we focused our analysis on candidate ATF3-interacting proteins (JDP2, ATF2, and HDACs) as potential ATF3 binding partners involved in transcriptional regulation of DNM3os.

TABLE 2.

Proteomic analysis of HMEC-1 cells transfected with an ATF3 expression plasmid, followed by immunoprecipitation with ATF3 antibody and mass spectroscopic analysis^a

Protein name	Gene symbol	NCBI version no.	Sequence coverage (%)b	SEQUEST score (Xcorr)c	No. of unique peptidesd
Activating transcription factor 2	ATF2	NP_001871.2	48	51	13
Activating transcription factor 3	ATF3	NP_001025458.1	44	11	2
Jun dimerization protein 2	JDP2	BAB83896.1	35	7	3
Histone deacetylase 6	HDAC6	NP_001308156.1	45	141	36
Histone deacetylase 7	HDAC7	AAF63491.1	26	36	14
Mitogen-activated protein kinase 8	MAPK8	P45983.2	49	40	15
Activator protein 1 (c-Jun)	JUN	P05412.2	25	11	3

^aHMEC-1 cells transfected with an ATF3 expression plasmid were solubilized in RIPA buffer and immunoprecipitated with ATF3 antibody, and immunoprecipitates were subjected to proteolytic digestion and cleanup on a C₁₈ column followed by MS/MS, as described in Materials and Methods. The data were analyzed using the program Proteome Discoverer, version 1.4.

^bSequence coverage is the percentage of the protein sequence covered by known identified peptides.

^cThe SEQUEST score is used for analysis of peptide tandem mass spectra using a cross-correlation (Xcorr) scoring routine to match tandem spectra to model spectra derived from peptide sequences, as described in Materials and Methods.

^dThe number of peptide sequences unique to a protein group.

TABLE 3.

Proteomic analysis of HMEC-1 cells treated with PlGF, followed by immunoprecipitation with ATF3 antibody and mass spectroscopic analysis^a

Protein name	Gene symbol	NCBI version no.	Sequence coverage (%)	SEQUEST score (Xcorr)	No. of unique peptides
Activating transcription factor 2	ATF2	NP_001871.2	13	9	4
Activating transcription factor 3	ATF3	NP_001025458.1	15	4	1
Jun dimerization protein 2	JDP2	BAB83896.1	10	3	1
Histone deacetylase 6	HDAC6	NP_001308156.1	20	48	15
Histone deacetylase 7	HDAC7	AAF63491.1	7	7	3
Mitogen-activated protein kinase 8	MAPK8	P45983.2	13	9	4
Activator protein 1 (c-Jun)	JUN	P05412.2	5	2	1

^aHMEC-1 cells treated with PlGF were solubilized in RIPA buffer and immunoprecipitated with ATF3 antibody, and the immunoprecipitate was subjected to proteolytic digestion, followed by MS/MS and proteomic analysis as described in the footnotes to Table 2 and in Materials and Methods. Sequence coverage, SEQUEST score, and number of unique peptides are as defined for Table 2.

Anti-ATF3-immune complexes from HMEC-1 cells transfected with an exogenous ATF3 expression plasmid were further analyzed by immunoblotting (IB), utilizing specific antibodies to ATF2, JDP2, and HDAC6 (Fig. 4A). The Western blots showed discernible protein bands for all three candidate binding partners, while immune complexes isolated with control IgG were negative (Fig. 4A). In an effort to determine whether these proteins formed multiprotein complexes with ATF3, immunoprecipitation was performed with anti-HDAC6, followed by immunoblotting for ATF2, ATF3, and JDP2 (Fig. 4B). The HDAC6 immune complex exhibited association with ATF3 and JDP2 but not ATF2 (Fig. 4B).

FIG 4.

Characterization of accessory proteins of ATF3 involved in binding to DNM3os promoter. (A to C) Characterization of ATF3 protein interactions with selected proteins (JDP2, ATF2, and HDAC6) as identified by coimmunoprecipitation. HMEC-1 cells were transfected with an ATF3 expression plasmid or treated with PlGF for 6 h, followed by cell lysis, immunoprecipitation (IP) with antibody to either ATF3 or HDAC6 or normal rabbit IgG, and Western blotting using antibodies to the indicated proteins. The data are representative of two independent experiments. (D) Binding of transcription factors ATF3, JDP2, HDAC6, c-Jun, HDAC7, and ATF2 to native DNM3os promoter in isolated chromatin as assessed by ChIP. HMEC-1 cells were treated with PlGF for 6 h. The soluble chromatin was immunoprecipitated with the indicated antibody or control rabbit IgG. The PCR primers employed flank the ATF3 binding sites (1 and 2) or the distal site (corresponding to a non-ATF3 binding site) in the DNM3os promoter (Fig. 3A) and are listed in Table 1. Immunoprecipitated chromatin was processed for PCR analysis, and expected product sizes are indicated. The bottom panels show amplification of the input DNA before immunoprecipitation. Data are representative of two independent experiments run in duplicate, and the indicated values are from densitometry values normalized to those of the untreated control.

Although ectopic ATF3 expression gave results similar to those with PlGF induction with respect to ATF3-associated proteins, it was necessary to confirm that these proteins associated with endogenous ATF3 following PlGF induction. The ATF3 immunoprecipitates from HMEC-1 cells treated with PlGF for 6 h showed by Western blotting the presence of JDP2 and HDAC6 (Fig. 4C, right panel). In the absence of PlGF, JDP2 and HDAC6 were notably absent in ATF3 immunoprecipitates (Fig. 4C, left panel). These results showed that ATF3 associated with HDAC6 and JDP2 as a possible ternary complex.

Binding of ATF3, JDP2, and HDAC6 to the DNM3os promoter as demonstrated by chromatin immunoprecipitation.

We further examined by ChIP analysis whether transcription factors ATF3, JDP2, and HDAC6 were present on the DNM3os promoter at locations centered near the two identified ATF3 binding sites (schematic shown in Fig. 3A) under basal conditions and in PlGF-treated cells. In response to PlGF induction, we observed a ∼2-fold increase in binding of ATF3, JDP2, and HDAC6 to ATF3 site 1 (nt −42 to −35) of the DNM3os promoter compared to levels observed under basal conditions. In contrast, ATF3 site 2 (nt −816 to −809) showed a 50% increase in ATF3 binding with a comparable 40% increase in HDAC6 association; however, significant JDP2 binding was not observed. Other transcription-associated factors were assayed by ChIP analysis of the DNM3os promoter as controls. HDAC7 showed a slight (20%) increase in association with ATF3 sites 1 and 2, whereas c-Jun binding decreased slightly for both ATF3 sites (Fig. 4D). ATF2 binding was not significant under either the basal or PlGF-induced condition. The observed ChIP recoveries of the DNM3os promoter were authentic since a similar analysis of a predicted, distal ATF3 binding site showed no significant chromatin recovery with the same panel of antibodies used for primary immunoprecipitation. ChIP controls for nonspecific IgG immunoprecipitation showed little or no specific DNA recovery; accordingly, inputs of basal and PlGF-induced chromatin material were essentially the same, as indicated by the bottom rows of each panel set (Fig. 4D). The data obtained from ChIP analysis indicated that JDP2, HDAC6, and ATF3 were colocated at or very near ATF3 site 1. ATF3 site 2 also showed significant association of ATF3 and HDAC6 but not of JDP2. These results are consistent with the observed repression of DNM3os transcription in response to PlGF whereby ATF3 sites proximal to the transcription site are populated by repressive DNA binding proteins.

HDAC and JDP2 participate in PlGF-mediated repression of DNM3os transcription.

Histone deacetylases (HDACs) are associated with repression of gene transcription following histone deacetylation, thereby promoting chromatin condensation (32). We examined possible roles of HDACs 1, 3, 4, 6, 7, 8, 9, and 11 in PlGF-mediated repression of DNM3os, utilizing commercially available HDAC shRNAs. As shown in Fig. 5A, PlGF treatment of HMEC-1 cells attenuated DNM3os RNA expression by ∼60%. This reduction was significantly reversed only by shRNAs for HDAC6 and HDAC7 (Fig. 5A).

FIG 5.

Histone deacetylase (HDAC) and JDP2 participate in PlGF-mediated repression of DNM3os and pre-miR-199a2 RNA transcription. (A and B) shRNAs to HDAC6 and HDAC7 reverse PlGF-mediated repression of DNM3os and pre-miR-199a2 RNA expression. HMEC-1 cells were transfected overnight with the indicated shRNAs or scrambled shRNA (scRNA), followed by treatment with PlGF for 6 h. RNA was isolated for qRT-PCR analysis. Data are means ± SEM (n = 3). (C) Effect of two different HDAC6 shRNA clones (shRNA1 and shRNA2) on HDAC6 protein expression in transfection of HMEC-1 cells as determined by Western blotting. The optical density data are normalized to β-actin values. HDAC6 shRNA1 is the same as that shown in panels A and B. The data are representative of two independent experiments. (D) Effect of transfection of JDP2 small interfering RNA (siRNA) on PlGF-mediated JDP2, DNM3os, and pre-miR-199a2 RNA expression (n = 3). *, P < 0.05; **, P < 0.01; ns, not significant (P > 0.05).

Since pre-miR-199a2 is part of the DNM3os transcription unit, we examined the synthesis of pre-miR-199a2 mRNA following transfection with HDAC shRNAs. As expected, the expression of pre-miR-199a2 was derepressed by HDAC6 and HDAC7 shRNAs but not by other HDAC shRNAs (Fig. 5B). The requirement of HDAC6 for repression of DNM3os transcription was supported by the corresponding 80% knockdown of the HDAC6 protein level by HDAC6 shRNA (Fig. 5C). It was possible that PlGF repression of DNM3os was due to changes in available HDAC6 and HDAC7; however, PlGF treatment did not reduce the HDAC6 or HDAC7 mRNA levels during 6 h of PlGF treatment (data not shown). Transfection of HMEC-1 cells with JDP2 shRNA followed by PlGF treatment reduced JDP2 mRNA expression by ∼60% (Fig. 5D). Furthermore, knockdown of JDP2 derepressed PlGF-mediated repression of DNM3os and pre-miR-199a2 (Fig. 5D). Taken together, the data showed that repression of DNM3os transcription, including pre-miR-199a2, involved JDP2 binding to the promoter of DNM3os and to HDAC6 and HDAC7, with the deacetylases involved in chromatin remodeling.

Tubacin, a putative inhibitor of HDAC6, reverses PlGF-mediated repression of DNM3os transcription.

To further characterize HDAC participation in DNM3os repression, we utilized chemical inhibitors selective for HDACs (33). Currently, selective chemical inhibitors are available for only a few HDACs. We tested the effects of trichostatin A (TSA), a broad-spectrum HDAC class I and II inhibitor, tubacin, a specific inhibitor of HDAC6 (34, 35), and mocetinostat (MGCD0103), a potent inhibitor for HDAC1 but not HDAC6 or -7. Inhibition of HDAC6 was observed to derepress DNM3os transcription, with optimal effects observed at drug concentrations of ≥5 μM (Fig. 6A). Both TSA and tubacin completely reversed PlGF-mediated repression of DNM3os transcription (Fig. 6B, lanes 3 and 4 versus lane 2), while mocetinostat or a vehicle (DMSO) control had no significant effect (Fig. 6B, lanes 5 and 6 versus lane 2). Moreover, the derepression of DNM3os by tubacin was locus specific as tubacin had no effect on DNM3 (sense strand) transcription in treated cells (data not shown). DNM3 is reportedly constitutively expressed under basal conditions in endothelial cells derived from human lungs (http://www.proteinatlas.org/ENSG00000197959-DNM3/tissue/lung). These results showed that PlGF-mediated repression of DNM3os was specific to this locus, with no long-range, untoward effect on the DNM3 sense strand promoter.

FIG 6.

DNM3os transcription requires HDAC6 and ATF3 and is inhibited by tubacin. (A) Dose-response effect of tubacin on DNM3os RNA expression. (B) Effect of trichostatin A (TSA; HDAC paninhibitor), tubacin (HDAC6 inhibitor), and mocetinostat (HDAC1 inhibitor) on DNM3os RNA expression. HMEC-1 cells were incubated with the indicated inhibitors for 30 min, followed by treatment with PlGF for 6 h and analysis of RNA by qRT-PCR. (C) Effects of HDAC inhibitors on wt DNM3os-luc and ATF3 mutated DNM3os-luc reporters as measured by luciferase activity. Cells were transfected with either a wt DNM3os-luc reporter plasmid or double mutation (DNM3osDM) of ATF3 sites in the DNM3os-luc vector, followed by incubation with the indicated inhibitors and treatment with PlGF for 24 h. Data are means ± SEM (n = 3). *, P < 0.05; **, P < 0.01; ***, P < 0.001; ns, not significant (P > 0.05).

Next, we utilized the wt DNM3os promoter-luciferase reporter to further study the effect of HDAC inhibitors. As shown in Fig. 6C, wt DNM3os-luc activity decreased ∼40% following PlGF treatment (Fig. 6C, lane 4 versus lane 1). The luciferase reporter activity was completely restored by both TSA and tubacin during PlGF treatment (Fig. 6C, lanes 5 and 6 versus 4). The repression of DNM3os by PlGF was dependent upon binding of ATF3 and presumably HDAC6 to the wt promoter (Fig. 3B). For this reason, we examined whether this promoter was affected by tubacin in the absence of ATF3 binding. The DNM3os double mutant reporter (double ATF3mt1/ATF3mt2 mutations) was tested for its response to tubacin. As shown in Fig. 6C, this mutant reporter was not repressed by PlGF treatment (Fig. 6C, lane 9 versus lane 7) and showed no significant change in activity in response to the addition of tubacin (Fig. 6C, lane 8 versus lane 7). From these results we concluded that HDAC6 required functional ATF3 binding sites in the DNM3os promoter and that any antagonism of PlGF effects by tubacin was not due to a general effect on transcription of the endogenous DNM3os gene or the wt DNM3os reporter construct.

Regulatory access of transcription factors to the DNM3os promoter is modified by chromatin structure (FAIRE).

To better understand the consequence of transacting factor binding to the DNM3os promoter during PlGF-mediated repression, we examined the chromatin status of the gene. We utilized FAIRE (formaldehyde-assisted isolation of regulatory elements)-qPCR methodology (28) to identify nucleosome-free, euchromatic regions of the DNM3os locus. PlGF-treated cells showed a ∼80% reduction of FAIRE signal at both ATF3 sites 1 and 2 compared to levels in untreated cells (Fig. 7A and B, PlGF versus the control); this was likely due to chromatin condensation, consistent with the observed repression of transcription. Conversely, treatment with tubacin in the presence of PlGF increased DNA recovery by ∼2-fold above the basal level for ATF3 site 1 and by ∼50% for ATF3 site 2 (Fig. 7A and B, respectively). Similarly, HMEC-1 cells transfected with ATF3 shRNA and HDAC6 shRNA showed significantly increased (∼2- to 3-fold) FAIRE signal compared to the signal with PlGF treatment alone (Fig. 7A and B, 4th and 5th lanes versus PlGF). Taken together, these data showed that PlGF-induced chromatin condensation repressed DNM3os transcription while tubacin prevented the repressive effect following PlGF signaling.

FIG 7.

PlGF induces chromatin condensation at proximal ATF3 sites in the DNM3os promoter following occupation by HDAC6 and ATF3 protein complexes, as characterized by FAIRE (formaldehyde-assisted isolation of regulatory elements)-qPCR and histone H3 acetylation marks. (A and B) Effects of tubacin, ATF3 shRNA, and HDAC6 shRNA on DNM3os promoter accessibility to ATF3 accessory proteins (i.e., HDAC6), as determined by FAIRE-qPCR signal. HMEC-1 cells were either cotreated with tubacin and PlGF for 6 h or transfected with the indicated shRNAs and then treated with PlGF for 6 h. Cell lysates were processed as previously described (28). Genomic DNA (input) and FAIRE DNA were subjected to quantitative PCR analysis, and results were normalized to those of untreated controls. (C and D) Effects of tubacin, ATF3 shRNA, and HDAC6 shRNA on histone acetylation marks H3K9Ac and H3K27Ac. HMEC-1 cells were either cotreated with tubacin and PlGF for 6 h or transfected with the indicated shRNA and then treated with PlGF for 6 h. Cell lysates were immunoprecipitated with antibodies against H3K9Ac, H3K27Ac, or control rabbit IgG. DNA samples were quantified by qRT-PCR analysis, and values were normalized to the input DNA. Data are means ± SEM (n = 3). *, P < 0.05; **, P < 0.01; ***, P < 0.001.

Deacetylation of histone H3 marks in the promoter of DNM3os is required for repression.

In silico analysis using ENCODE data of the DNM3os gene, located on chromosome 1, showed extensive histone acetylation marks for H3K27Ac in an ∼1.2-kb region centered on the TSS (https://genome.ucsc.edu/cgi-bin/hgTracks?db=hg38&position=chr1%3A172142217-172147725&hgsid=463512079_iALmHAFswXUUaHlV2xecqlvUAgqL). These histone marks are associated with chromatin remodeling events occurring at promoter and enhancer sites (36). Thus, we examined the presence of and changes in H3K9Ac and H3K27Ac marks in the promoter of DNM3os proximal to the identified ATF3 sites under basal conditions and under PlGF-mediated repression. Chromatin was isolated from HMEC-1 cells treated with PlGF, followed by immunoprecipitation with either anti-H3K9Ac or anti-H3K27Ac antibodies. The recovered DNA was analyzed by PCR using primers corresponding to the respective ATF3 binding sites in the DNM3os promoter (Table 1 and Fig. 3A). As shown in Fig. 7C, PlGF reduced from ∼40% to 5% the recovery of DNA with anti-H3K9Ac proximal to ATF3 site 1. DNA recovery was enhanced by inclusion of tubacin, ATF3 shRNA, and HDAC6 shRNA (Fig. 7C). Consistent with this result, PlGF treatment reduced from ∼12% to 2% the DNA recovery proximal to ATF3 site 2 (Fig. 7D). Similar to results at ATF3 site 1, DNA recovery increased from 2% in PlGF-treated cells to 20%, 5%, and 20% with tubacin, HDAC6 shRNA, and ATF3 shRNA, respectively, at site 2 (Fig. 7D).

The H3K27Ac mark was assayed for DNA recovery with anti-H3K27Ac as this modification is also involved in chromatin remodeling. As shown in Fig. 7C, PlGF treatment reduced recovery of DNA from ∼60% to ∼5%, proximal to ATF3 site 1 of the DNM3os promoter; DNA recovery of this segment was enhanced by treatment with tubacin, ATF3 shRNA, and HDAC6 shRNA. Similarly, PlGF attenuated DNA recovery from ∼30% to ∼5% proximal to ATF3 site 2 of the DNM3os promoter (Fig. 7D). DNA recovery was increased by coaddition of tubacin, HDAC6 shRNA, and ATF3 shRNA (Fig. 7D). Control incubations with nonspecific, primary rabbit IgG (Fig. 7C and D) were negative for DNA recovery. Together, these data showed that PlGF treatment reduced H3K27Ac and H3K9Ac histone marks at the DNM3os locus, consistent with chromatin condensation. Conversely, tubacin and HDAC6 shRNA treatment effectively antagonized this change to maintain an open chromatin state at the promoter-proximal region of DNM3os. In this system, ATF3, rather than acting as a general repressor, appears to be required for guiding chromatin remodeling enzymes (e.g., HDAC6) and/or other accessory proteins to effect the silencing of this promoter.

HIF-1α-dependent transcription can be modulated by ATF3 levels: effects on ET-1 expression.

Previous studies showed that PlGF induces HIF-1α expression, which in turn augments ET-1 expression (17); thus, we examined whether ATF3 activity was correlated with ET-1 mRNA and protein expression. PlGF induced expression of both HIF-1α and ET-1 mRNA (Fig. 8A, control versus PlGF), and the induction was inhibited by ATF3 shRNA (Fig. 8A, ATF3 shRNA versus PlGF). Moreover, PlGF augmented by ∼1.8-fold the secretion of ET-1 protein, and transfection with ATF3 shRNA followed by PlGF treatment reduced ET-1 secretion to the basal level (Fig. 8B, lane 3 versus lane 2). In the absence of PlGF, expression of exogenous ATF3 but not ΔATF3 led to increased ET-1 secretion (Fig. 8B, lane 4 versus lane 5). These data showed that changes in ATF3 levels led to proportional effects on ET-1 secretion from HMEC-1 cells.

FIG 8.

PlGF-mediated upregulation of ATF3 induces HIF-1α and its target endothelin-1 gene. (A) ATF3 shRNA attenuates PlGF-induced expression of both HIF-1α and ET-1 mRNAs in HMEC-1 cells. (B) ATF3 levels modulate secretion of ET-1 protein. HMEC-1 cells were transfected with either ATF3 shRNA or an ATF3 expression plasmid or ΔATF3 (lacking the leucine zipper domain). Cells were washed and either left untreated or treated with PlGF for 24 h. The culture supernatants were assayed for secreted ET-1 by ELISA, and data were normalized to the amount of total isolated protein. (C) Effects of tubacin and shRNAs for HDAC6 and ATF3 on PlGF-induced secretion of ET-1. Data are means ± SEM (n = 3). *, P < 0.05; **, P < 0.01; ***, P < 0.001; ns, not significant (P > 0.05).

Chromatin status indirectly affects ET-1 expression and secretion.

The DNM3os locus was demonstrated to undergo repression by a change in chromatin state/status. As shown above, HDAC6 activity was important for ATF3-mediated repression of DNM3os/miR-199a2 transcription. Thus, we determined whether tubacin affected expression of ET-1. As shown in Fig. 8C, PlGF-induced secretion of ET-1 was antagonized by tubacin treatment and HDAC6 shRNA. The effects of tubacin and HDAC6 shRNA are consistent with chromatin changes leading to conditions permissive for increased DNM3os/miR-199a2 transcription with concomitant reductions of HIF-1α and ET-1.

ATF3 expression in lung tissues of Berkeley sickle cell (BK-SS) mice.

Since PlGF-mediated upregulation of ATF3 leads to repression of miR-199a2 and concomitant upregulation of ET-1 expression in vitro, we examined ATF3 levels in lung tissues of BK-SS mice. As shown in Fig. 9A, there was a ∼4-fold increased ATF3 mRNA levels compared to those in control C57BL/6NJ mice (n = 6). Moreover, ATF3 protein levels were also higher in lung tissues of BK-SS mice than in control C57BL/6NJ mice (Fig. 9B). As previously reported, lung tissues of BK-SS mice show lower levels of pre-miR-199a2 and higher levels of ET-1 mRNA than control animals (18). These results link the increased ATF3 expression in lungs of BK-SS mice to reduced DNM3os/miR-199a2 transcription and resulting increased expression of ET-1 mRNA and protein.

FIG 9.

High levels of ATF3 are observed in lung tissues of BK-SS mice. Tubacin administration reduced plasma PlGF and ET-1 levels and ET-1 mRNA in lung tissues. (A and B) Comparison of ATF3 mRNA and protein in lung homogenates from BK-SS and C57BL/6NJ mice. Total RNA and protein were isolated from lung homogenates and subjected to qRT-PCR or Western blotting (C57BL/6NJ mice, n = 6; BK-SS mice, n = 6). Quantification of ATF3 protein by densitometry was normalized to the levels of β-actin in samples from control and BK-SS mice. (C to F) Effects of tubacin and niltubacin (inactive tubacin congener) on ET-1 and PlGF plasma levels (C and D) and on ET-1 (E) and HIF-1α (F) mRNA expression in lung tissues of BK-SS mice and control C57BL/6NJ mice. Tubacin or niltubacin (0.5 mg/kg body weight) was administered i.p. to 4- to 6-month-old BK-SS mice for 5 days (n = 3) and 10 days (n = 6). For 30-day continuous delivery of drugs (daily dosage, 0.33 mg/kg body weight), mini-osmotic pumps were used in BK-SS mice for tubacin treatment (n = 10) and niltubacin treatment (n = 9). The plasma was collected for ET-1 and PlGF ELISAs, and lung tissues were homogenized for qRT-PCR of isolated RNA. *, P < 0.05; ***, P < 0.001.

Administration of tubacin, a selective HDAC6 inhibitor, to BK-SS mice attenuates ET-1 and PlGF levels in lung tissues and plasma.

Tubacin, a selective inhibitor of HDAC6, and niltubacin (an inactive tubacin congener) (35) were administered to BK-SS mice daily at 0.5 mg/kg body weight via intraperitoneal (i.p.) injection for 5 or 10 days. Alternatively, drug delivery was performed via subcutaneous osmotic pump for 30 days. These drug regimens with tubacin showed a time-dependent reduction of PlGF (Fig. 9C) and ET-1 (Fig. 9D) in plasma compared to levels with niltubacin. The 30-day drug treatment effectively reduced plasma levels of PlGF by ∼60% (P < 0.001) (Fig. 9C) and plasma levels of ET-1 by ∼60% (P < 0.001) (Fig. 9D). After 30 days of drug administration, lung tissues showed a ∼40% reduction of ET-1 mRNA (P < 0.05) (Fig. 9E) in response to tubacin although HIF-1α mRNA levels were not significantly changed (Fig. 9F) compared to those in niltubacin-treated BK-SS mice. Although we were limited regarding isolation of various cellular subpopulations from drug-treated lungs, it was likely that decreased ET-1 levels arose predominantly from pulmonary endothelial cells, as previously shown (21). Taken together, these data showed that tubacin, a selective inhibitor of histone deacetylase 6 (HDAC6), attenuated ET-1 and PlGF levels in BK-SS mice by upregulating DNM3os/miR-199a2/miR-214 transcription.

DISCUSSION

Our previous studies in cultured endothelial cells showed that PlGF induces ET-1 expression via activation of HIF-1α independently of hypoxia (17). Moreover, our studies showed that PlGF attenuates levels of pre-miR-199a2 and subsequently mature miR-199a2 in vitro (18). In the BK-SS mouse model and in SCD patients, we showed that plasma levels of miR-199a2 are significantly reduced compared to those of normal controls (18). This is significant because miR-199a2 targets the 3′ UTR of HIF-1α mRNA and attenuates HIF-1α expression (18). In light of this regulatory loop, a consequence of PlGF-treatment of HMEC-1 cells and abnormally high PlGF levels, as seen in the SCD mouse model, we anticipated increased levels of HIF-1α with ET-1 expression, as was observed (18). Both in vitro and in vivo results showed that a consequence of reduced expression of miR-199a2 was increased expression of HIF-1α, especially in SCD, with a concomitant increase in HIF-1α-dependent gene expression, e.g., ET-1.

Transcription of DNM3os produces a long noncoding RNA, which is the primary transcript for miR-199a2 and miR-214 (18, 37, 38). We have extended our previous studies by providing mechanistic detail for negative regulation of DNM3os transcription mediated by PlGF. In this study, we show that PlGF-mediated induction of activated transcription factor 3 (ATF3) expression participated in repression of DNM3os transcription. In silico analysis of the DNM3os promoter showed seven putative cis-binding sites for ATF3; of these, the two sites nearest the transcription start site of DNM3os were required for PlGF-mediated repression. ATF3, as a member of the ATF/CREB family of basic leucine zipper transcription factors, is induced by an array of signals originating from cytokines, physiological changes, and pathological stimuli (20, 39–45). ATF3 protein can form heterodimers with AP-1 complex, JDP2, and ATF2 (19, 41, 42, 44, 46). This was corroborated by our finding that ΔATF3 lacking the basic leucine zipper domain was nonfunctional in the luciferase reporter assays.

The precise nature of the ATF3 complex involved in repression of DNM3os transcription was investigated further by proteomic analysis. Analysis of ATF3 immune complexes by mass spectroscopy revealed ∼80 polypeptides that associated either directly or indirectly with ATF3. A gene ontology classification allowed us to focus on proteins known to be regulators of transcription. Previous studies have shown that ATF3 forms homodimer and heterodimer complexes with basic leucine zipper proteins to regulate target genes both positively and negatively (31, 47, 48). Our results showed that ATF3 associated with ATF2, JDP2, HDAC6, and HDAC7 (Tables 2 and 3). The candidate ATF3 binding partners were thus examined as coregulators of DNM3os transcription. Our results showed that ATF2, JDP2, and HDAC6 were capable of associating with ATF3 in response to PlGF-mediated repression of DNM3os. Whether separate bipartite or higher-order complexes are required to repress DNM3os transcription was not evident, and further studies are warranted.

In order to gain insight into potential epigenetic regulation of DNM3os, chromatin modification events at the DNM3os promoter were examined utilizing the FAIRE methodology. Our studies showed that PlGF-mediated upregulation of ATF3 was associated with chromatin condensation encompassing the two TSS-proximal ATF3 binding sites of the DNM3os promoter and concomitant repression of DNM3os transcription.

Numerous studies have established that acetylation of core histones is associated with chromatin remodeling and transcriptional activation. In contrast, deacetylation of core histones is associated with changes in histone methylation favoring chromatin condensation and associated gene silencing (32, 36). Histone deacetylases (HDACs) modify nucleosome core histones and nonhistone proteins by removal of an acetyl group from specific lysine residues (49). Our studies showed that HDAC6 and HDAC7 were involved in PlGF-mediated repression of DNM3os. Repression was not dependent on increased expression of HDAC6 or HDAC7 mRNAs by PlGF, as seen for hypoxia-mediated upregulation of HDAC6 (50), thus implying some level of translational regulation. Moreover, tubacin, a selective inhibitor of HDAC6 (34, 51), interfered with DNM3os repression, but mocetinostat, which is an HDAC class I inhibitor, did not. The effect of tubacin was highly localized as transcription of the overlapping DNM3 gene on the sense strand was unaffected. Studies show that tubacin is specific for the tubulin deacetylase activity of HDAC6 and inhibits HDAC6 by increased acetylation of tubulin (34).

Further investigation into the nature of the repressor complex showed that HDAC6 in association with ATF3 was localized at or near the ATF3 binding sites adjacent to the TSS of DNM3os. We concluded from this result that these sites were crucial for chromosome remodeling initiated by HDAC6 and ATF3. Results of the ENCODE project showed that the DNM3os gene has extensive histone H3K27Ac acetylation marks flanking ∼0.5 kb (nt 172113424 to 172114622) in both 5′ and 3′ directions from the TSS of DNM3os. We observed that PlGF treatment resulted in loss of H3K27Ac and H3K9Ac marks proximal to the ATF3 sites of the DNM3os promoter, associated with chromatin condensation in that region and concomitant attenuation of DNM3os transcription. Conversely, silencing of HDAC6 or treatment with tubacin antagonized PlGF-induced H3K27Ac and H3K9Ac deacetylation and maintained open chromatin, allowing DNM3os transcription and miR-199a2 synthesis.

Next, we examined the lung tissues, the affected organ in pulmonary hypertension, of BK-SS mice and normal mice. As anticipated, ATF3 mRNA levels were significantly higher in lung tissues from BK-SS mice than in control mice. Moreover, pre-miR-199a2 levels were reduced in lungs of BK-SS mice compared to levels in control mice (18) and correlated with increased expression of ET-1. Thus, both in vitro and in vivo studies were consistent in that increased levels of ATF3 were associated with reduced expression of miR-199a2, leading to augmented expression of HIF-1α and its target, the ET-1 gene.

Since tubacin maintained the basal chromatin state of DNM3os in the presence of PlGF in vitro, we examined the effect of tubacin in 4- to 6-month-old sickle mice (BK-SS). Our studies showed that administration of tubacin at 0.33 mg/kg body weight to BK-SS mice for 30 days was effective in reducing plasma levels of ET-1 by ∼50% and PlGF by >80%. Furthermore, lungs from tubacin-treated BK-SS mice showed reduced ET-1 expression. This outcome likely resulted from increased transcription of DNM3os leading to synthesis of miR-199a2 and miR-214. Our recent studies have shown that miR-199a2 and miR-214 target the 3′ UTRs of HIF-1α and PlGF mRNAs, respectively (18, 52). Thus, allowing continued basal synthesis of miR-199a2 in the presence of tubacin, as observed both in vitro and in vivo, resulted in reduced HIF-1α-mediated ET-1 expression. Similarly, continued synthesis of miR-214 culminated in reduced PlGF levels as seen in vivo.

In conclusion, our studies showed that expression of activated transcription factor 3 (ATF3) is upregulated by placenta growth factor (PlGF) in cultured endothelial cells and in Berkeley sickle mice, which exhibit high plasma PlGF levels resulting from increased erythropoiesis (12, 16). The upregulation of ATF3 acts to decrease transcription of DNM3os and miR-199a2/miR-214 through ATF3 cis-binding elements proximal to the TSS of DNM3os (Fig. 10). ATF3-mediated repression of DNM3os occurred primarily by epigenetic changes leading to chromatin remodeling in the DNM3os promoter. Last, our studies showed that tubacin treatment of Berkeley sickle mice augmented expression of the DNM3os/miR-199a2 axis, with the beneficial outcome of reducing HIF-1α and associated ET-1 levels. These studies provide a rationale for a new therapeutic approach for treatment of SCD symptoms in human patients. Tubacin is currently in clinical trials for cancer patients; therefore, its use in human SCD patients is possible since safety issues have already been addressed (33).

FIG 10.

Schematic of PlGF-induced ATF3 expression and its role in DNM3os/miR-199a2 repression. Placenta growth factor elaborated by bone marrow erythroid progenitors in response to erythropoietin and hypoxia in SCD activates endothelium, leading to upregulation of ATF3 and activation of HIF-1α. HIF-1α upregulates the expression of ET-1. ATF3 acts as a repressor in association with HDAC6 and JDP2 on the promoter of DNM3os/miR-199a2. miR-199a2 is within the DNM3os transcription unit. Tubacin, a selective inhibitor of HDAC6, allows the expression of DNM3os/miR-199a2 both in vitro and in vivo. miR-199a2 targets the 3′ UTR of HIF-1α, leading to its downregulation with concomitant attenuation of the downstream target ET-1 gene in SCD.

Funding Statement

This work, including the efforts of Punam Malik and Vijay K. Kalra, was funded by HHS | National Institutes of Health (NIH) (RO1-HL111372). This work, including the efforts of Vijay K. Kalra, was funded by HHS | National Institutes of Health (NIH) (P30-DK048522) awarded to Neil Kaplowitz.

The content is solely the responsibility of the authors and does not necessarily represent the official views of NHLBI or the National Institutes of Health.