|Home | About | Journals | Submit | Contact Us | Français|
A hallmark of dysfunctional fibroblast to myofibroblast differentiation associated with fibrotic disorders is persistent expression of ACTA2, the gene encoding the cyto-contractile protein smooth muscle α-actin. In this study, a PURB-specific gene knockdown approach was used in conjunction with biochemical analyses of protein subdomain structure and function to reveal the mechanism by which purine-rich element binding protein B (Purβ) restricts ACTA2 expression in mouse embryo fibroblasts (MEFs). Consistent with the hypothesized role of Purβ as a suppressor of myofibroblast differentiation, stable short hairpin RNA-mediated knockdown of Purβ in cultured MEFs promoted changes in cell morphology, actin isoform expression, and cell migration indicative of conversion to a myofibroblast-like phenotype. Promoter-reporter assays in transfected Purβ knockdown MEFs confirmed that these changes were attributable, in part, to de-repression of ACTA2 transcription. To map the domains in Purβ responsible for ACTA2 repression, several recombinant truncation mutants were generated and analyzed based on hypothetical, computationally-derived models of the tertiary and quaternary structure of Purβ. Discrete subdomains mediating sequence- and strand-specific cis-element binding, protein-protein interaction, and inhibition of a composite ACTA2 enhancer were identified using a combination of biochemical, biophysical, and cell-based assays. Our results indicate that the Purβ homodimer possesses three separate but unequal single-stranded DNA-binding modules formed by subdomain-specific inter- and intramolecular interactions. This structural arrangement suggests that the cooperative assembly of the dimeric Purβ repressor on the sense strand of the ACTA2 enhancer is dictated by the association of each subdomain with distinct purine-rich binding sites within the enhancer.
The myofibroblast is a unique cell type that exhibits an ensemble of phenotypic properties typical of a collagenous matrix-producing fibroblast and a contractile smooth muscle cell (1). In the body, pre-formed myofibroblasts play an important structural role in certain developing and adult tissues and organs (1, 2). On the other hand, emergent myofibroblasts are critical to the formation and remodeling of granulation tissue during wound healing as they provide the contractile machinery and mechanical strength necessary for wound closure (3-5). While transient differentiation of resident connective tissue fibroblasts to myofibroblasts is a normal physiological response to tissue injury, persistent myofibroblast activation is associated with hypertrophic scarring, pathologic organ fibrosis, aberrant vascular remodeling, and dysfunctional stromal responses to neoplasia (6-9). Consequently, an improved understanding of the molecular mechanisms underlying myofibroblast trans-differentiation may reveal novel drug targets to limit scarring, fibrosis, and tumor progression.
Among the markers of myofibroblast conversion, expression of ACTA2, the gene encoding smooth muscle α-actin (SMαA)3 is recognized as one of the key determinants of the transition to a contractile phenotype (10-12). Based largely on comparing ACTA2 reporter gene activity in myogenic versus non-myogenic cell lines, early reports suggested that activation of ACTA2 transcription in fibroblasts is mediated by serum-derived, growth factor-dependent signaling leading to induction of an otherwise repressed 5′ enhancer-promoter (13-15). Later studies revealed that the 5′ flanking region of ACTA2 contains a variety of discrete but functionally-interacting cis-elements that serve as binding sites for certain muscle-associated, growth factor-inducible, or basal trans-activators found in ACTA2-expressing fibroblasts (16-19). In particular, combinatorial interactions between a transcription enhancer factor 1 (TEF1)-binding muscle CAT (MCAT) motif, two serum response factor (SRF)-interacting CArG boxes, and several specificity protein 1 and 3 (Sp1/3)-binding GC-rich elements are necessary to drive high level ACTA2 transcription in differentiating myofibroblasts (17, 20). Conversely, in undifferentiated fibroblasts, the activity of a composite MCAT/CArG/GC box enhancer is apparently suppressed by several single-stranded DNA (ssDNA)-binding repressors that interact with the opposing strands of an asymmetric polypurine/polypyrimidine-rich (Pur/Pyr) tract containing the core MCAT motif (15, 21). Cell-based promoter mutagenesis studies in conjunction with nucleoprotein interaction analyses with double-stranded and single-stranded probes led to the identification of purine-rich element binding proteins A and B (Purα and Purβ) and Y-box binding protein 1 (YB-1) as the key factors in strand-specific Pur/Pyr tract recognition and repression of the composite ACTA2 enhancer (17, 22).
Purα and Purβ are members of a small family of nucleic acid-binding proteins that interact with purine-rich ssDNA or RNA sequences homologous to the so-called PUR element originally described in eukaryotic gene flanking regions and origins of DNA replication (23-25). Despite the fact that Purα and Purβ share ~70% sequence identity and exhibit similar ssDNA-binding and helix-destabilizing properties in vitro (26-28), comparative gain-of-function and loss-of-function analyses conducted in transiently-transfected fibroblasts and vascular smooth muscle cells point to Purβ as the dominant repressor of ACTA2 in these cell types (29, 30). In keeping with its general biological role as a potent repressor of genes that encode contractile proteins, Purβ has also been reported to negatively-regulate MYH6 and MYH7 in cardiac and skeletal myocytes (31-33). More recent studies suggest that Purβ repressor expression in muscle cells is tightly controlled at the post-transcriptional level by certain muscle-restricted microRNAs to ensure appropriate myofiber composition for sustained cardiac and skeletal muscle performance in response to stress (34, 35).
Apart from hydrodynamic analyses revealing that Purβ can reversibly self-associate to form an elongated homodimer in the absence of ssDNA (36), comparatively little is known about the higher order structural domains in either the Purβ monomer or dimer that confer specific and high-affinity interaction with purine-rich elements in ACTA2 or any other target gene. A previous report demonstrated that Purβ interacts in a sequential and cooperative manner with the sense strand of the MCAT-containing Pur/Pyr element from mouse ACTA2 to form a high affinity 2:1 Purβ:ssDNA complex (37). While the primary structure of Purβ is similar to Purα in terms of the presence of three distinct regions of internal homology (dubbed PUR repeats I, II, and III) (38), Purβ contains several unique intervening sequences with high glycine and proline content that may affect the structural and functional properties of the protein (26). Importantly, the x-ray crystal structure of a truncated version of Drosophila melanogaster (Dm) Purα (amino acids 40-185) revealed a monomeric Whirly fold-like DNA-binding domain formed by the intramolecular interaction of the first two PUR repeat sequences (39). On the other hand, we recently identified a core tryptic fragment of Purβ (amino acids 29-305) that contains all three PUR repeats, self-associates in the absence of nucleic acid, and retains the ability to interact with the purine-rich strand of the ACTA2-derived MCAT element with high-affinity and specificity (40). In this study, the putative biological role of Purβ in suppressing ACTA2 expression and restricting myofibroblast cyto-differentiation was first validated via a stable gene knockdown approach. In silico modeling of protein structure coupled with empirical analyses of protein function were then used to delineate the relevant domains in Purβ that mediate ACTA2-specific nucleoprotein interaction and repression of the composite MCAT/CArG/GC box enhancer.
AKR-2B MEF cell lines stably-transduced with lentiviral vectors encoding a PURB transcript-specific short hairpin RNA (shRNA) or a scrambled control RNA were generated as described in (30). Subcloned cell lines were propagated in McCoys 5A medium (Gibco™/Invitrogen) containing 5% heat-inactivated fetal bovine serum (FBS) and 10 μg/ml blasticidin in a humidified 5% CO2 incubator and studied at passage number 5 to 15. Phase contrast images of live cells were obtained on a Zeiss Axiovert model 200 inverted microscope equipped with an AxioCam MRm digital camera. Assays used to compare the growth and migratory properties of derived cell lines are detailed in Supporting Information. In timed growth factor-treatment experiments, cells were seeded at a fixed density and then switched to serum-free MCDB-402 medium (JRH Biosciences) for 36 to 48 h. Cells were then treated for 24 h with either 10% FBS or 2.5 ng/ml recombinant human transforming growth factor β1 (TGF-β1) (R & D Systems) diluted in MCDB-402 medium. Confluent monolayers of growth factor-stimulated cells were washed three times with ice-cold phosphate buffered saline (PBS) and then extracted with 1× Reporter Gene Assay Lysis Buffer (Roche Applied Science) supplemented with protease inhibitors 0.5 mM phenylmethanesulfonyl fluoride and 1 μg/ml each of pepstatin A, leupeptin, and aprotinin. Soluble lysates and cell remnants were collected by centrifugation at 15,800 × g for 10 min at 4°C. Total protein content in cleared lysates was measured by BCA™ Protein Assay (Thermo Scientific) using bovine serum albumin (BSA) as a standard. Insoluble pellets were further extracted with a denaturing solvent consisting of 8 M urea, 100 mM sodium phosphate, 10 mM Tris-Cl pH 8.0 plus protease inhibitors. Denatured lysates were cleared by centrifugation and assayed for protein content as described above.
Murine ACTA2 promoter-chloramphenicol acetyltransferase reporter constructs (pVSMP8-CAT and pVSMP4-CAT) have been described elsewhere (14, 30, 41). The corresponding ACTA2 promoter-luciferase reporters were constructed as follows. A ~3.6 kb fragment was released from pVSMP8-CAT by sequential treatment with SphI, mung bean nuclease, and BamHI. A ~240 bp insert was released from pVSMP4-CAT by sequential treatment with SalI, mung bean nuclease, and BamHI. Restriction fragments were ligated into SmaI/BglII-digested and alkaline phosphatase-treated pGL3-Basic vector (Promega) to generate pVSMP8-Luc and pVSMP4-Luc. Following transformation into E. coli HB101 cells, ampicillin-resistant clones were selected for propagation and plasmid purification (Roche Applied Science). The fidelity of plasmid constructs was confirmed by restriction enzyme digestion followed by automated DNA sequencing performed by the Vermont Cancer Center DNA Analysis Facility.
Bacterial and mammalian expression vectors encoding full-length, N-terminal hexahistidine-tagged mouse Purβ (pQE30-NHis-Purβ and pCI-NHis-Purβ) were described in previous reports (17, 22). Expression plasmids encoding NHis-Purβ truncation proteins corresponding to amino acids 41-112 (Purβ I), 125-210 (Purβ II), 209-303 (Purβ III), 41-210 (Purβ I-II), 125-303 (Purβ II-III), and 41-303 (Purβ I-II-III) were constructed following a similar strategy outlined in (29) and as further detailed in Supporting Information.
With the exception of antibody screening assays, all other procedures involved in generating murine and rat monoclonal antibodies (mAbs) recognizing Purα and/or Purβ were carried out by a commercial vendor (Green Mountain Antibodies). These included rodent immunization, fusion of mouse or rat splenocytes with NS-1 myeloma cells, subcloning of hybridoma cells (2 rounds), and in vitro production, purification, and isotype/subclass determination of derived mAbs. The specific antigens used for immunization were keyhole limpet hemocyanin-coupled peptides corresponding to amino acid sequences B42-69, B302-324, and A291-313 of mouse Purβ and Purα, respectively (22). Animals were immunized with each individual peptide or a combination of all three peptides. The relative affinity and specificity of purified mAbs were evaluated by direct enzyme-linked immunosorbent assay (ELISA) using immobilized peptides or full-length NHis-Purα or NHis-Purβ as described in Supporting Information and shown in Figure S1.
A homology model of Mus musculus (Mm) Purβ was generated based on the structure of Drosophila melanogaster (Dm) Purα repeats I-II (amino acids 40-185) using web-based modeling servers and a bioinformatics approach described previously (38, 39). A pair-wise multiple sequence alignment of Mm Purβ and Dm Purα indicating 52% sequence identity was obtained using ClustalW (42, 43). The sequence alignment was submitted to the HHrepID and PSIPRED web servers (44, 45) under default parameters to identify repeated sequences and predict secondary structural elements within Purβ (26). To generate a model consistent with these results, the FASTA sequence of Mm Purβ was submitted to the I-TASSER web server (46, 47) to generate a homology model of Purβ repeats I-II using Dm Purα I-II (3K44) as a template (39). On the basis of the internal sequence homology of the three Purβ repeats, Purβ III was independently predicted by SWISS-MODEL (48-50) by threading the sequence of the third Purβ repeat onto the structure of Dm Purα I-II. Predicted intermolecular interaction between two Purβ III repeats was modeled by rotating the Mm Purβ III repeats of two monomers into an extended conformation and aligning them on Dm Purα I-II using Coot (51). Energy minimization of the Purβ homodimeric homology model was completed using CNS version 1.2 (52, 53) to relax close contacts and to regularize local bond and angle geometry. Computationally-derived structures were viewed and depicted using PyMOL (54).
Full-length NHis-Purβ and selected truncation proteins were expressed in and isolated from E. coli JM109 cells using chromatographic methods optimized for each particular recombinant protein as detailed in Supporting Information. NHis-Purβ-enriched fractions obtained by metal chelate affinity, heparin affinity, or size exclusion chromatography (SEC) were monitored for purity by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) under reducing conditions and staining with Coomassie® Brilliant Blue R-250. Wide range SigmaMarker™ proteins (Sigma-Aldrich) were used as molecular weight standards. The protein concentration of pooled fractions was determined by absorbance measurement using theoretical molar extinction coefficients at 280 nm of 20,400 M−1 cm−1 for full-length NHis-Purβ and NHis-Purβ I-II-III, 7,450 M−1 cm−1 for NHis-Purβ I-II, and 12,950 M−1 cm−1 for NHis-Purβ III (55). Protein preparations were routinely screened to ensure the absence of contaminating nucleic acid as previously described (36). Protein preparations were also monitored for the presence of nuclease activity by incubating 1.0 μM protein stocks with 2 μg of either pBLCAT3 plasmid or M13mp18 ssDNA (Bayou Biolabs) for 1 h at 37°C. The integrity of the DNA substrates was then evaluated by agarose gel electrophoresis. Absence of contaminating nuclease activity in recombinant Purβ preparations was established based on comparison to DNA substrates treated in parallel with 10−1 to 10−5 units of DNase I (Invitrogen).
The quaternary state of purified truncation proteins was determined by SEC on a 1.5 × 100 cm Sephacryl® 200 HR column calibrated with molecular weight standards blue dextran, BSA, ovalbumin, carbonic anhydrase, cytochrome C, and DNP aspartate (36). The foldedness of recombinant proteins was assessed by CD spectroscopy. All proteins were analyzed on a Jasco model 815 spectrometer after dialysis into buffer consisting of 50 mM Tris-Cl pH 7.5, 300 mM NaCl, 0.5 mM EDTA, and 0.5 mM tris(2-carboxyethyl)phosphine hydrochloride (TCEP). Multiple wavelength scans were recorded at 1 nm intervals from 195 to 280 nm on 5.0 μM protein solutions in a 1 mm cuvette at 25°C. Raw CD data were analyzed as previously described (40).
Direct or competitive colorimetric microplate-based assays were conducted with purified proteins and a 3′ biotinylated ssDNA probe corresponding to the purine-rich strand of the murine ACTA2 5′-flanking sequence from −195 to −164 (PE32-bF) as previously described (40, 56). In the direct ssDNA-binding format, solid-phase Purβ-PE32-bF complexes were detected by ELISA using primary rabbit antibodies directed against amino acids 210-229 or 302-324 of mouse Purβ (22) or the NHis tag (His probe H-15 from Santa Cruz Biotechnology, Inc. or anti-6-His from Bethyl Laboratories, Inc.). In the competitive binding format, solid-phase nucleoprotein complexes were detected with ExtrAvidin®-peroxidase (Sigma-Aldrich). Log [protein] vs. absorbance datasets were fit to four parameter variable slope equations to determine EC50 or IC50 values depending on the format of the assay (Prism 5, Version 5.04, Graphpad Software, Inc.).
Protein-protein interaction was assessed in an ELISA format using microtiter wells (Costar® EIA/RIA 96 well plate, certified high binding, Corning Inc.) coated with 200 nM NHis-Purβ or selected truncation proteins as previously described (30, 56). Nuclear extracts prepared from exponentially growing AKR-2B MEFs served as a source of protein binding partners of Purβ (57). The primary antibodies used for detection of solid-phase protein-protein complexes included rabbit anti-mouse Purα 291-313 (22), rabbit anti-mouse YB-1 (MSY1) 242-269 (22), and rabbit anti-mouse TEF1 1-15 (17). Commercial rabbit polyclonal antibodies against SRF (G-20), Sp1 (H-225 and PEP-2), and Sp3 (D-20) were from Santa Cruz Biotechnology, Inc.
AKR-2B MEFs were seeded into 6 well plates at 4.0 ×104 cells per well in McCoys 5A medium with 5% FBS. Primary mouse aortic outgrowth smooth muscle cells (AoSMCs) (30) were seeded at 2.5 ×104 cells per well in DMEM with 10% FBS. After an overnight incubation at 37°C in a 5% (MEFs) or 10% (AoSMCs) CO2 incubator, adherent cells were transfected with 2 μg of total DNA using jetPEI™ reagent (PolyPlus-transfection) at a ratio of 1.5 μl per μg of DNA. Transfection solutions typically contained 0.9 μg of pVSMP8- or pVSMP4-CAT or luciferase reporters, 0.1 μg of pSV40-β-Gal control reporter, and 1.0 μg of expression plasmid. After 48 h incubation at 37°C, cells were washed with PBS and then extracted with 1× Passive Lysis Buffer (Promega) supplemented with protease inhibitors. Total protein content was determined by BCA™ or Bradford assay and individual reporters were measured with the use of a CAT or β-Gal ELISA kit (Roche Applied Science), Luciferase Assay System (Promega), or ortho-nitrophenyl-β-galactoside chromogenic substrate assay. Numerical datasets were subjected to one-way analysis of variance and Tukey’s multiple comparison test with significance set at p < 0.05 (Prism 5, Version 5.04, Graphpad Software, Inc.). In some instances, transfected cells were processed by sequential extraction using a subcellular protein fractionation kit as directed by the manufacturer (Thermo Scientific).
Samples were prepared for SDS-PAGE by dilution of concentrated cell lysates into 6× sample preparation buffer (120 mM Tris-Cl pH 6.8, 3% w/v SDS, 30% v/v glycerol, 0.03% w/v bromophenol blue). For less concentrated cell lysates, soluble protein was precipitated by adding 5 volumes of ice-cold ethanol to 1 volume of cell lysate and incubating for at least 1 h at −20°C. Precipitated protein was collected by centrifugation and dissolved in 1× SDS-PAGE loading buffer. Samples were supplemented with 5% v/v 2-mercaptoethanol, heated for 3-5 min at 100°C, and subjected to slab gel electrophoresis on 10%, 12% or 15% w/v acrylamide:bisacrylamide (29:1) mini-gels or 4-20% precast gradient gels (Lonza). Molecular weight standards were run in parallel on each gel (BenchMark™ Prestained Protein Ladder, Invitrogen). Proteins were then electrotransferred to Immobilon®-P or Immobilon®-PSQ polyvinylidene difluoride membrane (Millipore) in 25 mM Trizma base, 192 mM glycine, 20% v/v methanol at 125 V for 90 min at 4°C. Transblots were probed with selected antibodies as described in (22). Primary and secondary antibodies used are listed in Supporting Information.
To assess the phenotypic consequence of Purβ loss-of-function in a multipotent mesenchymal cell type, a lentiviral shRNA expression system was used to stably-transduce AKR-2B MEFs owing to their high steady-state levels of Purα and Purβ and sensitivity to inducers of myofibroblast differentiation (20, 22, 58). As shown in Figure 1A, specific knockdown of Purβ (faster migrating band of doublet) was confirmed in two independently-derived blasticidin-resistant cell lines by Western blotting with the use of a newly-developed rat monoclonal antibody directed against a conserved PUR repeat I sequence present in both Purα and Purβ (specifically, amino acids 42-69 of Purβ). Importantly, Purα expression (slower migrating band of doublet) was not altered in either Purβ-only knockdown cell line (βI-B4 and β1-G7) compared to control cells transduced with scrambled RNA (βS-E6). For the purpose of comparison, extract from a serendipitously-derived cell line deficient in Purα expression (βI-F3) was included on the gel to highlight differences in the electrophoretic mobility of bands corresponding to Purα and Purβ. To quantify relative differences in Purβ expression, a functional ELISA was used to measure Purβ ssDNA-binding activity in soluble extracts of each cell line (Figure 1B). In keeping with the results of Western blotting, Purβ ssDNA-binding activity was reduced by 2.5-2.7 fold in the βI-B4 and βI-G7 cells in comparison to the βS-E6 control cell line. Consistent with a dominant Purα loss-of-function phenotype (59), the Purα-deficient βI-F3 cell line exhibited a markedly enhanced rate of cell growth relative to Purβ-only knockdown and scrambled control cell lines and was thus excluded from detailed study (Figure S2).
Analysis of growing MEF cell lines by light microscopy revealed that Purβ-only knockdown cells adopt a more elongated spindle-like morphology in comparison to control cells expressing scrambled RNA (Figure 1C). A similar but less dramatic change in morphology is evident in cells co-deficient in Purα. These morphological differences were readily apparent in confluent cell monolayers as well (Figure S3). To investigate the physical basis for these changes in cell shape, the expression of cytoskeletal actin proteins was assessed by immunoblotting of detergent-soluble and detergent-insoluble lysates prepared from serum- or TGF-β1-stimulated MEF cell lines. In serum-stimulated cells, the soluble G-form of SMαA was the predominant actin isoform exhibiting enhanced expression in concert with Purβ knockdown although a slight increase in β-actin expression was detected as well (Figure 2A). Consistent with sensitization to TGF-β1-driven myofibroblast cyto-differentiation, a corresponding increase in both G- and F-form SMαA, but not β-actin, was seen in Purβ knockdown MEFs relative to the control cell line treated with TGF-β1 (Figure 2B). Purβ deficiency, however, did not promote acquisition of a smooth muscle cell-like protein expression pattern as other markers of smooth muscle differentiation including smooth muscle myosin and SM22α/transgelin were not detected by immunoblotting of either control or Purβ knockdown MEFs (data not shown). To determine whether increased expression of SMαA protein correlated with de-repression of ACTA2 transcription, transient transfection assays were conducted using a minimal MCAT- and CArG-dependent ACTA2 enhancer-promoter construct. As shown in Figure 2C, ACTA2-driven reporter expression was significantly increased in both Purβ knockdown cell lines in comparison to control cells implying that loss of Purβ repressor function is necessary and sufficient to promote ACTA2 expression and myofibroblast differentiation in MEFs. In keeping with the known motility-restrictive properties of SMαA-containing myofilaments in cultured fibroblasts (60), Purβ knockdown MEFs also demonstrated a modest reduction in chemotactic migration toward serum growth factors as measured by Boyden chamber assay (Figure S4).
To explore the physical basis for Purβ-mediated repression of ACTA2 transcription, we first set out to computationally model the higher order structure of Purβ based on 1) sequence homology with other members of the purine-rich element binding protein family, and 2) the known x-ray crystal structure of Dm Purα 40-185. Primary sequence analysis of Mm Purβ using selected homology detection and structure prediction algorithms indicated that Purβ possesses the same linear arrangement of PUR repeat modules (designated I, II, III) as originally described in Purα (38, 39) (Figure 3A). The sequences encoding Mm Purα and Mm Purβ were threaded on the x-ray crystal structure of Dm Purα residues 40-185 (3K44) using SWISS-MODEL to assess the overall homology of the intramolecular PUR domain formed by PUR repeats I and II of each protein. As expected, the major differences were restricted to putative loop regions connecting the individual PUR I and II repeats while the β-strand and α-helix forming sequences were virtually superimposable (Figure S5). To create a hypothetical structure of the full-length Purβ monomer, PUR repeat III was independently modeled using SWISS-MODEL and the tertiary structure of the entire protein was constructed on the basis of an I-TASSER generated template (Figure 3B). Like PUR repeats I and II, PUR repeat III is predicted to possess similar ββββα topology. However, a putative random coil-forming region located between PUR repeats II and III may impart some degree of flexibility as to the position of PUR repeat III relative to the intramolecular domain formed by PUR repeats I and II (Figure 3C). Consequently, dimerization may occur via formation of an intermolecular PUR domain composed of two self-associating PUR III repeats from two Purβ monomers (Figure 3D).
To test the models in Figure 3, a series of cDNAs were engineered to encode NHis-tagged Purβ truncation proteins corresponding to individual PUR repeat modules (Purβ I, Purβ II, Purβ III) or selected combinations thereof (Purβ I-II, Purβ II-III, Purβ I-II-III). The utility of recombinant NHis-Purβ as a reliable experimental surrogate for the native protein expressed in mammalian cells has been documented in previous studies (29, 37, 56). Sequence validated bacterial expression plasmids were transformed into E. coli cells and recombinant truncation proteins were produced for trial purification under both native and denaturing conditions. While metal chelate affinity enrichment of each Purβ truncation protein was possible under harsh denaturing conditions (data not shown), only Purβ I-II-III (residues 41-303), Purβ I-II (residues 41-210), and Purβ III (residues 209-303) were amenable to purification in non-denaturing solvents (Figure S6) and were found to exhibit CD spectra consistent with well-folded polypeptides (Figure S7). Hence, it appeared that Purβ I (residues 41-112), Purβ II (residues 125-210), and Purβ II-III (residues 125-303) were intrinsically unstable and/or mis-folded when expressed in E. coli. This observation is consistent with the putative requirement for intramolecular association of PUR repeats I and II or intermolecular association of two PUR III repeats to form a stably-folded subdomain (Figure 3). To confirm the predicted quaternary state of each isolated truncation protein, calibrated SEC was performed using loading concentrations well in excess of the reported Kd for the full-length Purβ dimer (36). As shown in Figure 4A and B, Purβ I-II-III eluted as a ~64 kDa dimer while Purβ I-II resolved as a ~18 kDa monomer. In agreement with its predicted role in mediating self-association, Purβ III eluted as a ~17 kDa dimer (Figure 4C).
To assess the functional role of each putative Purβ subdomain in ACTA2 repression, transient co-transfection assays were performed in MEFs and primary mouse AoSMCs using a full-length ACTA2 construct (−1070 to +2582, VSMP8-Luc) as well as a minimal MCAT/CArG/GC box-dependent enhancer (−146 to +46, VSMP4-Luc) as specific transcriptional targets of Purβ (30). As shown in Figure 5B, full-length Purβ and the core I-II-III construct demonstrated comparable repressor activity toward both ACTA2 reporters in MEFs. Importantly, Purβ I-II was the only other truncation protein to exhibit statistically significant repressor activity. Purβ III alone showed no inhibitory activity while Purβ II-III was only weakly repressive. In agreement with E. coli expression and purification studies, immunoblotting of non-denatured MEF lysates indicated stable expression of full-length Purβ, Purβ I-II-III, Purβ I-II, and Purβ III (Figure 5C). However, Purβ I, Purβ II, and Purβ II-III were either not readily detected or only seen after extracting detergent-insoluble cell remnants with a denaturing solvent (Figure 5D), again pointing to the intrinsic instability and/or mis-folding of these truncation mutants. Essentially identical results were obtained in co-transfection studies conducted with primary AoSMCs (Figure S8) supporting the conclusion that the relative ACTA2 repressor activity of the stably-expressed truncation proteins is Purβ I-II-III > Purβ I-II Purβ II-III or Purβ III. Titration studies conducted in AKR-2B MEFs using a fixed amount of VSMP8 reporter and varying amounts of expression plasmid confirmed that Purβ I-II-III and Purβ I-II are quantitatively distinct in terms of their ACTA2 repressor activity (Figure 6).
To assess whether the relative ACTA2 repressor activity of each Purβ truncation protein correlated with differences in the affinity and/or specificity for target sites in the ACTA2 promoter, colorimetric microplate-based assays were used to compare the ssDNA-binding properties of Purβ I-II-III, Purβ I-II, and Purβ III to full-length Purβ. The purine-rich sense strand of the 5′ ACTA2 Pur/Pyr element containing a consensus core MCAT motif (italics) (GGGAGCAGAACAGAGGAATGCAGTGGAAGAGA, PE32-F) was chosen as a probe because it has multiple interacting binding sites (underlined) that permit formation of a high-affinity (macroscopic Kd ~ 0.3 nM) 2:1 Purβ:ssDNA complex (37). Initially, a titration experiment was performed to identify a minimal concentration of 3′ biotinylated PE32-F probe necessary to detect the interaction of each Purβ truncation protein by ELISA. Surprisingly, stable formation of Purβ I-II nucleoprotein complexes required a markedly higher concentration of ssDNA implying a significant difference in binding affinity of the monomeric subdomain relative to full-length Purβ, Purβ I-II-III, and Purβ III (Figure S9A). Titration assays conducted with a limiting concentration of PE32-bF (0.5 nM) highlighted the striking differences in the apparent ssDNA-binding affinity of the individual subdomains, Purβ III (EC50 = 1.80 ± 0.74 nM, n = 4) and Purβ I-II (EC50> 100 nM), relative to the full-length protein (EC50 = 0.16 ± 0.05 nM, n = 4) and Purβ I-II-III (EC50 = 0.22 ± 0.05 nM, n = 4) (Figure 7A and B). Consistent with a functional distinction between the separated dimerization and intramolecular subdomains, competition assays revealed that Purβ III (IC50 = 46 ± 15 nM, n = 3) and Purβ I-II (IC50 > 1000 nM) were much less effective than the composite Purβ I-II-III construct (IC50 = 1.0 ± 0.3 nM, n = 5) or the full-length protein (IC50 = 1.2 ± 0.2 nM, n = 5) at inhibiting the interaction of 0.5 nM PE32-bF with immobilized Purβ (Figure 7C). Despite these substantial differences in apparent ssDNA-binding affinity, the isolated Purβ III and Purβ I-II subdomains retained similar binding site specificity as demonstrated by their reduced interaction with mutant versions of PE32-bF containing heptathymidylate substitutions (T7) in place of 5′ or 3′ PUR and/or internal MCAT motifs (Figure 7D and Figure S9B).
To ascertain whether the isolated subdomains exhibited similar protein binding properties as full-length Purβ, an ELISA-based profiling assay was conducted using nuclear extract from AKR-2B MEFs as a natural source of potential Purβ interaction partners. Recombinant Purβ proteins were immobilized on microtiter wells at a saturating coating concentration (200 nM) and then assayed for their ability to capture specific transcription factors implicated in ACTA2 activation or repression. Consistent with previous findings (30, 56), full-length Purβ demonstrated preferential interaction with its co-repressor partner MSY1 relative to other factors screened using this assay format (Figure 8). Interestingly, Purβ I-II-III exhibited an even greater binding capacity for MSY1, while the isolated Purβ I-II and Purβ III subdomains displayed markedly reduced interaction with MSY1. In contrast, Purβ I-II showed a clear preference for interaction with the ACTA2 trans-activator Sp3, while Purβ III exhibited little or no Sp3 binding activity. Essentially identical results were obtained using nuclear extract diluted in binding buffer supplemented with reducing agent (Figure S10). This was done to ensure that the differences observed in the binding properties of individual Purβ truncation proteins were not attributable to anomalous protein oxidation.
The transient differentiation of stromal fibroblasts to contractile, SMαA-expressing myofibroblasts is an essential and tightly-regulated component of the wound healing process. Conversely, sustained stromal myofibroblast activation is pathologic as it often promotes aberrant tissue remodeling (61). Because SMαA expression is a biochemical hallmark of the myofibroblast phenotype, a better understanding of the regulatory factors that mediate ACTA2 transcription and translation in fibroblasts may reveal novel targets for therapeutic intervention to limit destructive fibrocontractile remodeling associated with scarring, fibrosis, and tumor progression. Among the factors implicated in ACTA2 regulation in fibroblasts, Purα and Purβ are unique in that they apparently repress gene transcription by forming nucleoprotein complexes with purine-rich ssDNA in such a way as to block trans-activator recognition of cognate double-stranded binding sites within the composite MCAT/CArG/GC box enhancer (16, 17, 30, 62).
In view of a growing body of evidence suggesting that Purβ may play a central role in repressing genes encoding muscle-restricted isoforms of actin and myosin in both myogenic and non-myogenic cell types (31-35), we initially sought to confirm that deficiency of Purβ in MEFs would necessarily promote the acquisition of a myofibroblast phenotype in vitro. To do so, we transduced MEFs using a lentivirus-based shRNA transgene delivery system to knockdown the expression of Purβ in a specific, stable, and constitutive manner. Analyses of two independently-derived cell lines showed that a relatively modest decrease in Purβ expression (~60-70% knockdown) was sufficient to switch cells to a myofibroblast-like phenotype as exemplified by characteristic changes in cell morphology, SMαA expression, TGF-β1 inducibility, and chemotactic migration (Figures 1, ,2,2, S3 and S4). Importantly, these changes occurred in the absence of any substantive effect on cell growth suggesting that Purβ does not participate in the direct regulation of cell cycling as has been reported for Purα (63-68).
To better understand the structural basis for Purβ-mediated repression of ACTA2, we employed web-based homology modeling servers to generate computational models of the Purβ monomer and dimer based on the known x-ray crystal structure of Dm Purα residues 40-185 (39). As previously described for Purα (38), the HHrepID web server identified three regions of internal sequence homology termed PUR repeats I, II, and III (Figure 3). Homology modeling suggested that each PUR repeat is similarly structured with respect to the arrangement of four β-stands and one α-helix. By analogy to the tertiary structure of Dm Purα I-II, Mm Purβ I-II is predicted to fold in such a way as to form an intramolecular PUR domain with features resembling a Whirly class-like DNA-binding fold (39). Although the PUR III repeat of Purβ is also predicted to adopt ββββα topology, the α-helical region is substantially longer than in PUR repeats I and II owing to the presence of the so-called “psycho” motif spanning residues 264-291 (26). This sequence is predicted to form an extended amphipathic α-helix, which may facilitate protein-protein interaction. In one hypothetical model of the Purβ monomer, the PUR III repeat is depicted as packing against the PUR I-II intramolecular domain (Figure 3B). However, previous hydrodynamic studies showed that full-length Purβ reversibly self-associates to form an elongated homodimer (36). Therefore, we speculated that the glycine-rich sequence spanning residues 210-229 may impart some degree of internal flexibility allowing the PUR III repeat region to extend away from I-II (Figure 3C). The interaction of two PUR III repeats to form an intermolecular PUR domain would necessarily give rise to an elongated Purβ homodimer composed of three distinct modules (Figure 3D).
Based on these deduced homology models, we created a set of expression vectors encoding His-tagged Purβ truncation proteins containing single or selected combinations of PUR repeats I, II, and III. Trial purifications from E. coli indicated that stable expression and folding of Purβ requires the formation of intra- and/or intermolecular subdomains. For example, Purβ I, Purβ II, and Purβ II-III constructs were poorly expressed and/or only detectable in denatured lysates. Conversely, full-length Purβ, Purβ I-II-III, Purβ I-II, and Purβ III were each highly expressed and readily purified under non-denaturing conditions. Moreover, their respective CD spectra were indicative of well-folded polypeptides. The results of calibrated SEC analysis confirmed the predicted quaternary state of each purified truncation protein and indicated that PUR repeat III constitutes the dimerization domain of Purβ (Figure 4). These findings are entirely consistent with the reported quaternary structures of Dm Purα I-II and Dm Purα I-II-III (39). Although the Purα homodimer has been proposed to adopt a Z-like shape based on results of small angle-X-ray scattering (39), the exact orientation and relation of the intra- and intermolecular subdomains in the Purβ homodimer is currently unknown.
To evaluate the capacity of each Purβ truncation protein to repress ACTA2 transcription, we performed ACTA2-luciferase reporter gene assays in both MEFs and AoSMCs. As expected, the relative expression/stability of each Purβ construct in mammalian cells was similar to that seen in E. coli. Of the truncation proteins expressed, only the dimerization-competent Purβ I-II-III core construct and the Purβ I-II intramolecular subdomain were consistently found to repress both the complete and minimal ACTA2 enhancer-promoter in both cell types (Figures 5 and S8). However, while Purβ I-II-III repressed the promoter to the same extent as the full-length protein, Purβ I-II exhibited ~50% less repressor activity. The intrinsically weaker activity of Purβ I-II was validated by titration experiments conducted over an extended range of expression vector concentrations pointing to the necessity of the PUR III repeat for full repressor function (Figure 6). Interestingly, forced expression of the Purβ III dimerization domain by itself did not affect the ACTA2 enhancer-promoter owing, in part, to the apparent inability of this construct to enter the nucleus when separated from Purβ I-II intramolecular domain (Figure S11). However, in the context of the full-length protein and the Purβ I-II-III core construct, we surmise that the PUR repeat III likely promotes more efficient ACTA2 repression by mediating the formation of a dimeric repressor capable of multisite ssDNA-binding within the confines of the nucleus.
Previous high resolution structural analyses of the 5′-flanking region of ACTA2 during myofibroblast differentiation revealed that an asymmetric Pur/Pyr tract spanning nucleotides −210 to −150 is hypersensitive to modification by chemical probes that preferentially react with unpaired nucleobases (58). This region contains a consensus MCAT motif and a TGF-β1 response element that appear to function in conjunction with downstream CArG and GC boxes to mediate high level ACTA2 transcription in fibroblasts (16, 17, 20). Consequently, we have chosen to focus our efforts on characterizing the interaction of Purβ with the MCAT region of the ACTA2 enhancer-promoter due to its high Pur/Pyr asymmetry and apparent propensity to transiently adopt non B-form structures in vivo (58). In this regard, a prior study from our lab reported that Purβ interacts with the 32 nt purine-rich strand of the ACTA2 MCAT element (dubbed PE32-F, −195 to −164) via a cooperative binding mechanism to generate a high-affinity 2:1 Purβ:ssDNA complex (37).
To test the importance of protein dimerization in facilitating the interaction of Purβ with PE32-F, the relative ssDNA-binding affinity and specificity of full-length Purβ and the core I-II-III protein were compared to the isolated intra- and intermolecular subdomains using both direct and competitive ssDNA-binding assays. Because these microplate-based colorimetric assays were performed under intrinsically non-equilibrium conditions, it was not possible to determine precise quantitative differences between Purβ I-II-III and the full-length protein. Despite this technical limitation, the Purβ I-II-III core construct did appear to bind PE32-F with comparable affinity and specificity to full-length Purβ under the assay conditions employed (Figure 7). Moreover, the biochemical properties of Purβ I-II-III (residues 41-303) defined in this study are analogous to those of a His tag-free core tryptic fragment of Purβ (residues 29-305) described in an earlier report (40). Interestingly, while the Purβ I-II and Purβ III subdomains each displayed a lower apparent affinity for PE32-F than Purβ I-II-III, the intermolecular subdomain bound more tightly to ssDNA than the intramolecular subdomain. The functional non-identity of the isolated Purβ I-II and Purβ III subdomains suggests that the native Purβ homodimer contains three separate but unequal ssDNA-binding modules. This structural arrangement reinforces the concept that stable nucleoprotein complex assembly on the ACTA2 MCAT element likely involves the recognition of multiple binding sites by Purβ (37). In support of this assertion, mutation of all three PUR elements in PE32-F was necessary to completely eliminate ssDNA-binding by full-length Purβ, the I-II-III core, and each subdomain (Figure 7D).
Another defining attribute of Purβ structure and function uncovered in this report is that the individual subdomains of Purβ differ in their capacity to interact with certain transcription factors relevant to ACTA2 regulation in fibroblasts. In particular, the Purβ I-II intramolecular subdomain was a more avid binder of Sp3 than the Purβ III intermolecular subdomain (Figure 8 and Figure S10). As Sp1 and Sp3 are known to interact with several sequence elements located within the composite MCAT/CArG/GC box enhancer (20), it is quite possible that the relatively strong repressor activity of Purβ I-II observed in transfected cells was due, in part, to its ability to bind and sequester Sp3 away from the enhancer. This would also explain why Purβ I-II retained repressor activity in the face of its relatively weak ssDNA-binding affinity compared to Purβ I-II-III. On the other hand, all three PUR repeats were required for efficient interaction of Purβ with MSY1 (mouse YB-1), the co-repressor protein that interacts with the pyrimidine-rich antisense strand of the ACTA2 Pur/Pyr tract (17, 22). This feature may account for why Purβ I-II-III was such an effective repressor when expressed in cells as direct physical interaction between Purβ and MSY1 is probably essential for efficient assembly of these co-repressors on the Pur/Pyr element and ensuing disruption of the core MCAT motif (37). Coordinated binding of MSY1 to the pyrimidine-rich strand may also serve to potentiate the intrinsic helix-destabilizing activity of Purβ (28). Unraveling the degree to which Purβ can stably alter the secondary structure of specific cis-elements in ACTA2 will clearly require a more systematic evaluation of the helix-destabilizing properties of Purβ and its isolated subdomains based on the biochemical criteria established for Purα-mediated melting of duplex DNA (27, 28, 69).
In summary, our findings reveal that Purβ is a potent inhibitor of myofibroblast differentiation by virtue of its ability to repress ACTA2 transcription via specific protein-ssDNA and protein-protein interactions. The functionally relevant unit of Purβ that mediates ACTA2 repression appears to be the homodimeric form of the protein. Subdomain-specific inter- and intramolecular interactions account for the formation of three separate ssDNA-binding modules within the Purβ homodimer. The tripartite organization of the assembled homodimer readily explains the structural basis for the cooperative binding of Purβ to multiple purine-rich sites within the MCAT region of the composite ACTA2 enhancer as well as the preferential association of Purβ with its co-repressor partner MSY1/YB-1.
†This work was supported by NIH grant R01 HL054281 and AHA grant 09GRNT2170060 to R.J.K. A.E.R. and L.A.F. were supported by NIH training grant T32 HL007594 (Kenneth G. Mann, PI).
SUPPORTING INFORMATION AVAILABLE