Aromatic Residues Dictate the Transcriptional Repressor and Single-Stranded DNA Binding Activities of Purine-Rich Element Binding Protein B

Abstract

Purine-rich element binding protein B (Purβ) is a single-stranded DNA (ssDNA) and RNA binding protein that functions as a transcriptional repressor of genes encoding certain muscle-restricted contractile proteins in the setting of cellular stress or tissue injury. A prior report from our laboratory implicated specific basic amino acid residues in the physical and functional interaction of Purβ with the smooth muscle-alpha actin gene (Acta2) promoter. Independent structural analysis of fruit fly Purα uncovered a role for several aromatic residues in the binding of this related protein to ssDNA. Herein, we examine the functional importance of a comparable set of hydrophobic residues that are positionally conserved in the repeat I (Y59), II (F155), and III (F256) domains of murine Purβ. Site-directed Y/F to alanine substitutions were engineered and the resultant Purβ point mutants were tested in various biochemical and cell-based assays. None of the mutations affected the cellular expression, structural stability, or dimerization capacity of Purβ. However, the Y59A and F155A mutants demonstrated weaker Acta2 repressor activity in transfected fibroblasts and reduced binding affinity for the purine-rich strand of an Acta2 cis-regulatory element in vitro. Mutation of Y59 and F155 also altered the multisite binding properties of Purβ for ssDNA and diminished the interaction of Purβ with Y-box binding protein 1, a co-repressor of Acta2. Collectively, these findings suggest that some of the same aromatic residues, which govern the specific and high affinity binding of Purβ to ssDNA, also mediate certain heterotypic protein interactions underlying the Acta2 repressor function of Purβ.

Graphical Abstract

A computational model of the Purβ homodimer is depicted in gray with specific aromatic residues highlighted in different colors, namely, Y59 (purple), F155 (red), and F256 (green). Residues 210–229 (cyan) and 302–324 (orange) correspond to the sites recognized by specific Purβ antibodies used in this study.

INTRODUCTION

The contractile proteins actin and myosin serve many essential cellular functions from transport of cellular cargo to facilitating muscle contraction and cell migration.¹ Expression of tissue- and cell-type specific isoforms of these proteins is tightly regulated and is key in determining cellular phenotype.^2–4 For example, smooth muscle α-actin is restricted to smooth muscle cells and myofibroblasts and enables the unique ability of smooth muscle to modulate vascular tone⁵, and for myofibroblasts to contract and to mend damaged tissue upon injury.⁶ In contrast, dysregulation of contractile protein expression can result in an altered cell phenotype leading to aberrant tissue remodeling. For instance, synthetic smooth muscle cells contribute to atherosclerotic plaque formation⁷, and persistent activation of myofibroblasts promotes fibrosis after myocardial infarction.⁸

Purine-rich element binding protein B (Purβ/PURB) is a structurally and functionally unique DNA and RNA binding factor, which acts as a negative transcriptional regulator of several genes encoding contractile proteins. Purβ represses the gene encoding smooth muscle α-actin (Acta2) in growth-activated fibroblasts and smooth muscle cells.^9–11 Purβ also inhibits the expression of certain myosin heavy chain genes (Myh6 and Myh7) in skeletal myoblasts and cardiac myocytes.^12–14 Furthermore, overproduction of Purβ generated by disrupting certain microRNA-dependent pathways or by imposing other tissue stress factors has been implicated in the pathological remodeling of cardiac and skeletal muscle.^15–18 Purβ has also been reported to be a circular RNA binding target in the regulation muscle-specific gene expression.¹⁹

Purβ is a member of a small family of nucleic acid-binding proteins, which are so named because of their preference for binding to purine-rich single-stranded DNA (ssDNA) or RNA molecules.²⁰ Members of the purine-rich element binding protein (PUR) family exhibit diverse functional activities such transcriptional activation/repression, translational regulation, and mRNA transport.²¹ Purα/PURA is the founding and most widely studied member of the family owing to its critical biological role in brain development^22–24 and because haploinsufficiency of PURA in humans causes a severe neurodevelopmental disorder known as PURA syndrome.^25–27 Research on Purβ has focused largely on its role in the transcriptional repression of genes encoding smooth or striated muscle isoforms of actin and myosin although more recent studies suggest that Purβ is capable of regulating other genes in non-muscle cell types.^28–31

The mechanism by which Purβ represses muscle gene expression has been extensively investigated using the Acta2 gene as a model system. Purβ apparently represses Acta2 expression by binding with high affinity and specificity to a single-stranded purine-rich sequence element in the promoter-enhancer region of Acta2.³² Purβ functions in collaboration with Y-box binding protein 1 (Ybx1), a co-repressor protein that binds specifically to the complementary pyrimidine-rich strand.³³ The co-interaction of Purβ and Ybx1 with their opposing ssDNA targets in the Acta2 promoter presumably disrupts the recognition of a core, duplex muscle-CAT (MCAT) motif by transcription enhancer factor 1 (TEF1).⁹ The structural basis underlying the ability of Purβ to bind to ssDNA in a sequence-specific fashion and to Ybx1 to form a multiprotein repressor complex remains largely unknown.

Insights into the role of protein structure on the function of PUR family members have been revealed by x-ray diffraction analysis of protein crystals composed of various Drosophila melanogaster (Dm) Purα subdomains.^{34, 35} Information provided from the high resolution structural studies of Dm Purα afforded the opportunity to computationally model the tertiary and quaternary structure of full-length mammalian Purβ.^{36, 37} Members of the PUR family are structurally related owing to the presence of three repetitive sequence modules (dubbed PUR repeats I, II, and III), each of which adopt a highly conserved secondary architecture consisting of four beta strands and a single alpha helix. Intramolecular association of PUR repeats I and II forms a stably folded PC4-like domain,^{38, 39} which is capable of binding to ssDNA.³⁴ Intermolecular association of two PUR III repeats generates an additional PC4-like domain that mediates the self-association of Purα or Purβ monomers to form stable homodimers.^{34, 36} An intriguing functional distinction between PUR proteins from certain species resides in the relative affinity of the intra- and intermolecular domains for ssDNA and RNA. A study of Dm Purα revealed that the repeat III dimerization domain binds to ssDNA and RNA with significantly lower affinity than the repeat I-II intramolecular domain.³⁵ Conversely, the dimerization domain of Mus musculus (Mm) Purβ binds to purine-rich ssDNA with relatively high affinity and is essential for elaboration of robust Acta2 repressor activity in the full-length protein.³⁶

The identity of specific amino acid residues that mediate the binding of Dm Purα to ssDNA was uncovered by the solved x-ray crystal structure of repeats I-II in complex with GCGGCGG ssDNA.³⁵ Interestingly, aromatic residues Y57 in repeat I and F145 in repeat II form pi-pi stacking interactions with guanine nucleobases in the DNA. Specifically, F145 replaces a neighboring cytosine at a kink in the ssDNA, and in a second spatially distinct DNA binding site, Y57 stacks with the 3’ end of the 7 mer. Based on our published computational models of mouse Purβ quaternary structure,^{36, 37} residues Y59 and F155 of Purβ are positionally conserved with Dm Purα Y57 and F145. Additionally, since previous results from our group indicated that the isolated PUR repeat III domain of Purβ is capable of binding to ssDNA, we wanted to explore if an analogous aromatic residue in repeat III contributes to the interaction of full-length Purβ with ssDNA. Since F256 in repeat III is the positional equivalent to F155 in repeat II, we hypothesized that residues Y59, F155, and F256 may dictate the transcriptional repressor and ssDNA-binding functions of Purβ. To test this idea, we generated the corresponding tyrosine or phenylalanine to alanine point mutants (Y59A, F155A, and F256A) and evaluated the effect of each substitution on the structure and function of Purβ using a combination of cellular and biochemical assays. Our findings point to a critical role for these hydrophobic residues in mediating both protein-DNA and protein-protein interactions that determine the Acta2 repressor activity of Purβ.

MATERIALS AND METHODS

Materials

The QuikChange II XL Site-Directed Mutagenesis Kit was purchased from Agilent Technologies, Inc., Santa Clara, CA. PageRuler Prestained Protein Ladder, Pierce^™ Enhanced Chemiluminescence Reagent, Pierce^™ BCA Protein Assay Kit, Pierce^™ Coomassie Plus Protein Assay Reagent, Invitrogen^™ SYPRO^™ Orange dye, and 6×His tag antibody HIS.H8 were obtained from Thermo Fisher Scientific, Grand Isle, NY. StreptaWell strips and poly [dI-dC] were acquired from Roche Diagnostics, Indianapolis, IN. Mammalian cell culture media and supplements were procured from Mediatech, Inc., Manassas, VA. Heat-inactivated fetal bovine serum (FBS), Dulbecco′s Phosphate Buffered Saline (D-PBS), ExtrAvidin^®-peroxidase, and synthetic oligonucleotides were obtained from Sigma-Aldrich, St. Louis, MO. Polyplus jetPRIME^® transfection reagent was purchased from VWR International, Radnor, PA. Luciferase Assay System and Reporter Lysis 5× Buffer were acquired from Promega Corp., Madison, WI. Immobilon^®-P polyvinylidene difluoride (PVDF) transfer membrane, 2, 2’-AZINO-bis [3-ethylbenziazoline-6-sulfonic acid] (ABTS) peroxidase substrate solution, and mouse anti-glyceraldehyde-3-phosphate dehydrogenase (GAPDH) monoclonal antibody clone 6C5 were obtained from EMD Millipore, Billerica, MA. BluBlot^™ HS autoradiography film was acquired from Krackeler Scientific, Inc., Albany, NY. Horseradish peroxidase (HRP)-conjugates of goat anti-rabbit IgG, goat anti-mouse IgG, and mouse IgG kappa binding protein were purchased from Santa Cruz Biotechnology, Inc., Dallas, TX.

Site-Directed Mutagenesis and DNA Cloning

Point mutations Y59A, F155A, and F256A were introduced into the mammalian expression plasmid pCI containing the mouse Purβ coding sequence with an N-terminal 6×-histidine (N-His) tag^{9, 33} using a QuikChange II XL Site-Directed Mutagenesis Kit according to the manufacturer’s protocol and as previously described.^{37, 40} Mutant primers were designed using the QuikChange Primer Design Tool (Agilent, Santa Clara, CA). Primer sequences are listed in Table S1. The complete open reading frame (ORF) of each point mutant was validated by Sanger sequencing performed by the Vermont Integrated Genomics Resource using a combination of ORF- and pCI-specific primers. Sequence alignments were made using Benchling. Plasmids were then propagated and purified from 1 L cultures of XL10-Gold E. coli cells by double cesium chloride gradient centrifugation as described previously for use in mammalian cell transfection experiments.³⁶ Selected mutant ORFs were subcloned from the pCI vector into a pQE30 bacterial expression vector as previously detailed.³⁷ The pQE30 constructs were transformed into E. coli strain JM109 and sequence-validated prior to use in protein expression.

Protein Purification

Full-length NHis-Purβ WT, Y59A, F155A, or F256A proteins were expressed from pQE30 plasmids in E. coli JM109 cells and purified by successive steps of metal chelate affinity, heparin ion exchange, and size exclusion chromatography (SEC) as described previously.^{37, 40, 41} Column fractions were monitored for protein content by absorbance measurement at 280 nm. Protein purity was assessed by sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) in gels stained with Coomassie Brilliant Blue R-250. The SEC column was calibrated with molecular weight standards to assess the ability of the mutants to form stable homodimers. The concentrations of final mutant protein preparations were ascertained by Bradford assay (Coomassie Plus Protein Assay Reagent) using NHis-Purβ WT as a standard. The concentration of NHis-Purβ WT was determined by absorbance measurement at 280 nm and a molar extinction coefficient as defined previously.^{36, 41} Stocks of purified proteins were stored at 4°C in SEC buffer consisting of 50 mM sodium phosphate, pH 8.0, 300 mM NaCl, 10 mM imidazole, and 10 mM β-mercaptoethanol. Structural models of specific proteins were depicted using PyMOL with the adaptive Poisson-Boltzmann Solver plugin for electrostatic representations.^{42, 43}

Cell Culture and Transient Transfection of Mouse Fibroblasts

AKR-2B mouse embryonic fibroblasts (MEFs) were cultured in growth medium consisting of the Iwakata and Grace modification of McCoy’s 5A medium supplemented with 2 mM L-glutamine and 5% v/v heat-inactivated FBS. Cells were maintained in a humidified 5% CO₂ incubator at 37°C and used for transient transfection experiments at passage 5 to 15. Briefly, AKR-2B MEFs were seeded at a density of 4 × 10⁴ cells per well in 6-well plates and cultured in 2 mL growth medium for 24–48 hours prior to transfection. Once the cells reached 40–50% confluence, the cells were transfected with a combination of plasmids consisting of 0.5 μg pCI NHis-Purβ wild-type (WT) or mutant expression plasmid, 0.9 μg Acta2 promoter-luciferase reporter plasmid (VSMP8)³⁶ and 0.1 μg of pSV40-βgal reporter plasmid. In some instances, a titration experiment was performed in which the amount of WT or mutant NHis-Purβ expression plasmid was varied while keeping the total amount of pCI plasmid fixed at 0.5 μg using empty pCI vector. The three plasmid mixture (1.5 μg total DNA) was pre-incubated with jetPRIME^® transfection reagent at a ratio of 1:2 DNA (μg):jetPRIME^® (μL). The mixture was adjusted to 150 μL with JetPRIME^® buffer and added dropwise to the medium in each well. Transfections were performed in triplicate or quadruplicate. Cells were incubated for four hours after which the medium was changed and the cells were returned to the incubator for another 44 hours. Adherent cells were prepared for harvesting by aspirating the growth medium and then washing the cells twice with ice cold D-PBS. Cells were lysed by addition of 250 μL of 1× Reporter Lysis 5× Buffer supplemented with a mixture of protease inhibitors (0.5 mM phenylmethylsulfonyl fluoride and 1. 0 μg/mL each of leupeptin, pepstatin A, and aprotinin). Wells were gently scraped and cell lysates were transferred to microcentrifuge tubes and stored at −20°C for use in biochemical assays.

Luciferase Reporter and Protein Assays

Acta2 promoter activity in NHis-Purβ wild-type (WT) or mutant-transfected cells was evaluated by luciferase reporter assay. Transfected cell lysates were thawed and subjected to centrifugation for 10 minutes at 15,300 × g to pellet the insoluble cell debris. The soluble fraction in the supernatant was assayed for total protein content and luciferase enzyme activity. Protein content was determined by bicinchoninic acid (BCA) assay using the Pierce^™ BCA Protein Assay Kit with bovine serum albumin (BSA) as a standard. Luciferase activity was measured using a Luciferase Reporter Assay System according to the manufacturer’s protocol. Luminescence was detected using a TD-20/20 luminometer (Turner Designs, Sunnyvale, CA). Units of luciferase activity were adjusted for protein content. Results are reported as fold change in repressor activity compared to empty pCI vector.

Western Blotting

Immunoblotting was used to evaluate the expression of NHis-Purβ WT or mutant proteins in transfected cell extracts. The preparation, electrophoresis, and transfer of cellular proteins from SDS-PAGE mini-gels to PVDF blotting membrane were performed as described in previous accounts.^{37, 44} Blots were gently rocked overnight with blocking buffer composed of 5% w/v nonfat dry milk in Tris-buffered saline (20 mM Tris-Cl pH 7.5, 150 mM sodium chloride) with 0.1% v/v Tween-20 (TBST) at 4°C. Blots were then incubated with 5 mL of primary antibody diluted in blocking buffer for two hours at room temperature. Blots were washed three times with TBST and incubated with 5 mL HRP-conjugated secondary antibody diluted in blocking buffer for one hour at room temperature. Blots were then washed four times with TBST, incubated in freshly prepared Pierce^™ ECL Western blotting substrate for one minute, and visualized on a LAS-4000 imaging system (Fujifilm Corp.) or on autoradiography film developed in a Protec Optimax x-ray film processor. Densitometry analysis was performed using Multi Gauge V3.1 software (Fujifilm Corp.) or ImageJ software by taking the sum of the pixel intensity for each entire band (not line scan). The primary and secondary antibodies used for Western blotting and their working concentrations are summarized in Table S2.

Southwestern Blotting

Southwestern blotting was used to qualitatively assess the ssDNA-binding capacity of recombinant NHis-Purβ proteins purified from E. coli. Proteins were resolved by SDS-PAGE and transferred to PVDF membrane in an analogous fashion to the Western blotting method described above. Blots were gently rocked overnight in blocking buffer composed of 2% w/v BSA in TBST. Blots were then incubated with 5 mL of 10 nM biotinylated oligonucleotide corresponding to the Purβ recognition sequence in the Acta2 promoter (PE32-bF, Table S3) diluted in 0.2% w/v BSA, TBST with 2 μg/mL poly[dI-dC] for two hours at room temperature. Blots were washed three times with TBST and then incubated with a 1:2000 dilution of ExtrAvidin^®-peroxidase for one hour at room temperature. The blots were washed with TBST, incubated with ECL reagent, and imaged as described for Western blotting. Densitometry analysis was performed using Multi Gauge V3.1 software (Fujifilm Corp.) or ImageJ software.

Thermal Shift Assay

The thermostability of purified proteins was assessed by differential scanning fluorimetry as described previously ^{37, 40} with a few technical adjustments. Purified protein stocks of WT or mutant Purβ were diluted to 160 μg/mL, 98 μg/mL, 75 μg/mL, or 50 μg/mL (~4.6 to 1.4 μM) in SEC buffer with 5× SYPRO^™ Orange dye. Samples were assayed in triplicate in 96-well PCR reaction plates at 40 μL/well using a QuantStudio 3 RT-PCR System and QuantStudio Design & Analysis Software v1.4 (Applied Biosystems) with an x4-m4 filter set (excitation 580 nm, emission 623 nm). Melt curves were generated using a continuous temperature ramp from 25°C to 95°C with run mode set to fast. This generated 8–9 readings per one degree Celsius increase in temperature with a total run time of approximately 25 minutes. Fluorescence intensities (F) were averaged to yield a single value per one degree Celsius increment. Background signal was then subtracted using corresponding F values obtained from negative control wells containing 5× SYPRO and SEC buffer but no protein. Background corrected F values generated in each well were normalized to the Fmax and Fmin of the respective well using the equation (F-Fmin)/(Fmax-Fmin). Normalized values between the Fmin and Fmax for each well were then plotted and fit to the Boltzmann equation to determine a melting/unfolding temperature (T_m) for each respective protein analyzed.

Protein-DNA and Protein-Protein Interaction Assays

A series of 96-well microplate assays were used to quantitatively compare the binding of purified Purβ WT and mutant proteins to either ssDNA, Ybx1, or selected Purβ antibodies. Briefly, the interaction of Purβ with various biotinylated ssDNA probes (Table S3) immobilized on streptavidin-coated wells was assessed by enzyme-linked immunosorbent assay (ELISA).^{37, 40} The interaction of Purβ with purified mouse Ybx1⁴⁵ immobilized on microplate wells was similarly evaluated by ELISA.^{37, 40} The relative reactivity of several different Purβ antibodies for WT or mutated Purβ adsorbed to microplate wells was also assessed by ELISA.^{37, 40} The primary rabbit polyclonal antibodies employed in the various immunoassays recognize distinct epitopes spanning residues 210–229 or 302–324 of mouse Purβ.³³ A competition assay was conducted as a secondary appraisal of Purβ binding to ssDNA in the absence of antibody-based detection of nucleoprotein complexes.^{45, 46} More detailed descriptions of each of these binding assays are provided in the Supporting Information.

Data Analysis and Statistics

All graphical and statistical analyses were conducted using GraphPad Prism 9 (GraphPad Software, San Diego, CA). Numerical data generated from cell-based transfection assays are presented as the mean and standard error of the mean (SEM). Numerical data generated from biochemical assays using purified proteins are reported as the mean and standard deviation (SD). Normalized luciferase activity data obtained from two to three separate transfection experiments were subjected to one-way or two-way analysis of variance (ANOVA) where applicable. Dunnett’s multiple comparisons test was applied using Purβ WT as the control in datasets analyzed by one-way ANOVA. Normalized fluorescence intensity data from the thermal shift assay were plotted and fit to a Boltzmann equation. Midpoint T_m values were determined for each protein tested at multiple concentrations in three different experiments. Differences among group means were assessed by one-way ANOVA along with Dunnett’s multiple comparisons test using Purβ WT as the control. Absorbance data from the protein-ssDNA interaction ELISA were plotted and fit to a four-parameter logistic curve (log(agonist) vs. response – variable slope). EC50 and B_max values were determined for each protein in three different titration experiments. Differences among group means were evaluated by one-way ANOVA along with Dunnett’s multiple comparisons test using Purβ WT as the control. Absorbance data from the competition assay were plotted and fit to a four-parameter inhibitory dose-response curve (log(inhibitor) vs. response – variable slope). IC50 values were determined for each protein in four different titration experiments. Absorbance data from the protein-protein interaction ELISA were plotted and fit to a hyperbola or spline curve. Apparent K_d values were determined for Purβ proteins that generated Ybx1 binding isotherms conforming to a hyperbolic shape. For the statistical tests described above, a p value less than 0.05 was considered to be indicative of a significant difference between the control and experimental groups.

RESULTS

Positional Conservation of Putative Base Stacking Residues in fruit fly Purα and mouse Purβ

The positional conservation of Dm Purα residues Y57 and F145 in mouse Purβ was assessed by comparing the primary sequence and structure of each protein. The empirically determined x-ray crystal structures of the Dm Purα PUR repeat I-II (5FGP) and PUR repeat III (5FGO) domains were accessed from the Protein Data Bank.^{34, 35} The structure of mouse Purβ was computationally deduced by homology modeling^{36, 37} and independently validated with AlphaFold.^{47, 48} A common feature of both proteins is that intramolecular association of PUR repeats I and II gives rise to a type I domain that mediates protein binding to ssDNA, while intermolecular interaction of two PUR repeat III modules from two monomers produces a type II domain that mediates protein dimerization (Figure 1).³⁹ Dm Purα residues Y57 and F145 have been implicated in the binding to ssDNA via stacking interactions between the amino acid side chain and guanine nucleobases.³⁵ Sequence and structural alignment of Purα and Purβ confirmed that these particular nucleobase-contacting residues correspond to Y59 in repeat I and F155 in repeat II of mouse Purβ (Figure 1 and Figure S1). Residue F256 in repeat III of mouse Purβ is the positional homolog of F155 in repeat II of mouse Purβ and Y218 in repeat III of Dm Purα. While Y218 was not found to influence Purα binding to ssDNA, presumably because the repeat III domain of Dm Purα only binds weakly to ssDNA,³⁵ we elected to study residue F256 in Purβ since the repeat III domain of Purβ binds strongly to ssDNA³⁶ perhaps via stacking interactions. Consequently, we chose to replace Y59, F155, and F256 with alanine to assess the importance of each of these residues in mediating the functional activities of Purβ.

Figure 1.

Positional conservation of aromatic residues located in predicted DNA binding sites of Mm Purβ compared to Dm Purα. A, B, C) Purβ has three PUR repeat modules designated I (turquoise), II (light blue) and III (dark blue). Each repeat contains an aromatic residue targeted for mutagenesis in this study: Y59 (purple), F155 (red) and F256 (green). A) Schematic of the primary structure of Mm Purβ. Predicted secondary structural elements are shown below each repeat: arrow = beta strand, round-capped bar = alpha helix. B) Cartoon representations of PUR repeats I (residues A47-Q108), II (residues A130-Y205), and III (residues E230-K301) of Mm Purβ³⁷ (homology model) superimposed with PUR repeats I (yellow, residues A45-S106), II (gray, K120-F182), and III (pink, D192-K255) of Dm Purα (PDB ID: 5FGO and 5FGP). Repeats I and II associate to form an intramolecular ssDNA-binding domain (left and middle) while repeat III from two monomers interact to form an intermolecular dimerization domain (right). Each adjoining repeat is shown in transparent mode. Purβ residues Y59 (Dm Purα Y57), F155 (Dm Purα F145) and F256 (Dm Purα Y218) are displayed as sticks to illustrate their relative positions in each respective repeat. Line representation of 7-mer ssDNA molecule (GGCGGCG) base stacking with Dm Purα residues Y57 (left) or F145 (middle) (PDB ID: 5FGP) is shown to highlight the feasibility of analogous stacking interactions between Mm Purβ residues Y59 or F155 and ssDNA nucleobases. C) Computational homology model of the Purβ homodimer³⁷ viewed from two orientations. The intramolecular or type I domain³⁹ formed by repeats I and II in each monomer are located at the top and bottom. The intermolecular dimerization or type II domain³⁹ is located in the middle. The type I domain at the top is shown with two ssDNA molecules (yellow) associated with putative binding sites. Aromatic residues reside on the surface of beta sheets within predicted ssDNA-binding regions. C and D) Yellow ssDNA molecules from Dm Purα 5FGP are overlayed according to the alignments shown in (B), to illustrate putative sites of ssDNA interaction in Mm Purβ repeats I and II. D) Electrostatic representations of the intramolecular domain (repeats I-II, top) and intermolecular dimerization domain (two repeat III molecules, bottom) oriented as in (C). Red = negative/acidic, blue = positive/basic, white = neutral net surface charge. E) Sequence of the forward/sense strand of a 32 nt Acta2 promoter-enhancer element (PE32-F), which is a binding target of Purβ. Three purine-rich elements highlighted in yellow were each mutated to heptathymidylate to study the sequence specificity of Purβ. An inverted muscle-CAT (MCAT) motif is present near the middle of the PE32 sequence (underlined).

Acta2 Repressor Activity of Purβ Y/F Mutants

The transcriptional activity of Purβ Y59A, F155A, and F256A was assessed by transient co-transfection assay in which each mutant was expressed in mouse fibroblasts to determine its ability to repress a transgene consisting of the Acta2 promoter linked to luciferase reporter gene. The Acta2 promoter-luciferase reporter construct known as VSMP8 has been previously shown to provide a reliable readout of Purβ repressor function in both fibroblasts and smooth muscle cells.^{36, 37, 40, 44} Cells transfected with the pCI-Purβ WT construct and the VSMP8 reporter served as a positive control for full repressor activity. Cells transfected with the empty pCI vector and VSMP8 served as a baseline control for Acta2 promoter activity in the AKR-2B MEF cell line. When compared to fibroblasts expressing Purβ WT, cells expressing Purβ Y59A or F155A produced more luciferase indicative of a reduction in Acta2 repressor activity (Figure 2A). Purβ Y59A was the most impaired, exhibiting 50% less Acta2 repressor activity than Purβ WT. Purβ F155A also showed reduced Acta2 repressor activity, but less so than Y59A, with about 40% less activity than Purβ WT. The F256A mutant exhibited a slight diminution in repressor activity but the magnitude of the effect was not significantly different compared to Purβ WT. Western blot analysis of soluble lysates indicated comparable expression of NHis-Purβ WT, Y59A, F155A, and F256A in transfected cells, confirming that the differences detected in luciferase levels were not due to lower expression of the mutant proteins (Figure 2B).

Figure 2.

The Y59A and F155A mutations diminish the Acta2 repressor activity of Purβ expressed in fibroblasts. A) AKR-2B MEFs were co-transfected with an Acta2 promoter-luciferase reporter construct (VSMP8) and a fixed amount of pCI expression plasmid encoding NHis-Purβ WT, Y59A, F155A, or F256A. Cell lysates were assayed for luciferase reporter activity and total protein content. Fold repression reflects the normalized luciferase activity measured in Purβ-transfected cells relative to empty pCI vector-transfected cells (defined as 1). Data are from three independent experiments each with three or four transfections performed per treatment group. Bars show the mean ± SEM (n = 10). p < 0.01 (**), p < 0.001 (***), p < 0.0001 (****) compared to WT based on one-way ANOVA and Dunnett’s multiple comparisons test. B) Cell lysates from (A) were analyzed by Western blotting with a His tag antibody to detect the comparable expression of NHis-Purβ WT and the point mutants. C) AKR-2B MEFs were co-transfected with an Acta2 promoter-luciferase reporter construct (VSMP8) and varying amounts of pCI expression plasmid encoding NHis-Purβ WT or F155A. Data are from two independent experiments each with three transfections performed per treatment group. Symbols with error bars show the mean ± SEM (n = 6). p < 0.01 (**) compared to WT based on two-way ANOVA. D) Cell lysates from (C) were analyzed by Western blotting with a His tag antibody to detect the dose-dependent expression of NHis-Purβ WT and F155A. B, D) A total of 15 μg lysed cell protein was loaded per lane. Each lane contains protein pooled from three replicate samples in a single experiment. The His tag blots were reprobed with a GAPDH antibody as a loading control. Numbers on the left designate the size of marker proteins in kDa.

Because Purβ F155A showed a relatively modest reduction in Acta2 repressor activity at a fixed concentration of expression vector, we wanted to determine if this difference would be observable over a range of expressed protein concentrations. Titration assays were performed using varying amounts of expression plasmid and a fixed amount of reporter plasmid in mouse fibroblasts. Once again, the repressor activity of Purβ F155A was attenuated across the entire titration series compared to Purβ WT (Figure 2C). Western blot analysis indicated that Purβ WT and F155A were expressed in a comparable, dose-dependent manner based on the amount of expression plasmid transfected, confirming that the reduction in Acta2 repressor activity was not due to lower expression of Purβ F155A (Figure 2D).

Purification and Structural Characterization of Purβ Y/F Mutants

Purβ Y59A, F155A, and F256A were selected for further biochemical characterization using purified proteins. Recombinant NHis-tagged mutants were expressed in E. coli and then purified using sequential chromatographic steps involving metal chelate affinity, ion exchange, and SEC (Figure S2). There was no detectable difference in overall protein purity between heparin-Sepharose and SEC steps. Calibrated SEC was performed as an additional quality control measure to verify that the elution profile of each mutant was comparable to the WT protein. Purβ is known to self-associate to form a ~70 kDa homodimer at high (micromolar) protein concentrations.⁴¹ Calibrated SEC analysis of each point mutant indicated that all three mutants were capable of forming homodimers as the major peak of each mutant eluted as expected just before the BSA standard (66 kDa) between 96–100 mL (Figure S3). The paucity of earlier eluting, high molecular weight material in the void volume was consistent with the absence of misfolded aggregates. The lack of a later eluting, lower molecular weight material implied that the mutations did not affect the monomer-dimer equilibrium. Consequently, the results of SEC suggested that the purified Purβ Y59A, F155A, and F256A mutants are properly folded and maintain the ability to self-associate to form homodimers.

The dominant homodimeric peaks obtained by SEC were collected and subjected to further analysis to assess the structural stability of each mutant. A thermal shift assay was used to monitor the unfolding of Purβ WT and each mutant protein as a function of increasing temperature (Figure 3A). The empirically determined melting temperatures (T_m) of Purβ Y59A, F155A, and F256A were essentially identical to Purβ WT over a range of protein concentrations (Figure 3B). These data indicate that there is no significant difference in the thermostability of the mutants compared to the WT protein. Consistent with the results of SEC profiling, the outcome of the thermal shift assays suggest that the Y59A, F155A, and F256A substitutions do not disrupt the folding or structural stability of the protein.

Figure 3.

The Y59A, F155A, and F256A mutations do not affect the thermostability of Purβ. Thermal shift assays were conducted to assess the structural stability of purified Purβ mutants Y59A, F155A, and F256A in comparison to Purβ WT. A) Protein unfolding was monitored as a function of temperature via differential scanning fluorimetry at multiple protein concentrations (50–160 μg/mL). Symbols show the normalized fluorescence intensity signal (F) averaged across multiple wells and experiments. Shadowing indicates the standard deviation of each mutant and black dashed lines highlight the standard deviation of the WT protein. B) Melting temperatures were determined from the midpoint of the normalized fluorescence intensity curve for each protein. Empirical T_m values were determined at different protein concentrations in three or four independent experiments. Bars show the mean ± SD (n = 12–15 trials). There were no significant differences in the T_m values determined for each mutant compared to Purβ WT based on one-way ANOVA and Dunnett’s multiple comparisons test.

As an additional screening tool to assess the structural fidelity of the Purβ point mutants, we evaluated the interaction of several Purβ antibodies with each mutant. Equimolar concentrations of Purβ WT and mutant proteins were immobilized on 96-well plates and then incubated with varying concentrations of primary antibodies directed against amino acids 210–229 or 302–324 of mouse Purβ. Solid-phase antibody-Purβ complexes were detected by ELISA. The resultant antibody binding profiles were essentially superimposable for Purβ WT and the Y59A, F155A, and F256A mutants using either pre-SEC or post-SEC purified proteins (Figure S4). A previously described P223L variant, which demonstrates reduced interaction with the 210–229 antibody, was included as an internal control.³⁷ These findings indicate that the amino acid substitutions do not affect the recognition of epitopes that are remote from the site of mutation and further highlight the overall structural similarity of the point mutants relative to Purβ WT.

Assessment of the ssDNA-Binding Activity of Purβ Y/F Mutants

The ssDNA-binding activity of each mutant was evaluated utilizing both qualitative and quantitative approaches. Initial screening was conducted by Southwestern blotting to gauge the overall ssDNA-binding capacity of each mutant relative to the WT protein. Southwestern blotting is technically analogous to Western blotting with the exception that a DNA probe is used in place of an antibody to assess the nucleic acid binding activity of a protein immobilized on a PVDF membrane. A caveat of this approach is that some renaturation of the protein molecules in the band must occur in order for a DNA-protein complex to form on the membrane. The biotinylated oligonucleotide probe used for detection of Purβ corresponds to a 32 nucleotide long purine-rich promoter sequence located in the Acta2 5’ flanking region (designated PE32-bF, Table S3). This sequence, which is also present in the VSMP8 plasmid used for Acta2 promoter-reporter assays, contains three distinct purine-rich elements and is bound by Purβ with high affinity (K_d ~ 0.2 nM) and a 2:1 protein:ssDNA stoichiometry.⁴⁹ As expected, Purβ WT was capable of recognizing and binding this probe after the protein was denatured, electrophoresed, and transferred to a PVDF membrane (Figure S5). The band intensity of the F256A mutant was similar to Purβ WT when compared at two different loading amounts. However, the Y59A and F155A mutants displayed much weaker signals than Purβ WT or F256A implying either reduced renaturation of these mutants, or more likely, a decrease in their overall ssDNA-binding activity.

A quantitative ELISA approach was next used to compare the DNA binding properties of Purβ WT to the Y59A, F155A, and F256A point mutants. This particular assay system has been described in previous publications and is well suited to detecting differences in the DNA binding affinities of Purβ mutants and variants in a biologically relevant range of concentrations.^{36, 37, 40} Our objective was to more definitively determine if the deficiencies in repressor activity of the Y59A and F155A mutants observed by Acta2 promoter-reporter assay in transfected cells could be due to changes in ssDNA-binding affinity or specificity. In this 96-well assay system, the purine-rich PE32-bF oligonucleotide from the Acta2 promoter was immobilized on streptavidin-coated wells at a limiting concentration of 0.5 nM. A pyrimidine-rich oligonucleotide dubbed 3I5T7 was also included in the assay as a control for the non-specific binding of Purβ to ssDNA. The 3I5T7 oligonucleotide is mutant version of PE32-bF containing seven consecutive thymines in place of each purine-rich element in the 3’, 5’ or internal region of PE32 (Table S3). The wells were then incubated with selected concentrations of purified Purβ WT or mutant proteins ranging from ~0.003 to 300 nM. Formation of solid-phase Purβ-ssDNA complexes was detected using a primary rabbit polyclonal antibody directed against residues 302–324 of mouse Purβ. This particular antibody was chosen because the 302–324 epitope is remote from the predicted DNA binding surfaces of Purβ and does not overlap with the Y59A, F155A, or F256A substitutions. Moreover, the 302–324 antibody recognizes Purβ-ssDNA complexes³³ and binds equivalently to Purβ WT and the Y/F point mutants (Figure S4).

Similar to the results of the cell-based assay, the Purβ Y59A and F155A mutants both exhibited a substantial reduction in binding to purine-rich ssDNA (Figure 4A). This was reflected in a significant increase in the logEC50 (~3 to 4-fold increase in EC50) of the Y59A and F155A mutants, which is a measure of relative binding affinity for PE32-bF (Figure 4B). Interestingly, the F256A mutant, which showed little to no defect in Acta2 repressor activity in the cell-based assay, also displayed a modest ~2-fold reduction in ssDNA-binding affinity in the ELISA. The shape of the binding curve corresponding to the Y59A mutant was particularly striking, as the curve plateaued at a maximum absorbance (B_max) roughly half that of the WT protein suggesting a possible difference in stoichiometry (Figure 4C). This is in contrast to the F155A and F256A mutants, for which less dramatic differences in the maximum absorbance were observed at high protein concentrations in comparison to Purβ WT. The attenuated binding of each aromatic mutant to purine-rich ssDNA was also seen if a different primary antibody recognizing residues 210–229 was used for detection of Purβ-ssDNA complexes (Figure S6).

Figure 4.

The Y59A, F155A, and F256A mutations reduce the interaction of Purβ with purine-rich ssDNA. A) The binding of each Purβ point mutant to ssDNA was assessed by ELISA using microplate wells coated with 0.5 nM of the 32 nt purine-rich recognition sequence from the Acta2 promoter (PE32). A pyrimidine-rich oligonucleotide in which all three PUR elements residing in PE32 were replaced with thymidylate (3I5T7) served as a control for binding specificity. Each Purβ mutant was tested over a range of protein concentrations. Purβ WT was assayed in parallel as a positive control. Wells without any DNA served as a control for background binding of each protein. The initial binding step was performed at 4°C while the immunological detection of Purβ-ssDNA complexes was done at room temperature using rabbit anti-mouse Purβ 302–324 as the primary antibody. Symbols with error bars represent the mean ± SD (n = 2–8) of data obtained in three independent experiments. Binding isotherms were generated by fitting the data points to a four-parameter dose-response curve. B, C) LogEC50 and B_max values were determined for Purβ WT, Y59A, F155A, and F256A based on fits to binding data obtained in each experiment. Bars show the mean ± SD (n = 3). p < 0.01 (**), p < 0.001 (***), p < 0.0001 (****) compared to WT based on one-way ANOVA and Dunnett’s multiple comparisons test.

Influence of Temperature, Salt, and DNA Sequence on Purβ Y/F Mutant Binding to ssDNA

The effect of temperature and monovalent salt on the interaction of the Purβ point mutants with ssDNA was also investigated to determine if the mutants behaved similarly under different assay conditions. The distinction in the ssDNA-binding activity of the mutants compared to Purβ WT was maintained if the primary binding step was performed at room temperature and without the addition of excess fluid-phase dT32 oligonucleotide (Figure S7). However, the non-specific binding of Purβ WT and the point mutants to the purine deficient 3I5T7 probe predictably increased in the absence of dT32. Furthermore, substitution of KCl for NaCl as the monovalent salt in the assay buffer did not appreciably alter the binding profile of each mutant relative to the WT protein (Figure 5). A slight increase in overall affinity (1.2 ± 0.1 fold) for PE32-F was observed across all proteins tested in buffer containing KCl. However, the identity of the salt did not significantly affect the proportionate differences in EC50 values determined for the Y59A, F155A, and F256A mutants relative to Purβ WT. These findings indicate that the inhibitory effects of the aromatic amino acid substitutions on Purβ binding to ssDNA are reproducible at different temperatures and in neutral buffers containing either Na⁺ or K⁺ ions.

Figure 5.

The Y59A, F155A, and F256A mutants exhibit binding isotherms consistent with reduced affinity for ssDNA in buffers containing different monovalent salts. A) The binding of each Purβ point mutant to ssDNA was assessed by ELISA using microplate wells coated with 0.5 nM of the 32 nt purine-rich recognition sequence from the Acta2 promoter (PE32). The assay was conducted in a buffer system containing either 150 mM NaCl (closed symbols) or KCl (open symbols). Each Purβ mutant was tested over a range of protein concentrations. Purβ WT was assayed in parallel as a positive control. Wells without any DNA served as a control for background binding of each protein. The initial binding step was performed at 4°C while the immunological detection of Purβ-ssDNA complexes was done at room temperature using rabbit anti-mouse Purβ 302–324 as the primary antibody. Data were normalized to the maximum A₄₀₅ of WT in NaCl buffer. Symbols with error bars represent the mean ± SD (n = 4) of data obtained in two independent experiments. Binding isotherms were generated by fitting the data points to a four-parameter dose-response curve. B) LogEC50 values were determined for Purβ WT, Y59A, F155A, and F256A based on fits to the binding data obtained in each experiment. Bars represent the mean with 95% confidence interval. Solid fill indicates NaCl, and hatched indicates KCl.

The multisite binding ability of each mutant was also evaluated by ELISA using a series of ssDNA targets in which each putative binding site in PE32-F was substituted with seven thymidylate (T7) residues. Oligonucleotides containing single, double, or triple substitutions of the 3’, 5’ or internal PUR-like elements (Table S3) were studied at a limiting concentration of 0.5 nM and purified protein concentrations of 1.0 or 10 nM. The ssDNA binding profiles of Purβ WT and F256A were quite similar across all the mutated oligonucleotides tested while Y59A and F155A demonstrated more unique features (Figure 6). Consistent with the notion that Purβ is capable of binding to multiple sites, mutation of two or more purine-rich elements had the greatest effect on reducing the interaction of Purβ WT or the F256A mutant with ssDNA. Interestingly, the Y59A and F155A mutants showed a more substantial decrease in binding to oligonucleotides in which the 5’ purine-rich element (GGGAGCA) was substituted with T7 relative to Purβ WT or the F256A mutant. This result suggests that the absence Y59 and F155 reduces the affinity and may alter the selectivity of Purβ for different purine-rich sequences in ssDNA.

Figure 6.

The Y59A and F155A mutants exhibit unique multisite binding profiles for ssDNA. A and B) The binding site preferences of each Purβ point mutant for ssDNA was assessed by ELISA using microplate wells coated with 0.5 nM of the Acta2 PE32-F element or various single, double, or triple mutants of PE32-F where the 3’, internal, or 5’ purine-rich sites were changed to seven consecutive thymidylates (Table S3). This screening assay was conducted in a buffer system containing KCl. Each Purβ mutant was tested at a fluid-phase concentration of 1.0 nM (A) or 10 nM (B). Purβ WT was assayed in parallel as a positive control. Wells without any DNA served as a control for background binding of each protein (dotted line). The initial binding step was performed at 4°C while the immunological detection of Purβ-ssDNA complexes was done at room temperature using rabbit anti-mouse Purβ 302–324 as the primary antibody. Raw (non-normalized) absorbance values are reported for each protein tested against each ssDNA probe. Bars represent the mean ± SD (n = 4) of data obtained in two independent experiments. The bars corresponding to Purβ WT control (gray) are shown behind each mutant (colored) for ease of comparison of the binding profile of each mutant relative to the WT protein.

Alternative Assessment of the ssDNA-Binding Activity of Purβ Y/F Mutants

A competition assay was also used to evaluate changes in the ssDNA-binding activity of the Purβ point mutants as a complement to the ssDNA-binding ELISA. The format of the competition assay was designed to eliminate the need for antibody-based detection of nucleoprotein complexes. This alternative assay assessed the ability of fluid-phase Purβ WT or mutant proteins to compete for the binding of biotinylated ssDNA to solid-phase Purβ WT immobilized on a microplate well. Purβ mutants possessing weaker ssDNA binding affinity would be predicted to generate competition curves that yield higher IC50 values compared to fluid-phase Purβ WT. In keeping with the results of the ssDNA-binding ELISA, each mutant displayed a competition curve that was shifted to the right relative to the Purβ WT control (Figure S8). The corresponding IC50 values of each mutant protein were consistent with reduced affinity for ssDNA compared to Purβ WT with the Y59A mutant demonstrating the highest IC50 value followed by F155A and F256A. While clearly less sensitive than the ssDNA-binding ELISA, the competition assay corroborated the finding that the Y59A and F155A mutants are indeed defective in binding to ssDNA.

Analysis of the Ybx1 Binding Activity of Purβ Y/F Mutants

Purβ is known to bind to Ybx1 and protein-protein interaction between these co-repressors may be important for elaboration of the Acta2 repressor activity of Purβ.^{9, 10, 33} Consequently, the Purβ Y/F point mutants were tested for their ability to bind Ybx1 using a quantitative ELISA approach. Purified recombinant Ybx1 was immobilized on microplate wells at a concentration of 100 nM and then incubated with a fixed concentration of fluid-phase Purβ point mutants or WT as a control. Formation of solid-phase Purβ-Ybx1 complexes was detected using a primary antibody against Purβ residues 302–324 or 210–229. Irrespective of the specific antibody detection system, the Y59A and F155A mutants appeared to be deficient in binding to Ybx1 (Figure S9). Titration assays performed over a range of Purβ concentrations further demonstrated that the Y59A and F155A mutants have a 2.5- and 1.8-fold lower affinity for Ybx1 than the WT protein (Figure 7). The F256A mutant was similarly tested but generated a much more complex binding isotherm that could not be fit to a conventional binding model (Figure 7). The curve for the F256A was shifted slightly to the right of Purβ WT at protein concentrations from 10 to 100 nM, an occurrence consistent with somewhat weaker binding affinity. At concentrations in excess of 100 nM, the curve for F256A consistently showed a negative downward slope. This phenomenon could be indicative of a difference in F256A monomer-dimer equilibrium or perhaps reduced solubility of F256A at high protein concentrations. Regardless of the physical basis for the non-hyperbolic binding isotherm seen with the F256A mutant, it is apparent that all three aromatic mutants form less stable complexes with Ybx1.

Figure 7.

The Y59A and F155A mutants demonstrate reduced binding affinity for the Acta2 co-repressor Ybx1. The binding of each Purβ point mutant to Ybx1 was assessed by ELISA using microplate wells coated with 100 nM of purified mouse Ybx1. This titration assay was conducted in a buffer system containing NaCl. Each Purβ mutant was tested at a fluid-phase concentration ranging from 6.3 to 400 nM. Purβ WT was assayed in parallel as a positive control. Wells coated with BSA only served as a control for background binding of each protein. The initial binding step was performed at 4°C while the immunological detection of Purβ-Ybx1 complexes was done at room temperature using rabbit anti-mouse Purβ 302–324 as the primary antibody. Absorbance values were corrected for background binding to BSA and normalized to the maximum absorbance obtained for Purβ WT. Symbols with error bars represent the mean ± SD (n = 4) of data obtained in two independent experiments. Binding isotherms for Purβ WT, Y59A, and F155A were generated by fitting data points to a hyperbola, while the binding isotherm for F256A was generated using an Akima spline fit.

Because of the implication that residues Y59 and F155 are involved in mediating the interaction of Purβ with purine-rich ssDNA and with Ybx1, we sought to determine the effect of ssDNA on the formation of Purβ-Ybx1 complexes. The interaction of Purβ with Ybx1 was evaluated by ELISA in the presence of increasing concentrations of fluid-phase PE32-F ssDNA. As demonstrated in Figure S10, the presence of ssDNA at stoichiometric concentrations actually increased the binding of Purβ to Ybx1, or at least enhanced the stability of Purβ-Ybx1 complex. This result suggests that Purβ can interact simultaneously with purine-rich ssDNA and Ybx1. This interpretation is consistent with the predicted structure of the Purβ homodimer in which there are two separate intramolecular type I domains containing residues Y59 and F155 (Figure 1C). Moreover, the functional importance of these particular aromatic residues in combination is further substantiated by the finding that a Purβ Y59A/F155A double mutant is almost completely defective in Acta2 repressor activity when expressed in mouse fibroblasts (Figure S11).

DISCUSSION

Previous work from our group indicated that a combination of hydrophobic and electrostatic interactions are likely responsible for mediating the physical and functional interaction of Purβ with ssDNA and the Acta2 gene promoter.^{36, 40} Consistent with the importance of electrostatic interaction, several basic amino acid residues located in PUR repeats I (K82), II (R159), and III (R267) of mouse Purβ were found to be critical for the ssDNA-binding and Acta2 repressor functions of Purβ.⁴⁰ In this study, we now identify two aromatic residues (Y59 and F155) that appear to be responsible for mediating contact between Purβ and ssDNA. Interestingly, the same residues also appear to stabilize the interaction of Purβ with co-repressor Ybx1.

As early as 1992 and based on sequence analysis alone, Bergemann et al. predicted the presence of base-stacking interactions between conserved Phe and Tyr residues in human Purα and purine-rich ssDNA containing a core GGAGG recognition element.⁵⁰ Several of these predicted/positionally conserved aromatic residues were subsequently tested for effects on DNA binding based on the x-ray crystal structure of the isolated Dm Purα repeat I-II domain in complex with ssDNA.³⁵ The high resolution structure of Dm Purα I-II bound to a small oligonucleotide (PDB ID: 5FGP) indicated direct association of Y57 in repeat I and F145 in repeat II with ssDNA via pi-pi stacking interactions with specific purine nucleobases. Biochemical assays conducted with various point mutants of Dm Purα I-II validated the importance of F145 in mediating protein binding to ssDNA and in facilitating the unwinding of double-stranded DNA (dsDNA).³⁵ In contrast, mutation of residue F68 in repeat I (a positional homolog of F145 in repeat II) had little to no effect on the ssDNA-binding or dsDNA-unwinding activity of Dm Purα I-II.³⁵ This finding is consistent with an earlier study indicating that mutation of the homologous F70 residue in mouse Purβ did not affect its Acta2 repressor function.⁴⁵ With the exception of F70 and F274, the role of specific hydrophobic residues in mediating the functional properties of Purβ had not been previously explored.^{37, 45} Since several solvent-exposed aromatic residues are located in analogous positions in repeats I (Y59), II (F155), and III (F256) of mouse Purβ, our goal was to determine if these residues contribute to the transcriptional repressor and ssDNA-binding functions of Purβ.

The relative activities of the Purβ point mutants Y59A, F155A, and F256A were initially evaluated using promoter-reporter assays in transfected mouse fibroblasts. These experiments revealed that the Purβ Y59A mutant had markedly reduced repressor activity for Acta2 compared to Purβ WT, while the F155A mutant also exhibited diminished repressor function but to a lesser degree than Y59A. The F256A mutant was the least compromised and was not significantly different from Purβ WT. A double mutant containing both the Y59A and F155A substitutions was found to be virtually devoid of transcriptional repressor activity in transfected fibroblasts. These results indicated that aromatic residues in repeats I and II are indeed necessary for the Acta2 repressor activity of Purβ in living cells, which compelled us to investigate if these residues mediate the physical interaction of Purβ with purine-rich ssDNA.

Comparing our computational model of the Purβ homodimer³⁷ with the crystal structure of Dm Purα I-II bound to ssDNA³⁵, it appears highly likely that Y59 and F155 are capable of base stacking interactions with ssDNA. Consistent with this notion, the purified Y59A and F155A mutants exhibited a lower affinity for the purine-rich target sequence PE32-F derived from the Acta2 promoter when tested in vitro. This Acta2 cis-element is emblematic of the structural features that determine the specific and high affinity interaction by mammalian Purβ, namely, a high proportion of G residues arranged in repetitive GGA or GGN triplets in a ~22–35 nt long sequence.^{9, 12–14} Irrespective of the particular ssDNA-binding assay employed (ELISA, competition assay, Southwestern blot) or adjustments made in the experimental conditions (temperature, monovalent cation, excess ssDNA), an identical trend was observed in which Purβ WT and F256A bound to PE32-F with highest affinity followed by F155A and Y59A with lower affinity (WT ≥ F256A > F155A ≥ Y59A). The ssDNA-binding ELISA was intentionally performed at low (nanomolar) protein concentrations. Under these conditions, the monomeric form of free Purβ would likely dominate⁴¹ and the assembly of a high-affinity 2:1 Purβ:PE32-F complex would be predicted to occur via a sequential and cooperative binding mechanism involving at least two non-identical sites in the ssDNA.⁴⁹ Hence, the demonstrably weaker Acta2 repressor activity of the Y59A and F155A mutants in transfected cells is probably due, in part, to a defect in the ability of each mutant to form stable, high affinity complexes with ssDNA sequences in the Acta2 promoter.

Another intriguing finding from this study is that the Y59A and F155A substitutions apparently influence the binding of Purβ to different purine-rich sites in ssDNA. The interaction of the Y59A and F155A mutants with ssDNA was more dramatically reduced than that of Purβ WT or F256A when the 5’ purine-rich site in PE32-F was substituted with seven consecutive thymidylates. A similar trend was evident when the internal site of PE32-F was mutated, albeit to a lesser degree. Given the ability of the modular Purβ homodimer to interact with multiple DNA binding sites, these results suggest that Y59 and F155 could contribute to the recognition of particular purine-rich sites in PE32-F. For example, mutation of Y59 may reduce or eliminate ssDNA-binding by repeat I, leaving the sequence “preferences” of repeat II and III to dictate the overall specificity of the dimer. Alternatively, mutation of residues Y59 or F155 in repeats I or II may alter the cooperative binding of Purβ to different (non-identical) sites within the ssDNA.⁴⁹ This inference is consistent with the distinct patterns of ssDNA interaction generated when the mutants were presented with a PE32-F probe containing one or more putative binding site mutations. The modular nature of Purβ makes the study of its sequence-dependent ssDNA binding properties very challenging owing to an incomplete understanding of the degree to which each repeat contributes to the recognition of purine-rich ssDNA.³⁶ Therefore, the contribution of Y59 and F155 to the “sequence specificity” of Purβ must be interpreted in a cautious manner at the present time.

In contrast to the deficiencies found in Y59A and F155A mutants for both ssDNA-binding and transcription factor activity, the repeat III aromatic mutant F256A showed no significant reduction in repressor function as judged by the Acta2 promoter-luciferase assay in transfected fibroblasts. Interestingly, the purified mutant did exhibit a modest but significantly lower affinity (~2-fold) for ssDNA compared to the WT protein in vitro. A potential reason for the discrepancy seen in the results obtained with purified protein compared to the cell-based assay could be that the absence of F256 does not impair binding to ssDNA enough to cause a reduction in Acta2 repressor activity. Alternatively, it is possible that F256 does not contribute significantly to the binding of Purβ to ssDNA in the more complex cellular milieu. Yet another possibility is that other solvent-exposed aromatic residues in repeat III are involved in ssDNA interaction such as F245, Y253, Y266 or F274.³⁷ An interesting feature of Purβ repeat III is that it binds ssDNA with much higher affinity than domain I-II in vitro.³⁶ It also contains more solvent exposed aromatic residues than repeat I or II (III = 4–6 residues compared with I = 2–3, and II = 3–4). This begs the question as to whether there is some redundancy in DNA binding modes within repeat III, which may contribute to its higher affinity for ssDNA. In this case, mutation of multiple aromatic residues in repeat III may be required to observe a reduction in repressor activity.

The diminished Acta2 repressor activity of the Purβ Y59A and F155A mutants may not be due exclusively to impaired binding to ssDNA. A somewhat unexpected finding is that the Y59A and F155A mutations also reduce the interaction of Purβ with mouse Ybx1, another ssDNA/RNA-binding protein involved in Acta2 repression. Purβ and Ybx1 are thought to act together to repress transcription of Acta2 by interfering with the activity of the MCAT-interacting trans-activator TEF1.⁹ A consensus MCAT motif is found at the center of the Purβ/Ybx1 binding element referred to in this study as PE32. The opposing purine-rich and pyrimidine-rich strands of this sequence support the binding of Purβ and Ybx1 co-repressors, which presumably compete for the interaction of TEF1 with dsDNA or with other trans-activators.^{9, 33} Purβ and Ybx1 are also known to stably associate as a heterotypic protein complex in the absence of DNA.^{9, 37, 40, 45} Complex formation between Purβ and Ybx1 may allow the proteins to traffic together resulting in more efficient targeting of DNA regulatory elements such as the two complementary strands of the PE32 element in Acta2 in a coupled manner. Furthermore, because Purβ likely functions as a homodimer, it is possible that Y59 (or F155) in the type I domain of one subunit could interact with Ybx1 while Y59 (or F155) in the type I domain of the other subunit could mediate binding to ssDNA. This notion is supported by the observation that excess purine-rich ssDNA does not disrupt the binding of Purβ to Ybx1 at DNA and protein concentrations favoring Purβ dimerization. Thus, a reduction in Purβ-Ybx1 interaction as demonstrated by the Y59A and F155A mutants may result in less efficient recruitment of Purβ and Ybx1 to their respective ssDNA sites in Acta2, which could contribute to impaired co-repressor activity.

Based on computational models of Purβ protein structure, the putative DNA binding surfaces of repeats I and II are positively charged and repeat III also contains positively charged residues, consistent with the ability to interact with the negatively charged backbone of ssDNA. This assertion is supported by our previous findings that basic residues K82, R159, R243, and R267 contribute to the transcriptional repressor and/or high affinity ssDNA-binding activities of Purβ.^{40, 45} Residues Y59, F155, and F256 are surrounded by positively charged residues within the predicted DNA binding surfaces. An important feature of such solvent exposed aromatic residues is their ability to undergo stabilizing pi-pi stacking interactions with nucleobases.⁵¹ As demonstrated in Dm Purα,³⁵ the comparable residues in Mm Purβ, namely Y59 and F155, are likely available for base-stacking interactions. Based on analogy to F145 in Dm Purα, residue F155 in Purβ may also be involved in the destabilization/unwinding of duplex DNA. Although this biochemical property was not empirically evaluated in this study, others have demonstrated that Purβ does indeed possess DNA strand separation activity in vitro.⁵²

It is important to note that nearly all vertebrates carry genes for at least three different PUR proteins: Purα, Purβ, and Purγ. In contrast, the fruit fly and other invertebrates have only one PUR protein, which has not been definitively related more closely to either Purα or Purβ in humans or mice. Thus, we refrain from drawing strict comparisons between the function of aromatic residues in “Purα” versus “Purβ” in this report. However, some previous studies have revealed both similarities and differences between Purα and Purβ at levels of protein structure and function. This dichotomy is important to consider when extrapolating the contributions of each paralog to normal biological and disease processes. For example, we previously reported that knockdown of Purα but not Purβ in MEFs enhances cell proliferation.³⁶ In contrast, overexpression of either Purα or Purβ was able to rescue neuronal cells from GGGGCC repeat-induced cell death in a model of amyotrophic lateral sclerosis and frontotemporal dementia.⁵³ Furthermore, manipulation of Purα and Purβ expression in various fibroblast and muscle cell types suggests overlapping and distinct roles for each protein in regulating cell phenotype.^{10, 11, 54} These functional similarities and differences undoubtedly derive from shared and distinct structural and biophysical characteristics of Purα and Purβ. The most obvious shared feature is the sequence and apparent structural similarity of the repeat I-II and repeat III domains of Purα and Purβ seen across multiple orthologs.

This study supports the concept that certain positionally conserved aromatic residues in mammalian and invertebrate PUR proteins are also functionally similar. For example, the F155 residue described herein is necessary for the ssDNA-binding and transcription factor activities of Mm Purβ while the homologous F145 residue apparently mediates the ssDNA/RNA-binding, dsDNA-unwinding, and neurodevelopmental activities of Dm Purα.³⁵ However, other aromatic residues such as F70 in Mm Purβ and the homolog F68 in Dm Purα do not appear to play a significant role in the nucleic acid binding functions of each protein.^{35, 45} On the other hand, differences in subdomain structure may contribute to the functional divergence of PUR protein orthologs. We previously showed that the isolated repeat III dimerization domain of mouse Purβ binds to ssDNA with higher affinity compared to the I-II domain (III EC50 ~ 1.8 nM, I-II EC50 >100 nM, PE32-bF probe).³⁶ However, the opposite pattern was reported for Dm Purα in that the repeat III domain had only weak ssDNA-binding affinity, while repeat I-II bound ssDNA with higher affinity (I-II K_d ~ 0.3 μM, III K_d ~ 10.2 μM, MF0677 probe) in vitro.³⁵ This could point to repeat III as a possible source of the difference in biological functions between invertebrate and mammalian PUR proteins. It is still unknown whether the functions of positional homologs Y59 in Mm Purβ and Y57 in Dm Purα are similar or different. In the Dm Purα crystal structure, Y57 is observed to base-stack with terminal guanine nucleobases suggesting a direct role in mediating protein binding to ssDNA.³⁵ Our findings indicate that Y59 in Mm Purβ is indeed important for the ssDNA-binding and transcriptional repressor activities of the protein. If the pattern we found is the same for Dm Purα Y57, this particular residue may play a key role in dictating the biological and biochemical functions of PUR proteins in multiple species. Similarly, this residue would be a good candidate for mutagenesis across PUR paralogs to examine the contribution of specific aromatic amino acids to the independent biological activities of Purα, Purβ, and Purγ.

In summary, our findings highlight the importance of aromatic residues in mediating the ssDNA-binding and transcription factor activities of Purβ. Consistent with our previous study of basic residues³⁷ and insights gleaned from the crystal structure of Dm Purα in complex with ssDNA³⁵, our analysis of aromatic residues herein support the presence of at least three distinct DNA binding surfaces, each associated with a single PUR repeat. We do not exclude the possibility that multiple DNA binding modes could be present in a single repeat, and which may in fact allow for differential nucleic acid recognition based on nucleotide sequence, number of binding sites, polynucleotide identity (RNA vs. DNA) and/or nucleic acid secondary structure (linear, G-quadruplex etc.). Our data also reveal the presence of at least two multipurpose residues (Y59 and F155), which serve ssDNA- and Ybx1-binding functions. Moreover, our results suggest that residues Y59 and F155 in conjunction with the modular construction of Purβ are critical factors in explaining the unique ability of the protein to preferentially bind to purine-rich ssDNA in a manner affecting gene expression.

ABBREVIATIONS

BSA

bovine serum albumin

Dm

Drosophila melanogaster

dsDNA

double-stranded DNA

ELISA

enzyme-linked immunosorbent assay

FBS

fetal bovine serum

GAPDH

glyceraldehyde-3-phosphate dehydrogenase

HRP

horseradish peroxidase

MCAT

muscle-CAT box

MEF

mouse embryo fibroblast

Mm

Mus musculus

nt

nucleotide

ORF

open reading frame

Purα

purine-rich element binding protein A

Purβ

purine-rich element binding protein B

Purγ

purine-rich element binding protein G

SDS-PAGE

sodium dodecyl sulfate-polyacrylamide gel electrophoresis

SEC

size exclusion chromatography

SD

standard deviation

SEM

standard error of the mean

ssDNA

single-stranded DNA

WT

wild-type