The spliceosomal proteins PPIH and PRPF4 exhibit bi-partite binding

Abstract

Pre-mRNA splicing is a dynamic, multistep process that is catalyzed by the RNA (ribonucleic acid)–protein complex called the spliceosome. The spliceosome contains a core set of RNAs and proteins that are conserved in all organisms that perform splicing. In higher organisms, peptidyl-prolyl isomerase H (PPIH) directly interacts with the core protein pre-mRNA processing factor 4 (PRPF4) and both integrate into the pre-catalytic spliceosome as part of the tri-snRNP (small nuclear RNA–protein complex) subcomplex. As a first step to understand the protein interactions that dictate PPIH and PRPF4 function, we expressed and purified soluble forms of each protein and formed a complex between them. We found two sites of interaction between PPIH and the N-terminus of PRPF4, an unexpected result. The N-terminus of PRPF4 is an intrinsically disordered region and does not adopt secondary structure in the presence of PPIH. In the absence of an atomic resolution structure, we used mutational analysis to identify point mutations that uncouple these two binding sites and find that mutations in both sites are necessary to break up the complex. A discussion of how this bipartite interaction between PPIH and PRPF4 may modulate spliceosomal function is included.

Introduction

The spliceosome is a complex and dynamic macromolecular machine that removes intronic sequences from messenger RNA (pre-mRNA) and joins coding sequences together. The spliceosome is composed of five uridine-rich small nuclear RNAs (snRNAs) that interact with essential protein chaperones to form snRNP (small nuclear RNA–protein complex) subcomplexes. Human snRNPs contain ~90 conserved spliceosomal proteins, which in addition to chaperones include proteins that assist in recognizing splice sites and positioning the pre-mRNA and snRNAs appropriately for the two catalytic steps of splicing [1–4]. Comparison of purified yeast and metazoan spliceosomes has found additional protein factors that abundantly incorporate into snRNPs and into intermediate and catalytic spliceosomal complexes [5–7]. These additional proteins are likely to play regulatory roles in the organisms they are found in, but many contain no orthologs in lower organisms or homology to annotated domains. This hinders study of their function.

On the other hand, the study of conserved spliceosomal proteins in tractable model organisms has contributed greatly to functional characterization [8,9]. One example is pre-mRNA processing factor 4 (PRPF4), named due to defects in pre-mRNA processing when deleted from Saccharomyces cerevisiae. PRPF4 is found as part of the U4/U5/U6 tri-snRNP. Tri-snRNP integrates into the spliceosome by interacting with U2 snRNP, creating the intermediate B complex [4,10]. Mutations in several tri-snRNP factors, including PRPF4, cause autosomal dominant versions of the neurodegenerative disease retinitis pigmentosa (adRP) by compromising tri-snRNP stability [11–14]. The disease-causing mutation is in the highly conserved C-terminal region of PRPF4, which interacts with pre-mRNA processing factor 3 (PRPF3). Destabilization of tri-snRNP results in aberrant mRNA splicing that contributes to the disease. This highlights the contribution of protein–protein interactions to proper spliceosome function, in addition to protein–RNA interactions [11–14].

The N-terminal region of PRPF4 is not conserved in S. cerevisiae, but is found in a variety of eukaryotes. A short 30-mer peptide derived from this ~200-residue region was shown to interact with peptidyl-prolyl isomerase H (PPIH), which is also not conserved in S. cerevisiae [15–17]. PPIH belongs to a family of proteins called cyclophilins that can catalyze the interconversion between cis and trans stereoisomers of proline, inducing local changes in protein conformation around the isomerization site [18,19]. Cyclophilins are evolutionarily ancient, with homologs existing in Escherichia coli, Plasmodium falciparum, and viruses [20–22]. There are eight nuclear-localized cyclophilins in human, and all eight are found abundantly in human spliceosome. PPIH abundantly associates with tri-snRNP and intermediates B splicing complexes [1,4,17,23–25]. Cyclophilins are assumed to function as prolyl isomerases in vivo, but previous studies have found no role for the isomerase active site in mediating spliceosomal interactions [16,24,26]. More recently, we showed that incorporation of the nuclear cyclophilins PPIH, PPIL2, CWC27, and PPIG into assembling spliceosome is important for proper spliceosomal assembly and catalytic function in vitro. This function was also independent of isomerase activity [27].

A spliceosome-wide siRNA screen recently provided direct evidence that both PRPF4 and PPIH regulate alternative splicing in human cells, and indeed that tri-snRNP proteins are over-represented in this process [28]. This study indicates that modulating protein–protein interactions within the spliceosome can regulate alternative splicing. As part of a larger goal to understand how spliceosomal proteins interact with each other in order to modulate splicing, we have undertaken an in-depth analysis of the PPIH–PRPF4 interaction using human genes expressed and purified from bacteria. We found a novel interaction between PPIH and PRPF4, and optimized the previous interaction to achieve nanomolar affinity. We characterized these interactions and found a mutational scheme that completely eliminates binding between PPIH and PRPF4. This work identifies a new bipartite binding interface between PPIH and PRPF4, and indicates that there is a role for the isomerase active site in spliceosomal protein interaction.

Experimental procedures

Cloning, expression, and purification of PPIH, PPIA, and PRPF4

Subcloning of full-length PPIH (BC003412) into pET15-MHL vector (EF456738) was described previously [19]. PPIA (BC137057) and PRPF4 (BC001588) constructs were subcloned into pET15-MHL vector using the In-Fusion EcoHD kit (Clontech). Mutant PPIH and PRPF4 constructs were generated using mutagenic PCR primers (see Supplementary Methods for all oligo sequences). After PCR, samples were treated with DpnI (New England Biolabs) for 1 h at 37°C before transforming into DH5α cells (New England Biolabs). Plasmids obtained from positive colonies were purified using the Zyppy Plasmid Miniprep Kit (Zymo Research) and sequenced to verify mutation.

For protein expression, 50 ng of plasmid DNA was transformed into BL21(DE3) cells (Life Technologies). These cells were grown in Terrific Broth (Research Products International) at 37°C using a LEX bioreactor (Epiphyte), induced using 100 µM IPTG (CellGro) at 15°C, and harvested 12 h later by centrifugation at 4000×g for 20 min. PPIA was prepared in the same manner, except expression was performed at 25°C overnight. Pellets were stored at −80°C until use. Frozen bacterial cell pellets were thawed in warm water and then resuspended in lysis buffer [50 mM Tris (pH 8), 500 mM NaCl, 3 : 1 ratio of BPER : YPER detergents (Thermo Fisher), DNase I, lysozyme, and Protease Inhibitor P2714 (all from Sigma)]. Cells were sonicated (frequency 50% maximum, 30 s on/30 s off, 5 min sonication time, Virsonic).

Immobilized metal ion chromatography (IMAC) purification of PPIH was as in Davis et al. [19], with the substitution of His-Link resin (Promega) for a Ni-NTA resin. The protein contains an N-terminal glycine and serine was added to the native sequence, left behind after protease cleavage of a glutathione-S-transferase fusion tag. Original preparations of PRPF4 followed the same purification procedures as those for PPIH, but after identifying PRPF4 as an intrinsically disordered protein, a denaturing purification scheme was optimized in order to remove associated E. coli proteins. For this protocol, 8 M urea was added to lysis buffer to make denaturing buffer, and after sonication, lysate was mixed with His-Link resin for 30 min at 4°C. Post-incubation, the resin was placed in a gravity column and washed sequentially with denaturing buffer, wash buffer [denaturing buffer + 10 mM imidazole (pH 8)], and elution buffer [denaturing buffer + 250 mM imidazole (pH 8)]. The eluate was then sequentially dialyzed in dilution buffers 1 [50 mM Tris (pH 8), 500 mM NaCl, 6 M urea, 1 M l-Arginine, and 5 mM β-ME], 2 [50 mM Tris (pH 8), 500 mM NaCl, 4 M urea, 1 M l-Arginine, and 5 mM β-ME], and 3 [50 mM Tris (pH 8), 500 mM NaCl, 2 M urea, 0.05 M l-Arginine, 5% glycerol, and 5 mM β-ME] for 1 h each before being filtered through a 0.45-µm syringe filter and run over a gel filtration column. Size exclusion chromatography (SEC) for all proteins was accomplished using a Sephadex 200 column (GE Healthcare) on an NGC Chromatography System (Bio-Rad) using gel filtration buffer [50 mM Tris (pH 8), 500 mM NaCl, 5 mM β-mercaptoethanol (β-ME), and 1 mM EDTA]. Fractions were analyzed using SDS–PAGE and pooled. If needed, protein was concentrated using Vivaspin-20 concentrators with a 5000 or 10 000 MWCO cutoff (GE Healthcare). ¹⁵N-¹H heteronuclear single quantum correlation (HSQC) spectroscopy was performed on protein obtained through both soluble and denaturing purification methods, and spectra were identical (data not shown).

Isothermal calorimetry

Experiments were performed in a Nano ITC (TA Instruments). Purified PPIH (40–100 µM) in isothermal calorimetry (ITC) buffer [50 mM HEPES (pH 8), 200 mM NaCl, and 5 mM β-ME] was placed into the sample cell, and PRPF4 proteins were injected at concentrations of 0.5–1.25 mM. Injections of 1–2 µl were performed at 25°C with an interval of 1 s and a constant stirring rate of 250 rpm. All samples were degassed for 10–30 min before each run. All proteins were injected into ITC buffer, which was used to perform baseline subtraction. The corrected data were analyzed and the curves were fit to an independent binding model using NanoAnalyze (TA instruments).

Surface plasmon resonance

A ProteOn XPR-36 instrument (Bio-Rad) was used to conduct surface plasmon resonance (SPR) experiments at 25°C. PPIH, W133A PPIH mutant, or PPIA was immobilized on a ProteON GLC chip (Bio-Rad, Hercules, CA). The chip channels were activated by injecting a solution of 25 mM N-ethyl-N′-(3-dimethylaminopropyl) carbodi-imide (EDC) freshly mixed with 8 mM sulfo-N-hydroxysuccinimide (sulfo-NHS) at a flow rate of 50 µl/min. Ligands were diluted in acetate buffer (pH 5.5, Bio-Rad Laboratories) to a final concentration of 0.5 µM and immobilized to 2000 response units (RUs). Excess ester groups on the sensor surface were capped with the injection of 1 M ethanolamine–HCl (pH 8.5, Bio-Rad) at a flow rate of 50 µl/min. Chip equilibration into SPR buffer (20 mM Tris–HCl pH 7.5, 150 mM NaCl, 5 mM β-ME, 1 mM EDTA, and 0.01% Tween) was performed overnight. Proteins used as analytes were dialyzed into SPR buffer before use to minimize bulk effects. Binding sensorgrams were collected and initially processed using the integrated ProteOn Manager software (Bio-Rad Laboratories). Analyte responses were corrected for nonspecific interactions using a control human IL-2 (Bio-Rad) immobilized to 2000 RU. Corrected RUs were plotted against analyte protein concentration and fitted to a one-site specific binding model using the GraphPad Prism to obtain equilibrium-binding constants. All results shown are representative of at least three separate experiments.

Circular dichroism

Circular dichroism (CD) spectra of single proteins and complexes were taken using a J-810 Spectropolarimeter ( JASCO). Wavelength scans from 196 to 260 nm were carried out using a 1-mm pathlength cuvette in 30 mM sodium phosphate buffer (pH 7.5) and 75 mM NaCl. A target concentration of 8 µM was used for single proteins and for complexes after optimization of CD signal and HT voltage values. For complexes, protein concentrations measured after spectra were taken results in predicted binding saturation of 76% (PRPF4_1–169), 90% (PRPF4_106–169), and 33% (PRPF4_1–98). Spectra shown are a buffer-subtracted average of three scans and collected over 0.2 nm intervals at 25°C. The web server CAPITO was used to evaluate the % secondary structure and fold state (unfolded, molten globule, and globular) [29,30]. Datasets of folded and unfolded protein CD spectra are available from the CAPITO website for analysis. Thermal denaturation curves were recorded at a wavelength of 220 nm from 25 to 75°C. Three spectra were recorded at every 1°C interval and averaged. Non-linear regression fit to a Boltzmann sigmoidal equation in the GraphPad Prism was used to calculate T_m.

Analytical ultracentrifugation

Sedimentation velocity (SV) and sedimentation equilibrium (SE) experiments were carried out with a ProteomeLab XL-I analytical ultracentrifuge (Beckman Coulter) using a 50 Ti rotor. Concentrations of proteins for SE were 0.3–0.5 mM and for SV from 0.3 to 0.9 mM, depending on the predicted extinction molar coefficient for each protein or complex. Analytical ultracentrifugation analysis (AUC) buffer was 20 mM Tris–HCl (pH 8), 500 mM NaCl, 5 mM β-ME, and 1 mM EDTA. After cell assembly of a six sector centerpiece for SE, 100 µl of protein was added to the three sample chambers and 120 µl of buffer was added to each of the three buffer chambers. Samples were spun at 2400, 6600, 15 000, 26 500, 41 500, 60 000, and 80 000×g until all samples sedimented (~72 h). Four hundred microliters of protein were added to the SV 2 sector center piece sample chamber and 420 µl of buffer to the buffer chamber before spinning samples at 135 000×g for ~18 h. SV data were fitted using a continuous sedimentation coefficient [c(s)] distribution model in SEDFIT [31]. For SE data, molecular mass analysis was performed using the ‘species analysis’ model available in SEDPHAT with RI (radial-invariant) noise baseline correction [32]. Global fit to four to six runs was used to calculate the reported molecular mass. For analysis of complexes, equimolar complexes were formed with a minimum concentration of 300 µM. This resulted in a minimum predicted binding saturation of 96% (PRPF4_1–169), 98% (PRPF4_106–169), and 83% (PRPF4_1–98).

Thermal stability assays

5000× SYPRO Orange (Sigma) was diluted to 200× in gel filtration buffer [500 mM NaCl, 20 mM Tris (pH 8), 5 mM β-ME, and 1 mM EDTA]. One microliter of dye was added to 9 µl of purified protein, and a thermal ramp was run from 25 to 99°C in 0.5°C/s increments. Data were collected and analyzed using a CFX96 Touch™ Real-Time PCR Detection System (Bio-Rad). Non-linear regression fit to a Boltzmann sigmoidal equation in the GraphPad Prism was used to calculate T_m. For analysis of complexes, PPIH was held at a constant 25 µM, and PRPF4 proteins were titrated from 25 to 75 µM. This resulted in a range of predicted binding saturations of 86–99% (PRPF4_1–169), 95–99.8% (PRPF4_106–169), and 52–83% (PRPF4_1–98).

Results

Identification of soluble PRPF4 protein for protein–protein interaction analysis with PPIH

PRPF4 is a 522 amino acid protein. The gene is dominated by a C-terminal WD-40 domain (residues 220–522), which includes the site for interaction with PRPF3 and the mutation that causes dominant negative retinitis pigmentosa [12]. The N-terminal region of PRPF4 includes a site of interaction with PPIH [15,17], but initial publications focused on 30 of the N-terminal 220 residues [16] (Figure 1). We undertook expression trials, extending our analysis to as much of the full-length PRPF4 protein as possible. Nine constructs spanning the full-length gene (BC001588) were subcloned into the pET15-MHL vector (Addgene 26092) encoding a cleavable N-terminal hexa-histidine tag, and expression trials were conducted in the BL21(DE3) strain of E. coli. All constructs containing the WD-40 domain were insoluble; the longest construct that yielded soluble protein was one encoding for residues 1–169 (referred to throughout as PRPF4_1–169). PRPF4_1–169 was purified to homogeneity using Ni²⁺-IMAC and SEC (Figure 2), with typical yields of 5–10 mg/l of cell culture. Cloning, expression, and purification of the full-length gene for PPIH (BC003412) have been described previously [19].

Figure 1. Identification of soluble PRPF4 proteins for complex with PPIH.

(A) A 30-mer peptide from PRPF4 interacts with PPIH without contacting the PPIH active site. (Left) A cartoon representation of PPIH (teal) shown interacting with a 30-residue peptide isolated from PRPF4 (goldenrod, residues 106–137) (PDB:1MZW) [16]. The isomerase active surface of PPIH is shown in stick representation, as is the PRPF4 residue F122 mentioned in the text. (Middle) The model is rotated to emphasize the distance between F122 within PRPF4 and the PPIH residue W133. (Right) PPIH is shown in electrostatic surface representation. F122 from PRPF4 inserts into a pocket formed in part by the PPIH α1–β3 loop. (B) Domain architecture of PRPF4. The splicing factor motif (SFM), which includes residues 103–155, is defined by the SMART domain server (http://smart.embl-heidelberg.de/). This motif includes the sequence of the 30-mer peptide shown in (A).

Figure 2. ITC indicates bipartite binding between PRPF4 and PPIH proteins.

(A) Graphical representation of constructs used in the current study. (B and C) SEC was performed on (B) PPIH (teal), (C) PRPF4_1–169 (fuchsia), PRPF4_106–169 (sunset), and PRPF4_1–98 (blue). Representative chromatograms are shown. All four proteins are well-behaved and homogeneous, although PRPF4_1–169 and PRPF4_1–98 exhibit retention volumes indicative of extended protein. (C, middle) Representative gels showing SEC fractions of purified proteins. (C, bottom) ITC indicates that PRPF4_1–169 interacts with PPIH with 0.48 µM affinity and two potential binding sites. A K_d value of 0.08 µM is modeled for PRPF4_106–169 interaction with PPIH, and affinity was not calculated for PRPF4_1–98 as saturation was not reached.

PRPF4_1–169 binds to PPIH using two interaction sites

We first addressed the question of whether PRPF4_1–169 bound to PPIH using ITC. We found that PRPF4_1–169 bound to PPIH with a K_d value of 0.5 µM. This is tighter than the 2 µM affinity reported previously for the PRPF4 peptide consisting of residues 107–136 (Figure 2) [16]. Moreover, the molar ratio indicated that there might be more than one binding site between the two proteins (Figure 2 and Supplementary Table S1). We then purified two more versions of PRPF4 for testing: one encoding the first 98 residues (PRPF4_1–98) and one encoding the last 64 residues (PRPF4_106–169) (Figure 2). We identified very tight binding between PRPF4_106–169 and PPIH. The value of K_d of 0.08 µM is significantly higher affinity than that reported for the PRPF4_107–136 peptide, and six times tighter than binding to PRPF4_1–169 (Figure 2). Our ITC trials indicated that PRPF4_1–98 might also interact with PPIH, but we were unable to reach high enough concentrations of PRPF4_1–98 to reach saturation (Figure 2). We therefore turned to a complementary technique, SPR.

We immobilized PPIH onto a ProteOn GLC sensor chip, using amine coupling, for use in the Bio-Rad ProteON™ XPR36 system. PRPF4_1–169, PRPF4_106–169, and PRPF4_1–198 were used as analytes and flowed over the sensor chip at five to six protein concentrations. We used nonlinear curve fitting to the Langmuir-binding isotherm to obtain K_d (see Methods for details of data analysis). PRPF4_1–169 interacted with PPIH with an affinity of 0.39 ± 0.04 µM, and PRPF4_106–169 with an affinity of 0.08 ± 0.01 µM (Figure 3). These values are quite similar to those obtained by ITC (Figure 2). PRPF4_1–98 bound to PPIH with a modest K_d value of 10.84 ± 1.37 µM (Figure 3). Protein–protein interaction analysis using ITC and SPR revealed a nanomolar affinity-binding site between residues 106–169 of PRPF4 and PPIH. A binding site within the N-terminal 98 residues of PRPF4 was also discovered.

Figure 3. SPR confirms bipartite binding between PRPF4 and PPIH.

SPR was used to validate interactions between (A) PRPF4_1–169, (B) PRPF4_106–169, and (C) PRPF4_1–98 with PPIH. (Top) Normalized sensorgrams using PPIH as a ligand and PRPF4 proteins, as indicated, as analytes. (Bottom) Data fit to the Langmuir one-site model to obtain equilibrium-binding curves and calculation of K_d. At least five concentrations were tested to obtain the RUs per concentration reported. Error bars are standard deviations of at least three separate runs.

Identification of PRPF4_1–169 as an intrinsically disordered region of PRPF4

The 30-mer peptide from PRPF4 had been previously been crystallized in complex with PPIH (Figure 1) [16]. We tried to crystallize the high-affinity complexes formed between PRPF4_106–169 and PRPF4_1–169 with PPIH as well. After extensive crystal trials using a variety of complex concentrations and buffer conditions, all crystals obtained in our trials contained density only for PPIH. Both ESpritz [33] and IUPred [34] servers indicated that the entire PRPF4_1–169 region tends toward disorder, which could explain our lack of success in crystallization (Figure 4). To experimentally validate this hypothesis, we used CD to identify α, β, and coil propensity of our PRPF4 proteins. PRPF4_1–169, PRPF4_106–169, and PRPF4_1–98 were all found to be predominantly coil as assigned by the CAPITO server [35]. PRPF4_1–169, PRPF4_106–169, and PRPF4_1–98 spectra have a classically ‘irregular’ shape, as evidenced by negative ellipticity values in the far UV. There is also a slight propensity for α-helix, especially in PRPF4_106–169, as shown by the slight deflection at 222 nm (Figure 4). When compared with a database containing CD spectra of folded and disordered proteins, PRPF4_1–169 and PRPF4_1–98 have ellipticity values at 200 and 222 nm that place them in a region where molten globule proteins cluster (Figure 4) [29,30].

Figure 4. The PPIH-interacting regions of PRPF4 are intrinsically disordered.

(A) Prediction of disorder using the IUPRED server [34]. Residues that fall below the dotted line tend toward order, as exemplified by the trace for PPIH. PRPF4_1–169 has a high tendency toward disorder for residues 1–20 and 150–169, and also tends generally toward disorder throughout. (B) CD spectra of PRPF4_1–169, PRPF4_106–169, and PRPF4_1–98 are indicative of minimal or no structured elements. PPIH, a mostly β-protein, is shown for comparison. (C) Comparison of molar ellipticity at 222 and 200 nm places PRPF4 in the ‘molten globule’ region. Reference CD spectra are from refs [29,30]. PRPF4_106–169 is not included in this analysis due to accuracy issues in measuring the protein concentration of the sample used in analysis.

Additional validation of the disordered status of all three PRPF4 proteins was provided by AUC. PPIH, PRPF4_1–169, PRPF4_106–169, and PRPF4_1–98 were first characterized using SE-AUC. Since SE-AUC measures the size of the molecule in solution independent of shape, it provides a way to unambiguously define PRPF4 and PPIH as monomers or higher-order oligomers in solution. SE-AUC confirmed that PPIH, PRPF4_1–169, PRPF4_106–169, and PRPF4_1–98 were all monomers in solution (Figure 5 and Supplementary Table S3). After establishing that all proteins were monomeric, we examined the shape of our proteins in solution using SV-AUC. SV-AUC provides a measure of the extended nature of a protein in solution using a value called the frictional ratio [31]. Globular proteins are expected to exhibit frictional ratios of 1.2–1.5; numbers higher than this indicate a protein that is extended in solution. PRPF4_1–169 was modeled with an apparent molecular mass of 21.5 kDa, 100.8% of expected, and the frictional coefficient refined to 2.2. PRPF4_106–169 was found in a monomer : dimer equilibrium, with apparent molecular mass of 9.6 kDa (99.8% of expected) and 22.4 kDa (234%). The modeled frictional coefficient was 2.2. PRPF4_1–98 was found to have an apparent molecular mass of 18.6 kDa, 140% of the expected molecular mass, and with a modeled frictional ratio of 2.3 (Figure 5 and Supplementary Table S3). We believe that the dimer population seen here for the PRPF4_106–169 and the larger than expected molecular mass of the PRPF4_1–98 are largely due to the high concentrations of protein needed to observe signal in SV-AUC, because we do not see any evidence of multimerization in solution using other methods. Additionally, the large values for the frictional ratio are indicative of highly extended shape in solution, as would be expected from intrinsically disordered regions (IDRs), and this may skew the modeling of s-values, leading to larger than expected molecular mass.

Figure 5. The PPIH-interacting regions of PRPF4 are monomeric and extended in solution.

(A–D) SE-AUC data for PRPF4 proteins and for PPIH. All four proteins are modeled as monomers in solution. (E–H) SV-AUC for PRPF4 proteins and for PPIH. PRPF4_106–169 (F) shows an apparent monomer : dimer equilibrium under the relatively high concentrations of protein used in this experiment. All other proteins are monomeric. The frictional ratio for all PRPF4 proteins are higher than 2, indicating highly extended conformations in solution.

PRPF4 does not undergo a disorder-to-order transition upon binding PPIH

The 30-mer peptide underwent a disorder-to-order transition upon crystallization with PPIH [16]. Since our three constructs of PRPF4 exhibit a strong tendency toward disorder, we asked whether any, or all, of these regions undergo a disorder-to-order transition upon interaction with PPIH. We first examined PRPF4–PPIH complexes using CD. We found that the addition of PPIH to PRPF4 results in spectra that are practically identical with that of PPIH alone, suggesting that the addition of PRPF4 to PPIH does not result in appreciable gain of α or β structure (Supplementary Figure S1). Next, we verified using SE-AUC that complexes between PPIH and PRPF4_1–169 and PRPF4_1–98 have 1 : 1 stoichiometry. However, we find that there are two molecules of PRPF4_106–169 in complex with each PPIH. This is likely connected to soluble multimerization of PRPF4_{106– 169} at high micromolar concentrations (0.5 mM), as indicated by the SV-AUC data of PRPF4_106–169 alone (Figure 6 and Supplementary Table S4). For all future work with PRPF4_106–169, we only used lower micromolar concentrations, where we have never seen any evidence of PRPF4_106–169 multimerization. We asked whether formation of complex would change the extended shape of PRPF4 in solution, measured by the frictional ratio. We found for the PRPF4_106–169 complex, the only one where we could obtain both good signal and fully saturated complex, that there was no change in the frictional ratio compared with PRPF4_106–169 alone (Supplementary Figure S2 and Table S4). Finally, we performed nuclear magnetic resonance (NMR) experiments on PRPF4_1–169 alone and in complex with PPIH. ¹⁵N-¹H HSQC spectroscopy exhibited a lack of dispersion along the proton axis, and only a limited number of peaks were visible in the spectra — hallmarks of an unfolded protein or IDR (Supplementary Figure S3). Adding equimolar amount of unlabeled PPIH to the experiment resulted in nearly identical spectra, indicating that PRPF4_1–169 does not undergo a significant disorder-to-order event when bound to PPIH.

Figure 6. PRPF4 complexes with PPIH largely have 1 : 1 stoichiometry.

(A–C) SE-AUC data for PRPF4 complexes. PRPF4_106–169 (B) complex is modeled as 2 : 1 stoichiometry, in concordance with the data shown in Figure 5.

Mutagenesis combined with SPR confirms the interaction surfaces within the PRPF4–PPIH complex, and NMR indicates no isomerization of PRPF4_1–98

The data presented above indicate a novel, two-site binding mode for interaction between PRPF4 and PPIH. We next sought to rationally design mutants that might interfere with either or both of these binding sites. Based on the co-complex structure between the 30-mer peptide with PPIH, only one residue was a candidate for point mutation of the high-affinity site: F122, which we changed to alanine (Figure 1). When PRPF4_1–169 F122A was used as an analyte, there was more than a 10-fold reduction in binding affinity to PPIH (15.6 compared with 0.38 µM). The binding of PRPF4_106–169 F122A to wild-type PPIH was also weakened, by ~7-fold (0.63 compared with 0.08 µM). These results confirm the contribution of F122 to high-affinity interaction, although it is not enough to disrupt either complex entirely (Figure 7). The lack of structural information for the PRPF4_1–98 region limited our ability to predict effective point mutations to disrupt the low-affinity site. However, if PRPF4_1–169 remains largely unstructured and extended upon binding PPIH, the contour length would be ~400 Å. If the 30-mer structure approximates the location of the tight binding site on PPIH, then the distance from the most N-terminal residue (E107) to the isomerase active site is ~200 Å, making this a possible secondary binding site. Therefore, we tested a mutation in the isomerase active site of PPIH that lowers binding affinity for proline and abrogates proline isomerization (W133A) [19,36]. We immobilized PPIH W133A and tested wild-type PRPF4_1–169, PRPF4_106–169, and PRPF4_1–98. All three PRPF4 analytes bound to PPIH W133A without loss of affinity (Supplementary Figure S4). We then immobilized PPIH W133A and tested for binding to F122A mutants of PRPF4_1–169 and PRPF4_106–169. PRPF4_106–169 F122A interacted with PPIH with 1 µM affinity. We cannot detect any binding between PPIH W133A and PRPF4_1–169 F122A (Figure 7). These results indicate that the W133A mutation in PPIH is perturbing an interaction with the PRPF4_1–98 region, and F122A weakens the interaction with the PRPF4_106–169 region, but total disruption of the complex is not achieved until both mutations are used.

Figure 7. Mutational analysis of the PRPF4–PPIH complex.

SPR was used to interrogate interactions between mutant PRPF4 (F122A) and either wild-type or mutant PPIH. (Top) Normalized sensorgrams using PPIH as a ligand and PRPF4 as analytes. (Bottom) Fitting sensorgram data to the Langmuir one-site model in order to obtain equilibrium-binding curves and calculation of K_d. In all cases, at least five concentrations of PRPF4 proteins were tested in at least three runs to obtain the RUs per concentration reported in the bottom panels. (A) PRPF4_106–169 F122A interacts with wild-type PPIH. (B) PRPF4_106–169 F122A interacts with PPIH mutant W133A. (A) PRPF4_1–169 F122A interacts with wild-type PPIH. (B) PRPF4_1–169 F122A does not interact with PPIH mutant W133A.

These findings led us to ask two more questions about the PPIH–PRPF4 interaction. First, the identification of W133A as a binding determinant for PRPF4_1–98 led us to ask whether this fragment of PRPF4 might be a substrate for proline isomerization by PPIH. There are three potential prolines that might be substrates (P17, P23, and P28). We used an NMR-based assay to check for proline isomerization. In this experiment, called ¹⁵N-¹H ZZ-exchange spectroscopy, the cis and trans resonances of the residue preceding a proline are monitored; upon the addition of enzyme, additional exchange peaks appear in the spectra due to catalysis and the mixing of magnetization during relaxation [37–40]. As seen in Supplementary Figure S5, we do not see any exchange peaks indicating active isomerization under the standard experimental conditions tested. It is possible that either isomerization is too slow to observe exchange, or that the inherent cis/trans isomer population in PRPF4_1–98 is biased so much toward the trans isomer that we cannot see significant exchange between the two forms. This might be expected for prolines contained within a fully unstructured polypeptide structure. For now, we do not currently see isomerization under conditions that have worked for other active cyclophilins and their verified substrates [37–40]. Finally, we tested for specificity between PPIH and PRPF4 fragments by immobilizing PPIA and testing for the ability of PRPF4 to interact with PPIA when compared with binding to PPIH. PPIA is the canonical member of the cyclophilin family and is highly similar to PPIH in amino acid sequence (54% identity and 70% similarity across the 170 residues of the cyclophilin domain). Additionally, all nine of the residues that comprise the isomerase active site are identical between PPIH and PPIA [18]. We used PRPF4_1–169, PRPF4_106–169, and PRPF4_106–169 as analytes to test all binding sites for specificity. As seen in Figure 8, we find that all three PRPF4 fragments are exquisitely specific for PPIH, with no observable interaction with PPIA seen.

Figure 8. Specificity of the PPIH–PRPF4 interaction.

SPR was used to interrogate interactions between PPIH and the canonical cyclophilin PPIA for each of the PRPF4 fragments. (Left) Normalized sensorgrams using PPIH as a ligand and PRPF4 fragments as analytes. (Right) Normalized sensorgrams using PPIA as a ligand and PRPF4 fragments as analytes. (A) PPIH binds to PRPF4_1–169, while PPIA binding is undetectable; (B) PPIH binds to PRPF4_106–169, while PPIA binding is undetectable; (C) PPIH binds to PRPF4_1–98, while PPIA binding is undetectable. Similar ranges of analyte concentrations were used for PRPF4_1–169, PRPF4_1–98, and PRPF4_106–169 as in Figures 3 and 7.

Formation of the PPIH–PRPF4 complex significantly increases thermal stability

We have described a tight binding site between PPIH and PRPF4_106–169, and a weak binding site in PRPF4_1–98. We wondered how these interactions were contributing to the overall stability of the complex. We used the Thermafluor technique, utilizing SYPRO Orange binding to exposed hydrophobic regions upon heating, to assess the effect of PRPF4 binding on PPIH stability [41]. We tested PRPF4_1–169, PRPF4_106–169, PRPF4_1–98, and PPIH proteins alone. In this assay, we found no significant melting transition for PRPF4_1–169, PRPF4_106–169, or PRPF4_1–98 (Figure 9). PPIH exhibited a cooperative transition with an average melting temperature of 53.8°C. We then performed a series of titrations with PRPF4_1–169, PRPF4_106–169, or PRPF4_1–98, using protein concentrations to give 80–99% saturation (described in Methods). We found that binding to PRPF4_1–169 resulted in an upward shift of almost 12°, to an average of 65.5°C. This highly stabilizing effect was predominantly due to PRPF4_106–169, which induced stabilization up to an average of 63.8°C. Binding to PRPF4_1–98 did not impart any additional thermal stability to PPIH (Figure 9). We verified these findings using CD, and obtained practically identical results, with a thermal stability shift of 6°C (Figure 9). We conclude that binding to PRPF4_1–169, and especially to the high-affinity-binding site in PRPF4_106–169, creates a highly stabilized PPIH–PRPF4 complex.

Figure 9. PRPF4 interaction with PPIH increases thermal stability of the complex.

(A and B) Thermal stability assays using SYPRO Orange, which interacts with hydrophobic residues. PPIH protein alone exhibits a cooperative unfolding transition at 54°C, while PRPF4 proteins show no transition (traces not shown). Complexes of either PRPF4_1–169 or PRPF4_106–169 with PPIH result in a highly stabilized complex with T_m of 66 and 64°C, respectively. (C and D) CD performed under thermal denaturating conditions. In both panels, buffer-subtracted spectra for PPIH and the PRPF4 protein are shown, along with the corresponding complex. Helicity as approximated by molar residue ellipticity at 222 nm is monitored throughout. The effect of interaction with PRPF4_1–169 or PRPF4_106–169 is to stabilize PPIH by 6°C. No stabilization effect was seen for PRPF4_1–98 complexes. Note, in (D), a small amount of helical signal in PRPF4_106–169 protein, but not in PRPF4_1–169, that is destabilized by increasing temperature.

Discussion

PRPF4_1–169 binds to PPIH using two unique, conserved interaction sites

We have discovered that the interaction between PPIH and the N-terminal region of PRPF4 is more complicated than previously reported. Our work identifies a nanomolar affinity-binding site contained within PRPF4_106–169. We have also identified a micromolar affinity site between PPIH and PRPF4_1–98. Conservation of both these sites in PRPF4 genes indicates that both are relevant to function. The C-terminal WD-40 domain of PRPF4 is found in all organisms that retain the ability to splice mRNA, including S. cerevisiae. However, the yeast PRPF4 gene does not include the N-terminal 200 residues currently under study. S. cerevisiae also does not encode for a tri-snRNP-associated ortholog of PPIH. We therefore performed sequence alignment on selected eukaryotes that encode both PRPF4 and a paralog of PPIH. This group included animals, fungi, ameba, green algae, plants, heterokonts, and an alveolate (Supplementary Figure S6) [42,43]. The region from residues 90–190 (human PRPF4 numbering) is broadly conserved throughout eukaryotes (Supplementary Figure S6). The first 100 residues of PRPF4 are less conserved but are present in insects, vertebrates, and at least one annelid. This level of conservation suggests that there is a biological role for the first 100 residues of PRPF4.

The strong effect of mutating F122 indicates that this residue in PRPF4 is critical for high-affinity interaction with PPIH. However, we could not find a point mutation in PPIH with an equally strong effect, although since PRPF4_106–169 is specific for PPIH there must be some selective determinant (Figure 8). One report found that binding between full-length in vitro translated PRPF4 and PPIH, mapped using loop swapping, depended on a region proximal to the one visualized in the 30-mer crystal structure [44]. It is tempting to speculate that the longer polypeptide encoded by PRPF4_106–169 might interact with this region of PPIH (α1–β3 region), explaining the increase in affinity for PPIH we see experimentally. This would be unlikely if the α-helical conformation of the 30-mer crystal was accurate, as the two short helical sections fold back upon one another (Figure 1), pointing away from the α1–β3 region. Our data would also support this report, since the α1–β3 region is not conserved between PPIH and PPIA. Future work should address this question by further mutational analysis of PPIH to confirm the role of the α1–β3 region in PRPF4 interaction.

An active site mutation in PPIH, W133, also alters binding to PRPF4. This is likely due to interaction between a proline residue in PRPF4 and the active site of PPIH [19,36]. There are five prolines in PRPF4_1–169, three of which are found in the first 30 residues. Although we cannot formally exclude the prolines at positions 118 and 126, they are present in the PRPF4_106–169 fragment, which binds with nearly equal affinity to wild-type and W133A forms of PPIH. We, therefore, believe that it is likely the prolines within the first 30 residues of PRPF4 that represent part of the weak-affinity-binding site for PPIH. This is the first direct interaction found between a spliceosomal protein and the active site of a spliceosomal cyclophilin, although we note that there are multiple examples of isomerase-dependent interactions with other nuclear proteins [45–49]. PRPF4_1–98 is specific for PPIH over PPIA, indicating that there are interactions outside the isomerase active site that are important for this interaction (Figure 8). On the other hand, we see no evidence that catalyzed isomerization is occurring in the PRPF4_1–98 fragment (Supplementary Figure S5). Therefore, our model focuses on the binding of prolines, but not catalysis.

PRPF4_1–169 is an IDR of PRPF4 and does not undergo a disorder-to-order transition upon binding PPIH

The 30-mer peptide co-crystallized with PPIH and adopted an α-helical conformation while interacting with a surface on PPIH opposite to the active isomerase site [16]. Although we identified a much higher affinity-binding site within PRPF4_106–169, this extended region does not adopt local conformation in solution, nor upon binding to PPIH. In fact, the entire N-terminal region of PRPF4 is intrinsically disordered and does not order upon binding PPIH. Perhaps, this explains why recent tri-snRNP structures show no ordered density for the first 200 residues of PRPF4 [50].

Model and relevance of the PPIH–PRPF4_1–169 interaction

Taken altogether, we arrive at a model for the PRPF4–PPIH complex in which an IDR of PRPF4 encodes two interactions with PPIH: a low-affinity site localized in the first 100 residues and a high-affinity site localized in residues 100–200. Why would both these interactions exist and be conserved across eukaryotes? We believe that the two sites serve unique functions in the context of the spliceosomal machinery. First, disease-causing mutations in PRPF genes disrupt protein–protein interactions within the tri-snRNP. Weakening these interactions leads to aberrant splicing and defects in spliceosomal assembly, and then disease [11–14]. Our previous work indicates that PPIH also is needed for proper functioning of human spliceosomal complexes, and a recent siRNA study found that PPIH knockdown results in a host of regulated splicing changes [27,28]. Therefore, we believe that the high-affinity site in PRPF4_106–169 anchors PPIH into tri-snRNP during assembly and integration into B complex, assuring proper spliceosome function. The anchoring site should be outside of the isomerase active site, as this would prevent transient protein–protein interactions, which are likely mediated by PPIH, from dissociating PPIH out of spliceosomal complexes. Whatever function PPIH plays within the tri-snRNP and B complex, this tethering interaction is very highly conserved in eukaryotes, indicating its importance (Supplementary Figure S6).

The conservation pattern of the N-terminal 100 residues of PRPF4 indicates that it also serves an important biological role, but the larger question is why this region would bind so weakly and yet specifically to PPIH. The addition of a second binding site which lessens the overall affinity of PRPF4 for PPIH implies some sort of auto-inhibition model. Perhaps, in the resting state, PRPF4_1–98 is weakly bound to PPIH through interactions with the isomerase active site. This would prevent the nonspecific binding of other proline residues by PPIH. This is supported by the finding that the prolines within PRPF4_1–98 are not being turned over by PPIH, even though the overall affinity of PRPF4_1–98 for PPIH is similar, if not higher, than that seen for most cyclophilin substrates [51–53]. Binding without catalysis might prevent release of product, which might block the autoinhibitory effect (Supplementary Figure S5). However, there are likely other residues within PRPF4_1–98 that participate in binding, as the W133 mutation of PPIH does not abrogate PRPF4_1–98 binding, while substitution of PPIA which is identical in the proline-binding pocket does. IDRs with repetitive sequences often make a large number of low-affinity interactions with their binding partners, which is a way to maintain affinity through binding to multiple, interchangeable residues in a short region of sequence [54,55]. These IDRs, which are often highly extended, unstructured, and have repetitive polar/charged polypeptide sequences, are considered good targets for post-translational modification (PTM) [56]. PRPF4_1–169 is highly enriched in arginine, lysine, serine, threonine, and tyrosine residues (34% of all residues). NetPhos predicts high-confidence sites for serine/threonine and tyrosine phosphorylation throughout the entire region [57]. By comparison, PhosphoSite Plus, a database of experimentally observed PTMs, indicates five acetylation and SUMOylation sites, and three sites of phosphorylation, in PRPF4_1–98 [58,59]. However, only two acetylation sites and one phosphorylation site has been observed in PRPF4_106–169, although the number of target residues is roughly the same. One explanation for this observation is that there is relatively limited accessibility of PTM sites in the high-affinity region while bound to PPIH and the tri-snRNP. The role of the poor-affinity site could then allow PRPF4_1–98 to be regulated by PTMs, which would then cause it to release partially or completely from PPIH in order to bind some other regulatory element or component of the spliceosome. Multiple PTMs might even be necessary to overcome the promiscuity of binding, but regardless PPIH and PRPF4 would remain tethered to the spliceosome through the high-affinity site. This release of PRPF4_1–98 from PPIH would then allow other prolinecontaining substrates to interact with PPIH (Figure 10). Verification of this model will require the development of assays in cells to measure the effect of mutations and PTMs on splicing function.

Figure 10. Model for the role of bipartite binding in the PRPF4–PPIH complex.

See the section ‘Discussion’ for details.

Abbreviations

AUC

analytical ultracentrifugation analysis

β-ME

beta-mercaptoethanol

CD

circular dichroism

HSQC

¹⁵N-¹H heteronuclear single quantum correlation

IDR

intrinsically disordered region

IMAC

Immobilized metal ion chromatography

IPTG

isopropyl β-d-1-thiogalactopyranoside

ITC

isothermal calorimetry

mRNA

messenger RNA

NMR

nuclear magnetic resonance

PPIH

peptidyl-prolyl isomerase H

PRPF3

pre-mRNA processing factor 3

PRPF4

pre-mRNA processing factor 4

PTM

post-translational modification

RNA

ribonucleic acid

RUs

response units

SE

sedimentation equilibrium

SEC

size-exclusion chromatography

SPR

surface plasmon resonance

snRNA

small nuclear RNA

snRNP

small nuclear RNA–protein complex

SV

sedimentation velocity