Design of Podospora anserina repeats

We used the 42PR HMM to search for sequence members of this family and identified a 42PR-containing ORF, Pa_6_8860, present in the genome of the fungus P. anserina. The predicted domain architecture of Pa_6_8860 consists of an N-terminal partial NB-ARC domain (van Ooijen et al., 2008), followed by a region containing modest similarity to the heptad repeats in kinesin light chains (Cyr et al., 1991). The remainder of Pa_6_8860 encodes 15 putative 42PRs (Figure 2). These 42PRs are an extreme example of high (>91%) ISI.

Open in a separate window

Figure 2

Sequence features of Podospora anserina Pa_6_8860 ORF and Pa 42PR repeat design

(A) Predicted domain organization of Pa_6_8860. The predicted 42PR motifs are labeled as grey boxes. Sequence positions in the 15 central 42PRs that differ from the derived consensus after alignment are highlighted in yellow. (B) Designed Podospora anserina 42PR sequence (Pa AB) from the alignment in (A). Capping sequences N_AB and AC_B were created by substitution of non-polar residues for polar residues (red) to promote solubility. RS substitutions (blue) resulted from cloning of repeat arrays. The consensus 34PR sequence (Main et al., 2003), c34PR, is aligned as in Figure 1. A and B-helix boundaries, and special characters to indicate hydrogen bonding interactions are labeled as described in Figure 1. (C) Single-helix representation of Pa 42PR constructs used in this study, where x signifies the number of internal Pa AB units, ranging from one to four. A and B-helices are shown in red and magenta, respectively. Together with connecting turns, these two helices make up a full Pa 42PR. N_A and C_B capping helices are shown in grey.

To characterize the structure and stability of this unique Pa 42PR, and to compare it with the more common 34PR family, we designed constructs to express arrays of a core AB repeat (Pa AB; Figure 2B and 2C). Although proteins constructed solely from Pa AB units expressed well, they were insoluble and could not be characterized. Therefore, polar substitutions to putative solvent exposed hydrophobic residues were designed from helix wheel representations and made to the N-terminal A and C-terminal B-helices, to create capping repeats N_AB and AC_B (Figure 2B). Substitutions on surface-exposed sites are expected to minimize structural and energetic perturbations. This approach has been successful in promoting solubility in other repeat protein studies (Main et al. 2003; Wetzel et al. 2008; Aksel et al., 2011).

We used these capping repeats, along with unmodified AB units, to create constructs of the type N_AB(AB)_xAC_B, where the subscript x represents the number of AB core repeats, ranging from one to four (Figure 2C). We were able to express and purify these capped constructs with reasonable yields for biophysical characterization.

Solution structure of Pa 42PRs

To determine the hydrodynamic properties of our Pa 42PR constructs, we conducted analytical ultracentrifugation sedimentation velocity (AUC-SV) experiments. We modeled SV data using direct boundary (ΔC/ΔT) methods, as well as c(s) methods (Schuck 2000). The SV data indicate that all constructs populate predominantly (>97%) monomeric species at low (<10μM) concentrations (Figure 3). Higher concentrations result in weak associations; the extent and nature of association differs for each construct (Table 1).

Open in a separate window

Figure 3

Sedimentation velocity analytical ultracentrifugation of Pa 42PR N_AB(AB)_xAC_B constructs

Continuous c(s) distributions and global ΔC/ΔT fits of N_AB(AB)AC_B (panels A,B), N_AB(AB)₂AC_B (panels C,D), N_AB(AB)₃AC_B (panels E,F), and N_AB(AB)₄AC_B (panels G,H), respectively. Lower panels are residuals between ΔC/ΔT values and fitted models. For clarity, only a subset of ΔC/ΔT values and curves are shown. See Table 1 for a summary of fitted models.

N_AB(AB)AC_B and N_AB(AB)₂AC_B SV c(s) distributions are consistent with predominantly monomeric species with higher concentrations displaying small peaks at higher s-values (Figures 3A and 3C). These higher s-value peaks are shifted from what would be expected for dimeric species. Sedimentation velocity ΔC/ΔT curves spanning a wide concentration range were globally fit using a monomer-incompetent trimer model (Figures 3B and 3D). This model assumes a small proportion of the loading concentration is present as a trimeric species that does not equilibrate with the monomer (Figure S1A). Incompetent species have been identified in a number of other AUC studies (Lemaire et al., 2005; Wowor et al., 2011; Xu, 2004). Although the fitted s- values of the incompetent species are consistent with molecular weights corresponding to trimers of each protein, given the low fitted concentrations of these species (less than 3% of the total loading concentration), it is possible they represent small amounts of impurities.

N_AB(AB)₃AC_B and N_AB(AB)₄AC_B SV c(s) distributions are consistent with predominantly monomeric species with higher concentrations displaying small peaks at s-values consistent with dimeric species (Figures 3E and 3G). Sedimentation velocity ΔC/ΔT curves were globally fit to a monomer-dimer, incompetent dimer model (Figures 3F and 3H). This model assumes a rapid and reversible equilibrium between monomer and dimer, and an additional dimeric species that does not equilibrate with the monomer (Figure S1B). A summary of hydrodynamic models and parameters used for each construct is shown in Table 1.

Based on sequence elements that match 34PR motifs, we expect the 42PRs derived from Pa to adopt an α-helical structure. Consistent with this expectation, far-UV CD spectra of all N_AB(AB)_xAC_B constructs show a high level of α-helical character with well-defined minima at 222 and 208nm (Figure 4A). CD spectra for Pa 42PR constructs suggest a higher level of α-helical character than c34PR constructs (Main et al. 2003). Observed variations in molar residue ellipticity values are likely due to uncertainties in protein concentrations. Each repeat contains only one tyrosine, and no tryptophans (Figure 2); thus, extinction coefficients are low.

Open in a separate window

Figure 4

Far-UV CD spectroscopy and global Ising modeling of equilibrium unfolding of 42PR and 34PR constructs

(A) Far-UV CD spectra of Pa 42PR constructs N_AB(AB)AC_B (black), N_AB(AB)₂AC_B (purple), N_AB(AB)₃AC_B (blue), and N_AB(AB)₄AC_B (cyan). (B) Normalized equilibrium unfolding transitions of Pa 42PRs and c34PRs. Closed circles show Pa 42PR constructs, and are colored as in (A). Open circles show c34PR constructs: B(AB)S (grey), B(AB)₂S (black), B(AB)₃S (purple), and B(AB)₄S (blue). With this color scheme, constructs with the same integer number of repeats have the same color. Solid and dashed lines result from globally fitting one-dimensional Ising models to Pa 42PRs and c34PRs, respectively. c34PR samples contain 50 mM sodium phosphate, 150 mM NaCl, and are at pH 6.8. Pa 42PR samples contain 25 mM Tris-HCl, 350 mM NaCl, and are at pH 8.0. All samples are at 25°C. See Figure S2 and Table S1 for two-state analysis of the same data.

Thermodynamic stability of Pa 42PRs

To measure the thermodynamic stability of designed Pa 42PRs, we carried out urea-induced equilibrium unfolding of constructs ranging in length from three to six total repeats (Figure 4B, closed circles), in 350 mM NaCl, pH 8. Under these conditions, all transitions are completely reversible and have fully resolved native baselines. As repeats are added, the transition midpoints increase and become sharper, indicating both increased stability and a high level of cooperativity. These observations are reflected in apparent free energies and m-values for unfolding obtained from a two-state fit (Table S1).

To compare stabilities and cooperativities of Pa 42PRs to shorter 34PR motifs, we also measured the urea-induced equilibrium unfolding of c34PRs of equivalent numbers of total repeats (Figure 4B, open circles), at 150 mM NaCl, pH 6.8. To generate c34PRs with integral numbers of whole repeats, we added a single c34PR B-helix to the N-termini of c34PR constructs studied by Regan and coworkers (Main et al., 2003; Kajander et al., 2005; Cortajarena and Regan, 2011). We term these constructs B(AB)_xS, where x signifies the number of central c34PRs ranging from one to four, and S is the “solvation helix” designed by Regan and coworkers. The addition of the N-terminal B helix slightly increases the stability of each construct, and is in agreement with the intrinsic and interfacial helical coupling energies measured by Regan and coworkers (Kajander et al. 2005).

Although Pa 42PRs are larger, they have significantly lower urea midpoints, sharper transitions, and increased m-values than their c34PR counterparts (Figures 4, S2, and Table S1). In contrast to the m-values of c34PRs, which plateau at four repeats, Pa 42PR m-values continue to increase through six repeats. This reflects a higher level of cooperativity in Pa 42PRs than in c34PRs. The magnitudes of the fitted Pa 42PR two-state m-values correlate well with their larger motif size and the expected solvent accessible surface area (SASA) changes for unfolding.

One-dimensional Ising analysis

To obtain a mechanistic understanding of the apparent increase in cooperativity in Pa 42PRs compared to c34PRs, we analyzed unfolding transitions using a nearest-neighbor 1D Ising model (Aksel and Barrick 2009; Kajander et al. 2005; Mello and Barrick 2004; Aksel, Majumdar, and Barrick 2011). We globally fit all Pa 42PR unfolding transitions to an Ising model (Figure 4B, closed circles and solid lines) and compared them to a separate global fit of c34PR unfolding transitions (Figure 4B, open circles and dashed lines). The global parameters obtained from these fits are shown in Table 2. For both nPR series, cooperativity arises from unfavorable intrinsic repeat folding (ΔG_i), and from favorable interfacial coupling between adjacent folded repeats (ΔG_i,i+1). The cooperativity enhancement observed for Pa 42PR transitions results from an increase in magnitude of both ΔG_i and ΔG_i,i+1 compared to c34PR transitions, i.e. lower intrinsic stability and higher interfacial stability.

Sequence features of the 42PR motif

We identified a new class of nPR sequences, which we term 42PRs. The repeating unit is 42 residues and shares sequence characteristics with canonical 34PRs (Figure 1). The main differences between the two sequence classes appear to be N and C-terminal extensions of the 34PR-defined A and B-helices, respectively, in 42PRs. The 42PR HMM shows a high degree of conservation near the N terminus. This conservation may reflect the sequence characteristics defining a longer A-helix. The C-terminus of the 42PR HMM shows less conservation, indicating the rules defining the extension of the B-helix may be less strict.

Conserved positions in both 42PRs and 34PRs include small and hydrophobic residues at helix interfaces and in turn regions (Figure 1). Based on these structurally restrictive environments, it is likely that the conserved nPR residues are responsible for defining the fold. These residues may act as staples, around which helical extensions can be accommodated. In this fashion, the structural registry and packing of conserved nPR residues between helices is maintained.

Structural features of nPR motifs

Due to the repetitive architecture of nPR proteins, their structures can be defined by a small number of repeating parameters: helix crossing angles, distances, and contacts. These parameters, and the tertiary structures they define, are important for function and for stability. Helix crossing angles determine the extent to which nPR arrays form a concave binding surface for target peptides and proteins (Cortajarena and Regan, 2006; Cortajarena et al., 2010; Zhu et al., 2012). The number and type of helical contacts (A_i:B_i, B_i:A_i+1, and A_i:A_i+1) likely contribute to cooperativity in folding. By comparing the structural and energetic features of consensus, naturally occurring, and highly repetitive 42- and 34PRs, we can determine which structural parameters are general to nPRs, to 42- versus 34PRs, and which are modulated by local sequence variation within families.

Aside from differences in helix lengths, many of the structural features of 42PRs are similar to those of 34PRs. Although there are small deviations from repeat to repeat, average crossing angles for the A_iB_i, B_i:A_i+1, and A_i:A_i+1 helices are similar (Table 4). Likewise, the helical distances and solvent accessible surface area (SASA) burial between helices are similar across nPR families (Table 4).

In contrast, there are several notable differences in contacts within and between 42- and 34PRs (Figure 6). In addition to contacts within helices (main diagonals), nPR structures show a characteristic contact pattern, consisting mainly of contacts between successive helices (A_i:B_i and B_i:A_i+1; anti-diagonal features). Other contacts include those between successive A-helices (A_i:A_i+1; off-diagonal features), but are less frequent in 42PR structures than in 34PR structures. These A_i:A_i+1 contacts connect the regularly spaced, anti-diagonal contacts (Figure 6).

Open in a separate window

Figure 6

Contact maps of 42 and 34 residue nPRs

Contacts are defined as atom pairs from different residues that are within 2.2–4.0 Å. Points above the main diagonal represent backbone contacts. Points below the main diagonal represent backbone-side chain or side chain-side chain contacts. Protein structures are displayed in each column (cyan, 42PRs; yellow, 34PRs). Row I displays helical contacts over the entire structure. Rows II, III, and IV expand over the indicated residue range. Rows I and II display contacts within A-helices (red), B-helices (magenta), between A_i and B_i helices (blue), and between B_i and A_i+1 helices (grey). Contacts between A_i and A_i+1 helices appear as red, off-diagonal points in rows I and II. Rows III and IV show all hydrophobic and polar contacts, respectively, and are color coded with respect to distance.

The packing features describing A_i:A_i+1 helices can be visualized by alignment of A_iB_iA_i+1 units from each motif with respect to A_iB_i, and are summarized in Table 4. Although the helix geometries within nPR families are similar, 42PRs have fewer A_i:A_i+1 contacts than 34PRs. For 3CEQ and 1ELW (naturally occurring 42- and 34PRs, respectively), the differences in A_i:A_i+1 contacts can be explained by local helix geometry (longer A_i:A_i+1 distances in 3CEQ). Local helix geometry in Pa 42PRs also affects the number of A_i:A_i+1 contacts, as the N_A-helix kink results in more A_i:A_i+1 contacts in the N-terminal region of the contact plot, relative to the internal repeat region (Figure 6, row I). Despite these local effects, Pa 42PRs and c34PRs have overall similar A_i:A_i+1 helix distances. This suggests that sequence specific information also influences the extent of A_i:A_i+1 interaction.

To further analyze the types of interactions in the A_i:B_i, B_i:A_i+1, and A_i:A_i+1 interfaces, we sorted polar and nonpolar interactions into separate contact maps (Figure 6, rows III and IV). Sorted contact plots reveal that packing within all interfaces, both within (A_i:B_i) and between (B_i:A_i+1) repeats, is predominantly hydrophobic, although these hydrophobic contacts (Figures 5C, 5D, and Figure 6, row III) occur at greater distances than the polar contacts (Figure 6, row IV). Polar contacts are also present within the B_i:A_i+1 interfaces of Pa 42PRs, including a conserved Tyr OηH--⁻OεC Glu hydrogen bond on the convex side of the Pa 42PR superhelix, between adjacent 42PR motifs (Figures 1B, ​,5E).5E). Other repetitive polar interactions include a His-Ser-Gln hydrogen bond network connecting successive A_iB_iA_i+1 helices (Figure S5) and a His-Ser A-helix N-terminal capping motif (Figure S6).

Interestingly, the 42PRs of the human kinesin light chain (3CEQ) contain a region with B_i:B_i+1 helical contacts (Figure 6, Row I), which are not seen in 34PRs. This results from a slight kink in one B-helix, allowing for enhanced hydrophobic packing, along with other contacts between polar residues on the convex face of the superhelix. These sequence and length-specific structural variations highlight the structural malleability of the nPR motif. In naturally occurring repeat proteins, sequence variation can locally tune structural features. A consensus design approach applied to 42PRs would be expected to reveal representative interactions across all 42PRs.

Folding of Pa 42PRs

The cooperative folding of the Pa 42PR arrays in this study is striking. As repeats are added, both stabilities and m-values increase. This phenomenon is characteristic of other linear repeat proteins, especially ankyrin repeats, where energetic coupling leads to highly cooperative folding (Aksel et al., 2011; Wetzel et al., 2008). In contrast, the m-values of c34PRs plateau at four repeats, consistent with a high level of partially folded states (Kajander et al., 2005; Cortajarena and Regan 2011), and decreased cooperativity compared to ankyrin repeats and the Pa 42PRs presented here.

Global fits of a 1D Ising (nearest-neighbor) model to Pa 42PR and c34PRs provide a quantitative description of cooperativity in these two repeat systems (Figure 4B). We find the cooperativity enhancement in Pa 42PRs to result from a decrease in the intrinsic stability (an increase in ΔG_i) and an increase in the interfacial stability (a decrease in ΔG_i,i+1; Table 2). Stability studies under the same conditions will reveal the extent to which these differences depend on differences in solution conditions, although preliminary results suggest Pa 42PRs m-values to be insensitive to salt concentration (JDM and DB, data not shown). The decreased stability of individual Pa 42PRs is surprising, as each Pa 42PR helix is longer, and has more hydrogen bonds. Moreover, the 42PR helices are predicted to be more stable, based on Agadir (2.4 and 8.28%; Muñoz and Serrano, 1994), than the helices of c34PR (0.4 and 1.74%). The best-fit intrinsic m-values (m_i) for both repeat types are consistent with an increased level of denaturant sensitive helical structure in Pa 42PRs.

Circular dichroism spectroscopy

CD measurements were conducted using an Aviv Model 400 CD Spectropolarimeter (Lakewood, NJ). CD samples contained 25 mM Tris-HCl, 350 mM NaCl, pH 8 (Pa 42PRs) or 50 mM Na Phosphate, 150 mM NaCl, pH 6.8 (c34PRs). Far-UV CD spectra were recorded at 25°C using a 0.1 cm path-length quartz cuvette (Starna Cells Inc., Atascadero, CA) at protein concentrations ranging from 15–25 μM. Spectra were obtained by signal averaging every 1 nm for 30 s. Buffer spectra were subtracted prior to analysis.

Urea-induced equilibrium unfolding

Unfolding transitions were obtained by monitoring CD at 222 nm (Pa 42PRs) and 220 nm (c34PRs) in a 1 cm path-length quartz cuvette. High purity urea (Amresco, Solon, OH) was deionized by stirring with mixed-bed resin (Bio-Rad, Hercules, CA) as previously described in (Street et al., 2008). Urea concentration was determined by refractometry (Pace 1986). Titrations were performed at protein concentrations ranging from 1.0–2.5 μM using a computer-controlled Microlab titrator (Hamilton, Reno, NV). At each urea concentration, protein samples were equilibrated for 5–7 min at 25° C, and signal averaged for 30 s. Two-state analysis was performed previously described (Street et al., 2008), and shown in Figure S2.

Global analysis of equilibrium unfolding transitions of Pa 42PRs and c34PRs using 1D Ising models was performed by constructing partition functions from two-by-two transfer matrices as previously described (Aksel and Barrick 2009). These matrices contain intrinsic and interfacial free energy terms (ΔG_i and ΔG_i,i+1). A single ΔG_i parameter is used for the three internal and capping intrinsic energies of the energies of Pa 42PR (N_AB, AB, and AC_B). Likewise, another ΔG_i parameter is used for the two internal and capping intrinsic energies of the energies of c34PR (BA and BS). The use of single ΔG_i terms for capping and internal repeats is justified by preliminary results from cap deletion studies (JDM & DB, unpublished), which show similar stability decrements for capping and internal repeats.

Denaturant sensitivities (m_i=-dΔG_i/d[Urea]) were ascribed to instrinsic stability through a linear relationship. Functions describing the fraction of folded repeats were generated from the partition functions, and used with baseline parameters to model urea-induced equilibrium unfolding curves. Fitting was performed using a python program written by J.D.M., importing lmfit (Newville et al., 2014) to minimize a global objective function using non-linear least-squares.

The Ising model used here has a significantly lower parameter to construct ratio than analysis using two-state models (0.75 thermodynamic parameters per construct versus 2.0 for two-state analysis; compare Tables 2 and S2). This is accompanied by increased degrees of freedom (ν=249, Pa 42PRs; ν=306, c34PRs, compared to ~20 for two-state analysis of individual constructs) in fitting. Therefore, error analysis and interpretation of the fitted parameters are more robust.

Analytical Ultracentrifugation

Analytical ultracentrifugation sedimentation velocity (AUC-SV) experiments were performed using a Beckman XL-I analytical ultracentrifuge. Prior to AUC experiments, all proteins were extensively dialyzed into CD buffer. Protein concentrations ranged from 5–100 μM.

AUC-SV cells were assembled using SedVel60K 1.2 mm meniscus-matching centerpieces (SpinAnalytical) and sapphire windows. All other cell components were purchased from Beckman Coulter. Upon sample and reference (dialysate) loading, centerpieces were aligned in a An-60Ti rotor, and menisci were matched according to (Allgood and Barrick, 2011). After remixing, the rotor was thermally equilibrated under vacuum at 25° C for at least 90 minutes. Experiment s were run for approximately 8 hours at 45–50 krpm.

Protein Crystallization and Data Collection

Crystals of native N_AB(AB)₃AC_B were grown at room temperature (~22° C) by hanging -drop vapor diffusion. Protein solution (20 mg/mL) was mixed in either a 2:1 or 1:1 ratio with reservoir solution containing 0.1 M MES (pH 6.5) and 25–30% PEG 4K. Crystals appeared after approximately 3–7 days. Crystals were cryoprotected by transfer into a solution consisting of 0.1 M MES (pH 6.5), 35% PEG 4K, and 5–10% ethylene glycol, and then flash frozen in liquid nitrogen. The selinium-methionine-substituted Q17M variant gave rise to morphologically similar crystals under these conditions, with the addition of 5 mM TCEP.

Native and selenium derivative data sets were collected at the National Synchrotron Light Source (NSLS) beamlines X-25 and X-29 (Brookhaven National Laboratory, Brookhaven, NY)) and processed with HKL2000 (Otwinowski and Minor, 1997). Crystals belong to space group P2₁2₁2 and contain one molecule per asymmetric unit.

We would like to thank Drs Annie Héroux and Howard Robinson of the Macromolecular Crystallography Research Resource (PXRR) at the National Synchrotron Light Source, Brookhaven, NY for their technical assistance with data collection strategy and preliminary analysis. We also thank Dr. Phil Mortimer of the JHU mass spectrometry facility and Dr. Michael Love of the JHMI X-ray facility. This work was supported by NIH grant R01 GM068462 to D.B., J.D.M. was supported by NIH training grant T32-{"type":"entrez-nucleotide","attrs":{"text":"GM008403","term_id":"218383076","term_text":"GM008403"}}GM008403, J.M.K. was supported by NIH grant R01 GM099231 to Daniel J. Leahy, and G.D.B. was supported by NIH grant R01 GM084192.