Distinct conformational and energetic features define the specific recognition of (di)aromatic peptide motifs by PEX14

Thermodynamic parameters of the PEX14-PEX5 peptide interactions

We first examined the interaction and thermodynamics of the recognition of the eight (di)aromatic binding motifs in the PEX5 NTD with the PEX14 NTD using isothermal titration calorimetry (ITC). For this we titrated 15-mer peptides, comprising the seven Wxxx(F/Y) motifs flanked by five residues, as well as one construct comprising PEX5 residues 1–113, which harbors the W0 (LVxEF) peptide (Figure 1B). The binding processes represent a single transition with exothermic binding enthalpies with dissociation constants ranging from 60 nM to 6 µM (Table 1, Supplementary Figure 2). The relative differences are comparable to previously reported fluorescence polarization studies (Saidowsky et al. 2001), some minor differences variations may reflect distinct buffer conditions and temperatures used. The significant variations in the binding affinities for the different (di)aromatic peptide motifs in the PEX5 NTD suggest that – in addition to the two conserved aromatic residues – further amino acids contribute to the interaction. Interestingly, ITC experiments with yeast PEX14 NTD and PEX5 NTD show only binding to a reverse WxxxF (W3) motif with µM affinity (Supplementary Figure 3), consistent with previous reports (Kerssen et al. 2006), suggesting further variations in the binding interface.

The energetics of binding are notably different for each motif (Figure 2A). Under all measurement conditions, binding between the peptides and PEX14 NTD is exclusively enthalpy-driven (ΔH < 0) with an unfavorable entropic contribution (−TΔS > 0). The enthalpy changes for the binding of W1, W3, W5 and W6 motifs are larger (about −19 kcal/mol) compared to W0, W7 peptides (−11 kcal/mol). Moderate enthalpy values are observed for the W2 and W4 peptides (−15 kcal/mol). Noteworthy, binding of W0, which lacks the first aromatic residue, and W7 show the lowest entropic penalty.

Figure 2:

Thermodynamic and conformational features of the PEX14 NTD peptide interaction. (A) Bar Graph comparing thermodynamic parameters for the PEX14 NTD interaction with different (di)aromatic peptide ligands. Color codes are ΔG in blue, ΔH in green and –TΔS in red. (B) Correlation between free binding energy (−ΔG), calculated from the measured ITC data at 298 K (blue boxes) and thermal stability for the PEX14 NTD/peptide interactions are shown. Thermal unfolding was measured using CD at 222 nm, the transition midpoints (T _m ) are shown as red squares. (C) Correlation between free binding energy (−ΔG) and average ¹³C NMR secondary chemical shifts for the peptide motifs. The extent of positive NMR secondary chemical shift indicates increasing helical propensity. The correlation line is shown in red: y = 1.37x + 8.14 with R² = 0.42. (D) CD spectra (mean residue ellipticities) of different regions and the complete PEX5 NTD. A strong negative band below 200 nm indicates unstructured regions. The three 100-residue regions show different minor extent of helical conformations. The CD spectrum of the PEX5 NTD (1–315) shows some negative minima at 208 and 222 nm, consistent with the presence of partial helical folding.

We next characterized the thermal stability of apo PEX14 NTD and in complex with the W0–W7 ligands using circular dichroism (CD) measurements (Figure 2B). Temperature-dependent measurements allow the determination of the transition midpoint (T _m ) for thermal denaturation of the free PEX14 NTD and when bound to the W0–W7 ligands. The T _m values correspond to the temperature at which 50% of protein is unfolded and are directly correlated to the stability of protein complex (Rees and Robertson 2001; Tol et al. 2013). The T _m is 61 °C for the apo PEX14 NTD and ranges from 65 °C to 77 °C for the PEX14 NTD/ligand complexes in the presence of two-fold excess of peptide ligands (Figure 2B, Supplementary Figure 4). As expected, ligand binding did not cause any unfolding transition but stabilized the fold. Notably, the highest T _m value of 77 °C is observed for the complex with the W5 peptide, which also has the highest binding affinity (∆G = −9.86 kcal/mol), while the lowest stability (T _m  = 64 °C) was observed for W4, which features the lowest binding affinity (∆G = −7.10 kcal/mol). Overall, the melting temperatures and thermodynamic stabilities of the peptide complexes show a clear correlation with the free energies calculated from the ITC data (Figure 2B).

To assess conformational features of the PEX5 (di)aromatic motifs we used solution NMR and CD spectroscopy. Our CD spectra of the full PEX5 NTD (1–315) and the regions comprising residues 1–113, 110–230, 228–315 indicate a mostly disordered region with some α-helical propensity (Figure 2D). These results are in agreement with our previous reported NMR analysis, where we identified defined α-helical propensities within the PEX5 NTD by analysis of ¹³C secondary chemical shift (Supplementary Figure 5) (Gaussmann et al. 2021). Since these experiments did not cover the flanking sequence of W7 very well, we analyzed ¹³C secondary shifts of a larger construct for this study (Supplementary Figure 5G). Notably, in the context of the full PEX5 NTD W5 motif exhibits the largest extent of α-helical conformation (Table 2). There is clear correlation of the free energy of association (−ΔG) determined by ITC and helical propensity of the (di)aromatic motifs from average ¹³C secondary chemical shift values (Figure 2C). The weaker correlation observed for W6 and W7 may reflect additional contributions by neighboring residues outside the core motif. The good correlation between helical propensity observed by NMR and thermodynamics of binding determined by ITC also matches the stabilization effect of peptide binding to the PEX14 NTD indicated by thermal stability measurements (T _m ) by CD.

Computational analysis of the PEX14 NTD/PEX5 W0–W7 peptide interaction

To understand the contribution of structural features for the PEX14 NTD/PEX5 (di)aromatic peptide interactions, we performed 50 ns molecular dynamics simulations of 20-mer peptides comprising the W0 to W7 motifs bound to the PEX14 NTD. The simulations are based on the experimental structure of the PEX14-W1 peptide complex. For the simulation of other peptides, side chains were replaced by the corresponding residue with IRECS (Hartmann et al. 2007; Hartmann et al. 2009). The eight peptides bind to PEX14 NTD with different affinities, ranging from ∆G values of −9.89 kcal/mol to −7.10 kcal/mol determined by ITC (Table 1). To assess the importance of specific PEX14 NTD interaction pattern contributing to these values, we focused on the analysis of the enthalpic contribution ∆H (Table 1), as these values can directly be correlated to specific interactions and interaction energies. We then calculated the interaction energies for the five residues of the Wxxx(F/Y) core motif plus the following additional residue of the peptides based on optimized structures extracted from the last 10ns of MD simulations using the Dynadock program (Antes 2010). For computational analysis of binding energies, only the five residues of the core motif plus the following residue were considered (Figure 1B). Flanking residues are included to consider structural features obtained from the molecular dynamics simulation. We hence denote residue numbers of the core peptide (p) motifs with the number A1(p)-A6(p) and the flanking residues with a number of the position relative to the core motif (… −2(p), −1(p), A1(p)–A6(p), A+1(p), …), i.e. the W and F residues in the W1 peptide (WAQEFL) correspond to W1(p) and F5(p), respectively (Figure 1B). Residues within PEX14 are annotated according to the human protein sequence.

Based on the binding enthalpies, ΔH, the eight peptides can be classified into three groups: W1, W3, W5, and W6 have strong, W2 and W4 intermediate, and W0 and W7 weak enthalpic contributions. Notably, the experimental binding enthalpies (ΔH) and calculated interaction energies correlate very well (Table 2). The W0 motif, which lacks the first aromatic residue shows small experimental binding enthalpy and calculated interaction energy. The W6 peptide exhibits the most favorable experimental enthalpy and consistently shows the best calculated interaction energy. Peptides with intermediate experimental enthalpy values have interaction energies around −150 kcal/mol and the calculated energies for the weakest binder W7 was determined to −100 kcal/mol. These results show that our computational simulation provides realistic peptide-PEX14 structures, which can be used for an in depth analysis of the binding modes.

From a comprehensive analysis of the PEX14-peptide interaction (Tables 3, 4; Supplementary Table 1) three key features can be identified that are important for binding to the PEX14 NTD (Figure 3, top left): (i) a hydrophobic core interface, (ii) electrostatic interactions with K56 and (iii) electrostatic interactions with R40 and N38, as illustrated for the PEX14-W1 complex in Figure 3.

1. The overall binding pattern is characterized by a central hydrophobic core region consisting of the bottom of the peptide binding groove in PEX14 and the corresponding counterpart residues in the bound peptide. Note, that the K56 and R40 contribute additional hydrophobic interactions via their bulky aliphatic side chains contacting W1(p) and L6(p)/F5(p), respectively. Thus, the hydrophobic area in PEX14 NTD stretches over the whole binding site and forms complementary hydrophobic pockets for the corresponding peptide residues (Figure 3i). If this stable hydrophobic interaction pattern exists, a strong intra-peptide backbone hydrogen bond between the flanking residue −2(p) and residue 3(p) of the core motif can be observed (Figure 3iii), which stabilizes the helical conformation of the peptide.

2. The hydrophobic interface is flanked on either side by the two charged PEX14 residues R40 and K56, respectively. These residues contribute not only hydrophobic interactions but mediate strong electrostatic contacts involving their positively charged side chains with negatively charged side chains in the PEX14 NTD, where intra-ligand interactions can further stabilize these interactions. The hydrophobic/hydrophilic interaction pattern of K56 is conserved in all peptides containing the W1(p) and E4(p) residues and thus presents a second key feature (Figure 3, ii).

3. The third important binding feature involves R40, which establishes hydrophilic interactions with E−2(p) and N38 through its terminal guanidino group (Figure 3iii), featuring a similar but not that as highly conserved interaction pattern as seen for K56. Overall, the two flanking hydrophilic interactions additionally stabilize the central hydrophobic region such that the peptide is effectively locked in its bound position, which may explain the large enthalpy contributions.

Figure 3:

Key interactions in the PEX14 NTD/PEX5 W1-peptide complex. The PEX14 NTD is shown in yellow/green, the PEX5 peptide in pink. Important residues are given in licorice representation. The central picture shows the most prominent structure of W1 during the last 10ns of MD simulation. Salt bridges and hydrogen bonds are indicated by colored dashed lines (orange: Sidechain contacts, blue: backbone contacts). A schematic sketch of the binding pattern of the PEX14 NTD/PEX5 complex is shown on the top left, illustrating the three binding features (i), (ii) and (iii). (i) Surface representation of the residues contributing to the hydrophobic core. (ii) Hydrophilic interactions of K56. (iii) Hydrophilic interactions of R40.

Detailed analyses of the binding modes of the individual PEX14-peptide complexes are given in Tables 3, 4 and Supplementary Table 1. In addition, representative structures from the last 10 ns of simulation (see Materials and Methods) are shown in Figure 4 and Supplementary Figure 6. We find that the strong binding peptides, i.e. W1, W3, W5 and W6 exhibit all three interaction features (Table 3): (i) a stable central hydrophobic core, strong hydrophobic and hydrophilic interactions between the peptide and residues (ii) K56 and (iii) R40, respectively, leading to a strong intra-ligand −2(p)/3(p) backbone hydrogen bond (Table 4) and thus a very stable helical conformation of the bound peptide. For the peptides W3 and W6 lacking the glutamate at position −2(p), R40 stabilizes the binding to the PEX14 NTD by interacting with residues located more C-terminal (Supplementary Figure 6).

Figure 4:

Molecular dynamics simulations. Most prominent structures observed during the last 10 ns of the MD simulations of PEX14 NTD bound to wildtype or mutant PEX5 peptides are shown. The PEX14 NTD and PEX5 peptides are colored in green and pink, respectively. The three key features hydrophobic core, electrostatic interactions with K56 and interactions with R40/N38 are shown in boxes (i), (ii) and (iii) respectively. Hydrophobic interactions are indicated by green ϕ, charged and polar interactions with an orange “+” sign. Interaction features that are missing are indicated by a purple “−“ sign. Salt bridges and hydrogen bonds are indicated by dashed lines. Symbols above the respective peptide sequences indicate existing or missing features of the individual residues. Residues mutated compared to wildtype motifs are underlined (A) W2, (B) W4, (C) W0, (D) W7, (E) W1_E14(p)A, (F) W2_S2(p)A, (G) W4_Y2(p)L, (H) W7_D4(p)E.

Analysis of the complexes with intermediate and weak binding peptides (Table 3 and Figure 4) reveals that some of the features observed for the strongly binding peptides are missing. In the W2 and W4 peptide complexes the central hydrophobic core is disrupted by residue 2(p), either S or Y, substituting for the hydrophobic residues A or L. This residue is located right at the center of the hydrophobic core (Tables 3, 4; Figure 4A, B) and thus crucial for optimal packing. Residues in the flanking regions can also affect the binding affinity. The threonine residue (T) at −2(p) in W2 abolishes the favored R40 and N38 interactions seen with W1 (Figure 3D) while the H mutation at position 6(p) in W4 causes a charge clash with R40 which destabilizes the binding (Figure 4B).

The effects of these amino acid substitutions on the overall stability of the complexes are summarized in Table 4. The percentage of simulation time within the last 10 ns of the MD simulations during which either the 4(p)/K56 salt bridge or the −2(p)/3(p) backbone hydrogen bond exist, serves as a measure for the stability of the corresponding interactions and, as explained above, the stability of the helical peptide conformation in the complex. The E4(p)/K56 salt bridge exists during 90% of the simulation time (i.e. value of 0.9) for the W1 peptide, but only during 38% of the time for peptide W2. This is correlated to the presence of the 2(p)/3(p) backbone hydrogen bond, which is significantly weakened in W2 (0.58) compared to the strong binding peptides (>0.8). In addition, the peptide interactions of R40 are weak for peptide W2. This is due to strong conformational fluctuations of the guanidino-group, which in W2 can alternatively interact with the backbone carbonyl oxygen of residue 2(p). These fluctuations further perturb the already weakened central hydrophobic core (Figure 4A). The same trend holds for W4, due to the disruption of the hydrophobic core by Tyr. As this residue is also too large to fit into its binding pocket, the bulky tyrosine side chain is turned towards to solvent and clashes with the R40 side chain (Figure 4B). In addition, the side chain of R40 is stabilized by π-stacking interactions with Y2(p) and H6(p). Although this should stabilize the bound complex, it leads to weaker hydrophobic interactions due to imperfect shape complementary (i.e. a gap is introduced between peptide and protein surfaces, data not shown), as well as extra electrostatic repulsions between H6(p) and R40. Therefore, in both the W2 and W4 complexes, the central hydrophobic binding core is disrupted, which leads to an additional destabilization of the R40 peptide interactions.

The W0 motif (LVxEF) is distinct from W1-W7 in that it lacks the W1(p) tryptophan (the most conserved residue of the motif), which is replaced by a leucine. Nevertheless, W0 is one of the strongest binders in terms of ΔG but not considering ΔH (Table 1). The less favorable ΔH likely reflects that the replacement of W by L reduces the hydrophobic contact surface (Figure 4C). Yet, most of the key features required for a strong interaction as described above are present in W0. Hydrophobic interactions with R40 and K56 as well as the E4(p)/K56 salt bridge and −2(p)/3(p) H-bond exist (Table 3). In the last 10ns of MD simulation the E4(p)/K56 salt bridge and −2(p)/3(p) H-bond are present for 99 and 89% of the time respectively (Table 4). These results are comparable with W1 showing similar affinities (Table 1). When bound to the PEX14 NTD the W0 peptide undergoes conformational fluctuations as it is less well packed, consistent with the reduced helicity observed for the W0 peptide featuring only two helical turns (other ligands have four) and the lack of the bulky tryptophan side chain and incomplete electrostatic clamping (Figure 4C, iii), suggesting conformational entropy compensation. This is likely also reflected in the fast off-rate (k_off) observed in SPR experiments (Neuhaus et al. 2014).

In the W7 peptide E4(p), which normally forms the essential salt bridge to K56 is mutated to D4(p). The shorter side chain does not provide an optimal length for hydrogen bond formation with K56 and leads to strong fluctuating movements of the K56 side chain as it tries to adapt to the larger distance towards D4(p). This effect is enhanced by D6(p), as the terminal guanidino group of R40 tries to form hydrogen bonds with D6(p), which is sterically not possible, thus leading to flipping R40 conformations (Figure 4D). During the MD simulations, both movements of R40 as well as K56 cause strong fluctuations in the PEX14 backbone in the neighboring binding site region, which are not observed for the other peptides (data not shown). Through these fluctuations the central hydrophobic core is weakened, as seen by the lower percentage of occurrence of the −2(p)/3(p) H-bond in W7 (0.71) (Table 4).

In silico mutational analysis to identify sequence requirements for PEX14 binding

Based on the analysis of the simulation results we carried out in silico mutational studies for four peptides (W1, W2, W4, and W7), performing the same type of simulations as for the natural peptides for an overall of seven variants each featuring one single “strategic” mutation within the peptide. We designed two “failure” and five “rescue” mutations. The “failure” mutations are based on W1 and were designed to eliminate the important E4(p)/K56 salt bridge (and thus destabilize the interaction). For both variants W1_E4(p)A and W1_E4(p)L, the interaction energies indeed decrease considerably in both cases leading to much weaker binding (Table 2) and a partial opening of the binding pocket (compare Figure 4E and Supplementary Figure 6D with Figure 3). These results confirm the importance of the E4(p)/K56 salt bridge.

The five “rescue” mutations were designed to improve peptide-PEX14 interactions of W2, W4 and W7. Replacing S2(p) in W2 by Ala or Leu (W2_S2(p)A and W2_S2(p)L) restore the central hydrophobic core region. As expected, these mutations lead to stronger interaction energies (Table 2) as well as an increase in the percentage of occurrence of the E4(p)/K56 salt bridge. In the case of W2_S2(p)L also the −2(p)/3(p) hydrogen bond is stabilized (Table 4). The binding pattern of R40 is altered in both mutants, and R40 now forms stable interactions with the side chains of the central peptide residues, resembling the binding pattern of W1 and W5 (Figures 3iii, 4F, iii; Supplementary Figure 6B, iii). In the variant W2_S2(p)L, the aliphatic side chain of R40 contributes to the hydrophobic core (Supplementary Figure 6E, i).

Next, we mutated Y2(p) to Leu in W4 generating W4_Y2(p)L, which also leads to a stabilization of the central hydrophobic core and thus of the E4(p)/K56 salt bridge. In addition, a more stable interaction pattern of R40 with the C-terminal end of the peptide is observed, resembling the binding pattern of W6 and W3 (Figure 4G, i; Supplementary Figure 6A, i, 6C, i). The two variants of W7 replacing D4(p) by Glu (W7_D4(p)E) and D6(p) by Leu (W7_D6(p)L), respectively, show increased binding energies, a stabilizing of the E4(p)/K56 salt bridge (Table 4), and of hydrophobic interactions with R40 (Table 3). The R40/6(p) and K56-4(p) interactions for the D4(p)E variant are shown in (Figure 4H). Due to the longer side chain of the residue at position 4(p), a stable interaction pattern with K56 can be formed (compare to Figure 4D).

Analysis of additional (di)aromatic ligands in peptide overlay binding assays

To validate the general recognition features of PEX14 binding derived from the analysis above, we synthesized 80 15-mer peptides representing Wxxx(F/Y)-containing naturally occurring fragments of PEX5 proteins of various organisms. The immobilized Wxxx(F/Y)-containing peptides were analyzed by overlay incubation with purified human PEX14 NTD followed by antibody detection (Figure 5). The intensity of the staining roughly correlates with the dissociation constants as indicated for the Wxxx(F/Y) motifs W1 to W7 of human PEX5 (Figure 5A). In accordance with results obtained by ITC (Table 1), W1 and W5 gave the strongest signals, whereas W7 and W4 were not detectable.

Figure 5:

Peptide overlay binding assay with PEX14 NTD for various diaromatic peptide motifs and sequence conservation analysis of all PEX14 NTD binding motifs. (A) Peptide spot overlay assay with His-tagged PEX14 NTD and immobilized peptides representing Wxxx(F/Y) motifs of PEX5 of different species. Each peptide comprised 15 amino acids with a central Wxxx(F/Y) and five adjacent amino acids at each side. Bound PEX14 was visualized immunochemically with monoclonal anti-His₆ antibodies. Spots with reduced intensities indicate reduced binding affinity for PEX14. The number of Wxxx(F/Y) motifs varies within the sequences of PEX5 proteins, indicated by species name as abbreviation and position of Wxxx(F/Y) motif starting with the N-terminal W1. (B) Sequence logo representation of the strongest PEX14 NTD binding motifs found. The total height (in bits) of the stack indicates the degree of sequence conservation at the corresponding position and the height of each letter is proportional to its frequency at that position. The logo was generated using Berkeley’s WebLogo program (Crooks et al. 2004). Hs: Homo sapiens, Xt: Xenopus laevis, Gal: Gallus gallus, Dr: Danio rerio, Ce: Caenorhabditis elegans, Ag: Ashbya gossypii, Dm: Drosophila melanogaster, Apis: Apis apis, Tb: Trypanosoma brucei, Ld: Leishmania donovani, At: Arabidopsis thaliana, Sc: Saccharomyces cerevisiae, Fg: Fusarium graminearum, Hp: Hansenula polymorpha, Pp: Pichia pastoris, Ca:Cavia porcellus, Yl: Yarrowia lipolytica, Nc: Neurospora crassa, An: Aspergillus nidulans, Sp:Schizosaccharomyces pombe. (C) The refined core and expanded consensus motif for high affinity PEX14 binding.

Taken together, among the 80 tested peptides, we identified 33 binding peptides (Supplementary Table 2), from which 12 showed a strong interaction with human PEX14 NTD (Figure 5B, Supplementary Table 2). Interestingly, the relative number of diaromatic peptide motifs that interact with PEX14 in the peptide-overlay assay, is higher in plants and animals than in fungi and protists. For instance, 5 out of 7 motifs of human PEX5 show clear PEX14 binding, whereas all tested yeast and protists PTS1-receptors contain no more than one (di)aromatic peptide interacting with the human protein. This observation is also consistent with our ITC experiments which show binding of ScPex14 to a single motif in ScPex5 NTD with a 1:1 stoichiometry (Supplementary Figure 3) and other previous reports on Pex14-Pex5 interactions (Cyr et al. 2008; Hojjat and Jardim 2015; Watanabe et al. 2016).

Analysis of the peptide binding (Figure 5A, B) confirms that the predominant residue type at position 2(p) is alanine, while at position 6(p) mainly leucine is found, followed by alanine and isoleucine underlining the role of a hydrophobic residue. At position 4(p) almost all PEX14 binding peptides (n = 29) exhibit a glutamic acid side-chain whereas aspartic acid at position 4(p) is the most abundant amino acid among the non-binding peptides. Note, that the predominant residue outside of the core-motif in position −2(p) is glutamic acid. These data support the proposed recognition features that we identified in our analysis, i.e. the requirement of a stable central hydrophobic core region and stabilizing salt-bridges between the peptide and the residues K56 and R40 for high affinity binding to PEX14. With these data we propose a refined core consensus motif WΦxE(F/Y)Φ further expanded to ExWΦxE(F/Y)Φ for high affinity binding peptides (Figure 5C).

There are a few peptides, which do not completely fulfill the criteria identified in our analysis. For example, in Gallus gallus (Gal) strong interactions are observed for Gal1 and Gal4. The sequence of Gal1 is almost identical with the human W1 with a mutation A2(p) to T2(p), which seems not to disturb the overall stability of the binding. Although threonine is a polar residue, the terminal of the side chain is a methyl group. Our simulations reveal that the residue at position 2(p) tends to use the terminal of the side chain to interact with V41 forming the hydrophobic core (Figure 4; Supplementary Figure 6). Thus, we speculate that threonine at position 2(p) can also form the hydrophobic core and stabilize the binding compared to other polar residues, such as serine (Figure 5B). However, the analogue W4 motif Gal4 is less conserved in the flanking region. Especially, the substitution of H6(p) to Q6(p) changes charge and size specifies this position as unfavorable for large charged amino acids. In fact, none of the good binders harbors a R, K or H in the 6(p) position.

Cloning, protein expression and purification

Cloning, recombinant expression and purification of human PEX14 (16–80) (PEX14 NTD) (UniProtKB: O75381) and PEX5 (1–113), PEX5 (110–230), PEX5 (228–315) (UniProtKB: P50542) protein constructs were performed as described previously (Gaussmann et al. 2021; Neuhaus et al. 2014).

In brief, unlabeled PEX14 NTD was cloned into pETM11 vector, expressed in Escherichia coli BL21-(DE3) cells (Stratagene) in LB medium as a fusion protein containing His₆-tagged followed by a tobacco etch virus (TEV) cleavage site. After cell lysis (cell lysis buffer: 50 mM sodium phosphate, 300 mM sodium chloride, 10 mM Imidazole, pH 8.0), PEX14 NTD was purified using nickel-nitrilotriacetic acid-agarose (Qiagen) (elution buffer: 50 mM sodium phosphate, 300 mM sodium chloride, 500 mM Imidazole, pH 8.0), followed by TEV cleavage (TEV buffer: 50 mM sodium phosphate, 100 mM sodium chloride, 1 mM DTT, pH 8.0). His₆ tag was removed by a second Ni²⁺ affinity chromatography. The final purification was done by size exclusion chromatography on a HiLoad 16/60 Superdex75 column (GE Healthcare) in 20 mM ammonium bicarbonate buffer and pooled fractions are lyophilized.

Human PEX5 (1–113), PEX5 (110–230), PEX5 (228–315) protein constructs were expressed from pETM10 vectors with a non-cleavable N-terminal His6 tag. Unlabeled PEX5 fragments were expressed and purified with the same protocol used for PEX14 NTD with minor change. 8M urea was used in the cell lysis buffer to avoid unspecific proteolysis cleavage during purification. Urea was removed during Ni affinity chromatography by extensive wash with cell lysis buffer. Ni-Eluted fractions were further purified by size exclusion chromatography and lyophilized.

The larger PEX5 (281–639) construct was cloned into a pETM11 vector (EMBL) with cleavable N-terminal His6 tag using site-directed ligase-independent mutagenesis (SLIM) (Chiu 2004) with the following primers:

The ScPEX14 NTD (corresponding to amino acids 1–58) (UniProtKB: P53112) and ScPex5(239–280) (UniProtKB: P35056) constructs were subcloned into a pETM30 vector (EMBL) that encodes a His₆-GST tag followed by a TEV-cleavage site. The ScPex5(1–313) construct encodes a non-cleavable His₆-tag. Expression and purification of the yeast proteins followed the same procedure described above for the human variants, with the exception that for ScPEX14 NTD an additional wash step with 1M NaCl was included in the Ni-affinity step. Uniformly ¹⁵N,¹³C-labeled PEX5 (281–639) was expressed in deuterated M9 minimal medium supplemented with 1 g/L ¹⁵NH₄Cl (Cambridge Isotope Laboratories), 2 g/L [U-¹³C]-glucose-d12 (Cambridge Isotope Laboratories), as the sole source of nitrogen and carbon. The cell culture was induced with 0.5 mM IPTG at 18 °C for 14–16 h before harvesting. PEX5 (281–639) was purified in buffer containing 50 mM Tris pH 7.5 and 300 mM NaCl using Ni-affinity chromatography. After TEV cleavage and a reverse Ni-column, size exclusion chromatography on a HiLoad Superdex 75 16/600 column (Cytiva) was performed.

The PEX5 (1–113) protein fragment was used to represent W0 for the biophysical experiments. Synthetic 15-mer peptides (W1–W7) of human PEX5 were purchased from Peptide Specialty Laboratories (Heidelberg, Germany). Peptides with purity of ≥98% were dialyzed extensively against water before the experiment.

Isothermal titration calorimetry (ITC)

ITC measurements were performed at 25 °C using PEAQ-ITC or iTC200 microcalorimeters. All proteins and peptides used for titration were dialyzed overnight in ITC buffer consisting of 20 mM sodium phosphate pH 6.5, 100 mM sodium chloride, 0.02% (w/v) sodium azide. In individual titrations, 200 μM concentration of human PEX5 peptides were loaded into the syringe and injected in 1.5 μ L volumes at an interval of 150 s into a 280 μL cell contacting 20 μM concentration of PEX14 NTD, while stirring at 750 rpm. Calorimetric data were fitted to a single site binding model using MicroCal ITC-ORIGIN software supplied with the instrument. The binding stoichiometry (n), the dissociation constant (K _D ) and the enthalpy change (ΔH) were obtained from the fitted data. The Gibbs free energy (ΔG) and change in entropy (ΔS) were calculated from −RT lnK _D  = ΔG = ΔH−TΔS, where R is the gas constant and T is the absolute temperature (Rees and Robertson 2001). To account for heat of dilution, control experiments were performed and subtracted from the corresponding data.

NMR spectroscopy

NMR experiments for PEX5 (1–113), PEX5 (110–230), PEX5 (228–315) protein fragments were described previously (Gaussmann et al. 2021). NMR of PEX5 (281–639) was performed at 298 K on a Bruker Avance II 950 MHz spectrometer equipped with cryoprobe. Buffer was exchanged to 20 mM sodium phosphate pH 6.5, 50 mM sodium chloride and 10% D₂O using size exclusion chromatography. The protein was measured at 750µM in a 5mm Shigemi tube. Sequential assignment of backbone resonances was done by using TROSY versions of standard triple resonance experiments (Sattler M et al. 1999; Weisemann et al. 1993). NMR spectra were processed using Topspin (Bruker Biospin, Rheinstetten, Germany) or NMRPipe (Delaglio et al. 1995) and analyzed using CcpNMR Analysis 2.4.2 (Vranken et al. 2005). Secondary chemical shifts, Δδ (¹³Cα) – Δδ(¹³Cβ) were calculated by subtracting random coil chemical shifts from the observed ¹³Cα, ¹³Cβ chemical shifts (Kjaergaard and Poulsen 2011; Schwarzinger et al. 2001).

Circular dichroism

Circular dichroism (CD) spectra were recorded on a Jasco J-810 spectropolarimeter equipped with a peltier thermal controller. A final concentration of 30 µM of PEX14 NTD and 60 µM of peptide motif (1:2 ratio) were prepared in 10 mM sodium phosphate, 50 mM sodium chloride and pH 6.5. Thermal denaturation experiments were carried out by increasing the temperature from 10 to 95 °C at 1 °C/min in a cuvette with 0.1 cm path length and the CD spectra were collected at 222 nm. The protein-peptide complexes were incubated for 1 h before initiating the unfolding experiment. The midpoint of the folding and unfolding (T _m ) is derived from raw data by fitting to the sigmoidal equation, Y = A2 + (A1−A2)/(1+exp[(x−x₀)/dx)]. Where A1 and A2 are the folding and unfolding intercept respectively. x is the midpoint of the curve and dx is the slope of the curve (Greenfield, 2006). The curve was fitted using Origin. Far UV-CD date were collected at 25 °C in the wavelength range of 190–260 nm. Spectra from 10 accumulations were added and the spectrum of the buffer alone was subtracted.

Peptide overlay binding assays

Each peptide spot comprises regions of 15 amino acids of PEX5 proteins with the Wxxx(F/Y) motif as a central core motif and 5 flanking amino acids on each side. Peptides were directly synthesized on a cellulose membrane as described previously (Saidowsky et al. 2001). After blocking with 3% BSA in TBS (10 mM Tris/HCl pH 7.4, 150 mM NaCl), membranes were probed overnight at 4 °C with purified 10 nM His₆-tagged PEX14 NTD in TBS. Bound PEX14 NTD was immunodetected by monoclonal anti-His₆ antibodies in TBS +3% BSA, and horseradish peroxidase–coupled secondary antibodies in TBS +10% milk powder and ECL Western Blotting Detection Reagent (GE Healthcare Amersham, ECL Western Blotting Detection Reagent). Between steps, the membranes were first thoroughly washed with TBS-TT (20 mM Tris/HCl pH 7.5, 0.5M NaCl, 0.05% (v/v) Tween20, 0.2% (v/v) Triton X-100) and at the end with TBS only.

Molecular dynamics simulations

Molecular Dynamics calculations were performed for all peptides shown in Figure 1B bound to PEX14. As initial structures the complex of PEX14 NTD with PEX5 (PDB-ID: 2W84, peptide sequence W1, Table 2) was used. The sequence of PEX5 was mutated to the corresponding sequence for all other motifs using the IRECS (Hartmann et al. 2009; Hartmann et al. 2007) method as implemented in the DynaCell program (Antes 2010). he Amber14 force field parameter set (Duan et al. 2003) was used together with the Amber14 software packages (Case et al. 2014). The structures were prepared for minimization with the tleap utility (Schafmeister et al. 1995) and all calculations were conducted in a neutralized, rectangular TIP3P (Jorgensen et al. 1983) water box extending at least 12 Å from any protein atom at each side of the box. Energy minimizations were performed with sander or pmemd.MPI (Case et al. 2014). For every complex, two subsequent minimizations were conducted. First, 10,100 steps of restraint minimization (100 steps with the steepest descent algorithm and 10,000 steps with the conjugate gradient method) were done with the protein atoms restrained using a 50 kcal mol⁻¹ Å⁻² force constant. Second, 100,100 steps of energy minimization (100 steps with the steepest descent and 100,000 steps with the conjugate gradient method) were conducted without restraints. Both minimizations were considered as converged if the root-mean-square of the Cartesian components of the energy gradient was less than 0.0001 kcal mol⁻¹ Å⁻¹. The non-bonded interaction cutoff was set to 8.0 Å for both energy optimizations. Before conducting production runs, all systems were heated up by stepwise increasing the temperature over 660 ps while at the same time incrementally decreasing the number of restraint atoms as well as the force acting on them.

At each heating-up step, the systems initial velocities were randomly assigned from a Maxwell-Boltzmann distribution at the given target temperature. MD simulations were performed with 1-fs time steps. Non-bonded interactions were computed applying a cutoff of 14 Å. The Particle Mesh Ewald method was used to calculated long-range electrostatic interactions (Darden et al. 1993). The SHAKE algorithm (Ryckaert et al. 1977) was applied to constrain bonds to hydrogen atoms. The temperature was kept constant using the Berendsen thermostat (Berendsen et al. 1984) with a time constant of 1 ps to ensure constant temperature. The Berendsenbarostat was applied with a compressibility of 45 × 10⁻⁶ bar⁻¹ and a pressure relaxation time of 1 ps to keep a constant target pressure of 1 bar. All MD simulations were performed by the pmemd.MPI or pmemd.cuda programs from the Amber14 software package (Case et al. 2014).

After equilibration, a total of 50 ns of MD simulation was performed for each system (100 ns for W5). For the analysis of the system, all frames (4000) from the last 10 ns MD of the trajectory were minimized by DynaDock (Antes 2010) for all minimized structures. The averaged energy values were used as final interaction energy. In order to calculate the interaction energy of the variants, the corresponding residue was mutated by IRECS (Hartmann et al. 2007) based on the representative structure of the biggest structural cluster from the corresponding wildtype MD trajectory. For each variant, 50 ns of MD simulation were performed (100 ns for W2_S2L) and the last 10 ns of the trajectories were used to calculate the interaction energy. The running conditions for these MD calculations were the same as for the wildtype peptides.

The last 10 ns of the trajectories were analyzed using cpptraj in AmberTools 15 (Roe and Cheatham 2013) for hydrogen bond analysis together with an in-house Cytoscape (Shannon et al. 2003) plugin allowing a network-based representation of the obtained hydrogen bonds. The hydrophobic cluster analysis was performed visually using VMD (Humphrey et al. 1996).