Main

The stereochemistry of amino acids, and therefore of proteins, is biological canon. The chirality of the Cα atom means that the mirror images (enantiomers) of amino acids cannot be superimposed—that is, they have a ‘handedness’. Amino acids in nature are predominantly ‘left-handed’ or levorotatory (l), whereas their enantiomers are ‘right-handed’ or dextrorotatory (d) (Fig. 1a)—so named because of how they affect circularly polarized light3. Thus, l- and d-amino acids and proteins composed of l- and d-amino acids (l- and d-proteins, respectively) are mirror images of each other (Fig. 1b). The preference for l-proteins is so strong that we may generally say that l-proteins make up the molecular structure and machinery of Nature. However, d-amino acids do exist, and Nature typically exploits these in signalling, as free amino acids, or in defence systems, as parts of short peptides, or in peptidoglycans (for example, in the bacterial cell wall4,5), as neurotransmitters6, toxins and venoms7, and as antibiotics8 (reviewed in ref. 9).

Fig. 1: Chirality in protein–protein interactions.
figure 1

a,b, l- and d-amino acids (a) and l- and d-proteins are mirror images of each other, illustrated in b by ubiquitin (Protein Data Bank (PDB): 4GSW). c, The five protein pairs constituting our model system covering a continuum of disordered (ProTα:H1)22,24 to ordered (MCL1:PUMA)25 protein complexes, and three intermediate interactions of RST with ANAC046, DREB2A28 and ANAC013. d, l-protein pairs will interact, but the features that might allow l-proteins and d-proteins to interact are unclear.

Full size image

Proteins are key to the activity of biological systems; they function through interactions with one or several binding partners. It is widely accepted that the d-enantiomer of a protein would be unable to bind a partner l-protein. However, in a pharmaceutical context, it would be desirable to overcome this lack of binding, owing to the metabolic stability of d-peptides in biological systems, where they are not recognized by natural metabolic processes10. Thus, peptides based on d-amino acids have been explored as constituents of peptide drugs and synthetic d-proteins have been used as scaffolds for screening natural or de novo designed l-peptides in mirror-image phage displays11,12. In the ‘retro-inverso’ strategy, d-amino acid-based peptides mimic the l-peptide enantiomer when the d-amino acid sequence is produced in reverse13. This strategy relies on the d-peptide forming the same secondary structure as the l-peptide, enabling interaction with its l-protein binding partners. Examples can be found in the treatment of diabetes14,15, breast cancer16 and inflammation17.

The past 25 years have uncovered the functional relevance of intrinsically disordered proteins and protein regions (referred to collectively here as IDPs) existing in dynamic ensembles of interconverting conformations1. Although structural disorder can persist in complexes and has important functional roles there, it remains unclear whether IDPs are confined to chiral constraints18,19,20. The continuum of complexes formed by IDPs ranges from folded, induced-fit interactions to fuzzy, or fully disordered complexes with structural heterogeneity21,22. Not much is known about the atomic structure of heterogeneous complexes, and less is known about the structure of fully disordered complexes where the ligand at the extreme can be comparably dynamic in the free and the bound states22,23. This raises the fundamental question of whether these complexes are truly independent of 3D topology and thus independent of the chiral constraints of folded complexes, or whether there are configurational constraints, perhaps too subtle to be resolved experimentally.

To answer this question, we selected an assortment of interacting protein pairs in which the ligand is disordered in its unbound state, and either stays disordered in the complex or adopts different degrees of structure upon binding (Fig. 1c). Peptide ligands were synthesized using d-amino acids and their binding to folded partners was compared to that of their l-peptide enantiomers using a range of biophysical and structural methodologies (Fig. 1d). We found that sensitivity to chirality in binding correlates with retained disorder in the complex, with fully disordered protein complexes forming regardless of the ‘handedness’ of the ligand.

To test whether disordered protein interactions could persist regardless of chirality, we initially focused on the interaction of prothymosin-α (ProTα) with histone H1.0 (H1), which has been shown to be a high-affinity, disordered interaction22,24 (Fig. 1c). We used a 21-residue peptide from the C-terminal tail of H1, H1155–175 (Fig. 2a), which contains a high charge density with a fraction of charged residues of 0.52 (Extended Data Table 1), and procured l- and d-enantiomers (l-H1155–175 and d-H1155–175). Far-UV circular dichroism (CD) confirmed that the two peptides were mirror images of each other (Fig. 2b), and nuclear magnetic resonance (NMR) spectroscopy analyses showed identical chemical shifts (Extended Data Fig. 1a). The CD spectra also showed that the peptides were disordered, as expected. We next used NMR to measure the chemical shift perturbations (CSPs) of ProTα caused by each enantiomer upon their addition. In this case, we found that l-H1155–175 and d-H1155–175 produced similar CSPs in ProTα (Fig. 2c), which we quantified by calculating the difference between the CSPs induced by l- and d-enantiomers (ΔCSPl-d) at equimolar concentrations of each enantiomer of the H1155–175 peptide (Fig. 2d). We probed the affinity (Kd) and thermodynamic properties using isothermal titration calorimetry (ITC), finding the same values for the enantiomers in terms of Kd (Fig. 2e and Extended Data Table 2). We observed that the changes in binding enthalpy (−TΔS) and entropy (ΔH) were similar for l-H1155–175 and d-H1155–175 (Extended Data Table 2), and that the Kd value was in the low micromolar range for both enantiomers. ProTα and full-length H1 interact with nanomolar to picomolar affinity at near-physiological ionic strength22,24, but we observed micromolar affinity with the peptides, mostly because of the lower total charge of the H1155–175 fragment (ProTα: −43 (or −44 depending on isoform24); l/d-H1155–175: + 11; full-length H1: + 53). We also obtained binding affinities using single-molecule Förster resonance energy transfer (smFRET) spectroscopy, labelling ProTα with donor and acceptor fluorophores (Fig. 2f). The agreement between the affinities obtained by ITC and smFRET using very different ProTα concentrations suggests that the complex is predominantly of 1:1 stoichiometry (Fig. 2). Furthermore, analogous to the NMR CSPs, the smFRET data showed that the changes in transfer efficiencies on binding are very similar for l-H1155–175 and d-H1155–175, indicating that the conformational ensembles of ProTα bound to l-H1155–175 or d-H1155–175 are highly similar, again highlighting that there is no significant difference between the interactions of ProTα with the l- or d-enantiomers of H1155–175.

Fig. 2: Effects of chirality on protein interactions.
figure 2

a, ProTα and the H1155–175 peptide remain disordered during their interaction. b, Far-UV CD spectra of l-H1155–175 and d-H1155–175 peptides and their sum. MRE, mean residue ellipticity (cm2 dmol−1 residue−1; n = 1). c,d, CSP (c) and ΔCSPl-d (n = 1) (d) caused by the addition of l-H1 or d-H1 (500 µM) to ProTα (50 µM). e, ITC of interactions of ProTα with l-H1155–175 and d-H1155–175. Raw ITC thermograms (left) and fitted one-site binding isotherms (right). DP, differential power. f, smFRET of l-H1155–175 or d-H1155–175 with ProTα, fitting FRET efficiency (⟨E⟩) as a function of ligand concentration to obtain Kd values. g, MCL1 is folded and interacts with the disordered PUMA peptide, which forms an α-helix upon interaction via induced fit. h, Far-UV CD spectra of l-PUMA and d-PUMA peptides, and their sum. i, ITC performed under the same conditions for both l-PUMA and d-PUMA (n = 3, figure is representative). N/A, not applicable. j, ITC performed using higher concentrations of both MCL1 and d-PUMA (70 µM and 700 µM, respectively; n = 3, figure is representative). k, CSPs induced by the interaction of MCL1 (50 µM) with l-PUMA and d-PUMA at equal (90%) saturation (n = 1). l, Changes in NMR peak intensities upon addition of l-PUMA or d-PUMA to MCL1 at 90% saturation (n = 1). In all panels, l-peptides are represented in grey, and d-peptides are in orange. Blue diamonds indicate missing assignments, assigned residues that could not be tracked, or prolines.

Source Data

Full size image

As an example, representing the other end of the disorder–order continuum, we probed the interaction between induced myeloid leukaemia cell differentiation protein (MCL1) and p53 upregulated modulator of apoptosis (PUMA, also known as Bcl-2-binding component 3) l- and d-peptides (Extended Data Table 1). This nanomolar-affinity complex has previously been characterized as a folding-upon-binding induced-fit interaction that leads to the folding of disordered PUMA130–156 into a stable α-helix within the complex25 (Figs. 1c and  2g). We therefore postulated that the d-enantiomer of PUMA would be unlikely to bind MCL1. As PUMA tends to form homodimers in solution, we used the strictly monomeric M144I variant, which is a fully functional binding partner of MCL126. Far-UV CD and NMR analyses indicated that the PUMA peptide was disordered, with l- and d- enantiomers producing mirror-image CD spectra and identical chemical shifts (Fig. 2h and Extended Data Fig. 1). We first performed ITC using the same concentrations for both l-PUMA and d-PUMA, finding that l-PUMA bound with nanomolar affinity, whereas d-PUMA appeared not to bind (Fig. 2i and Extended Data Table 2). By increasing the concentrations of both MCL1 and d-PUMA considerably, we were able to observe a Kd value in the high micromolar to low millimolar range (Fig. 2j). All thermodynamic properties (Extended Data Table 2) were very different for l-PUMA and d-PUMA, which further suggests that there is a fundamental difference between l-PUMA and d-PUMA in their ability to interact with MCL1. We then compared the NMR CSPs induced at equally saturated levels (90% bound) of either l-PUMA or d-PUMA to MCL1. Only l-PUMA caused substantial CSPs for MCL1 (Fig. 2k). Since the interaction between MCL1 and l-PUMA was in slow exchange on the NMR timescale, the relative contributions of the bound and unbound species could be assessed from relative peak intensities. Reduced intensity of peaks representing the free state of MCL1 was observed in the presence of d-PUMA, suggesting some compatibility but a failure to achieve the folded bound state, as also supported by the very small chemical shift changes in MCL1 (Fig. 2l). This might indicate that d-PUMA can interact weakly with MCL1 but cannot undergo the folding required for an induced-fit interaction.

Overall, using l- and d-peptides and comparing interactions at the extremes of the disorder–order continuum confirms the need for homochirality in structurally ordered complexes. However, a fully disordered heterochiral complex can form, even with binding characteristics that are indistinguishable from the natural counterpart.

Having probed the extremes of the disorder–order continuum, the next step was to understand how intermediate systems respond to chirality. As intermediate systems, we used the RCD1–SRO–TAF4 (RST) domain from RCD1, interacting with various transcription factors that form different degrees of structure in their RST-bound states27 (Fig. 1c). We characterized interactions of RST with l- and d-peptides derived from the transcription factors ANAC046 (ANAC046319–338; Fig. 3a), DREB2A (DREB2A255–272; Fig. 3b) and ANAC013 (ANAC013254–274; Fig. 3c and Extended Data Table 1). Previous far-UV CD analyses suggested an induced helical structure for the DREB2A in complex with RST; an effect that was less pronounced for the two other ligands27. Whereas an NMR-based HADDOCK model exists for the RST–DREB2A complex28 (Fig. 3b), the structures formed by ANAC046 and ANAC013 in their complexes with RST were unknown. Thus, we generated a structural prediction for these interactions using AlphaFold329,30 (Fig. 3a,c and Extended Data Fig. 2).

Fig. 3: RST interactions with l- and d-peptides vary depending on remaining disorder in the complex.
figure 3

a–c, The peptides of ANAC046 (a), DREB2A (b) and ANAC013 (c) are disordered in their free state, and form varying structures upon binding to RST (top right). Far-UV CD spectra show the l- and d-enantiomers of each peptide as mirror images (sum of l- and d-peptide spectra is shown in blue). d–f, ITC for binding of 10–100 µM RST in the cell with 100–1,000 µM of ANAC046 (d), DREB2A (e) or ANAC013 (f) peptide in the syringe (n = 3, figure is representative). g–i, NMR CSPs of the interactions of RST with of ANAC046 (g), DREB2A (h) or ANAC013 (i) peptide showing the differences between the l- and d-enantiomers (>99% saturation, n = 1). Blue diamonds, indicate missing assignments, untraceable assigned residues or prolines.

Source Data

Full size image

We first confirmed using far-UV CD and NMR that the l- and d-peptides were disordered and enantiomeric (Fig. 3a–c and Extended Data Fig. 1). To determine the thermodynamics and the affinity of the interactions of l- or d-peptides with RST, we used ITC (Fig. 3d−f and Extended Data Table 2). We observed very different thermodynamic profiles for the three l-peptides, with more favourable enthalpy (\(\Delta {H}^{^\circ }\)) for DREB2A and ANAC013 than for ANAC046, whereas the opposite was the case for the entropy (\(-T\Delta {S}^{^\circ }\)), suggesting that the complexes had different characteristics. Comparing the effect of stereochemistry, we observed larger differences in Kd values as the interactions probably became more structured—that is, the difference between l-ANAC046 and d-ANAC046 was 15-fold (Fig. 3d), the difference between l-DREB2A and d-DREB2A was 72-fold (Fig. 3e) and the difference between l-ANAC013 and d-ANAC013 500-fold (Fig. 3f). The trend suggested that the amount of structure required for binding reduced the propensity of the d-enantiomer to interact with RST. This interpretation was further supported by comparing ΔCSPl-d of RST induced by ANAC046 (Fig. 3g), DREB2A (Fig. 3h) and ANAC013 (Fig. 3i). The ΔCSPl-d values of RST were substantial upon addition of ANAC013 and minimal upon addition of ANAC046. The same trend was observed in the pattern of peak intensity changes of RST (Extended Data Fig. 3), suggesting that ANAC046 is relatively disordered in complex with RST, whereas ANAC013 is more structured, and therefore less likely to interact with RST as a d-enantiomer. The ΔCSPl-d values of RST for binding of DREB2A were between those for binding of ANAC046 and ANAC013, consistent with the differences in binding affinity for RST and effect of the d-enantiomer. Finally, we extracted Kd values and the dissociation rate constant (koff) after fitting NMR titration data to a two-state model using NMR 2D lineshape analysis31,32 (Extended Data Fig. 4), finding only minor effects of stereochemistry on the transition state energies (−2 to 1 kJ mol−1 (ΔΔGunbound-‡,d-l)) (Extended Data Fig. 5 and Extended Data Table 3). This observation highlights that the major effect of stereochemistry occurs after the transition state, in agreement with previous observations that native contacts consolidate late in the binding reaction pathway of IDPs33,34,35.

In summary, it is apparent that the stereosensitivity of protein–protein interactions is not an all-or-none phenomenon, but instead occurs on a continuum. The results presented here indicate that sensitivity to chirality reflects the degree of disorder in the native complex.

To understand the sensitivity to chirality better, we investigated the RST complexes in more detail. We obtained insight into the RST-bound structures from Cα chemical shifts determined from chemical exchange saturation transfer (CEST) NMR for l-ANAC046 (Extended Data Fig. 6) and ZZ-exchange NMR for l-ANAC013 (Extended Data Fig. 7; 50% saturation), comparing them to similar available data for l-DREB2A28. A distinct pattern in structure emerges from the secondary chemical shifts (SCSs) (Fig. 4a). In their bound state, all three peptides have a similar core region comprising a short strand followed by a turn. Comparing the SCSs to the AlphaFold3-predicted structures of the complexes shows remarkable agreement between the NMR data and the predicted structures (Fig. 4a,b) unanimously suggesting the existence of a common structural core in all three bound RST partners. Aligning the peptides based on the core structure displaces the conserved Phe/Tyr residue of the sequence-based RST-binding motif27 to align with an acidic and hydrophobic residue, respectively, suggesting that features beyond specific residue types are important (Fig. 4c). Indeed, all core structures are highly acidic and hydrophobic with a large and similar salt dependence in RST binding (Fig. 4d and data in ref. 36), further supporting the idea that binding of the core structure to RST is dependent on geometrically relaxed hydrophobic, ionic and π–cation interactions.

Fig. 4: Ability to bind d-enantiomers scales with disorder of the complex.
figure 4

a, 13Cα SCS of l-enantiomers of ANAC046, DREB2A and ANAC013 in complex with RST. Four or more consecutive positive values indicate helical structure and three or more consecutive negative values indicate β-strand or extended structure. Grey regions highlight core structure. Key RST-interacting residues, shown as sticks in b, are highlighted in bold. b, Left, AlphaFold3-predicted structures of the RST complexes. Right, expanded view highlighting the binding sites, with interacting residues as sticks. c, Structure-based sequence alignment (top) and 15NH transverse relaxation rate constants (bottom) of free (light) and bound (dark) states of the RST-binding partners (error bars are standard errors of the fit). Asterisks indicate chemical exchange. The dashed line shows the outlier-excluded average R2 value for free RST28,50. Possible structures formed are indicated by the top schematic and key RST-interacting residues, shown as sticks in b, are highlighted in bold. d, Ionic strength dependence of Kd value (error bars are s.d. from two technical replicates). The line is a linear fit of log(Kd) versus [NaCl] with concentrations varying from 10–30 µM RST in the cell and 100–300 µM peptide in the syringe (n = 2 except at 100 mM NaCl, where n = 3). e, Thermodynamic parameters of l- and d-enantiomer binding to RST. f, Stereosensitivity of the five model protein systems shown as the difference in ΔG (ΔΔGd-l, left y axis) and the fractional loss of binding energy (ΔΔGd-l/∣ΔGL∣, right y axis). g, Correlation between total CSP of the partner protein (ProTα, RST or MCL1) induced by l-enantiomer binding and the sensitivity to stereochemistry expressed as fractional loss of binding energy. In f,g, the wedge indicates the extent of bound-state heterogeneity and error bars were calculated from s.d. of ITC parameter shown in Extended Data Tables 2 and 3.

Source Data

Full size image

DREB2A and ANAC013 form additional structures in the core flanking regions, extending into a helix C-terminally (DREB2A) and forming a helical turn N-terminally (ANAC013) (Fig. 4a). This additional structuring was confirmed by the transverse relaxation rates (R2 rates) obtained in the free and RST-bound states (Fig. 4c). In the bound states, DREB2A showed increased R2 rates in the C-terminal flanking region, and ANAC013 increased R2 rates in the N-terminal flanking region, compared with ANAC046 (Fig. 4c). Within the core structure, the number of residues exhibiting dynamics on an intermediate timescale originating from exchange between free and bound states or from heterogeneity in the bound state differed, as indicated by increased R2 rates or missing signals. In complex with RST, ANAC046 showed increased dynamics in its core structure compared with that of DREB2A, which again was more dynamic than in ANAC013 (Fig. 4c). This observation provides an explanation for the observed variation in sensitivity to chirality, suggesting that the RST complexes with greater bound-state dynamics are less sensitive to chirality.

The enthalpic and entropic contributions to binding provide insight into the interaction mechanisms37,38 (Fig. 4e). For the RST complexes, a linear relationship between the changes in enthalpy and entropy is observed across the binding partners, with disordered and d-enantiomer interactions producing smaller changes and an entropic driving force (Fig. 4f). d-DREB2A and d-ANAC013 still bind and do so with an affinity and a thermodynamic profile analogous to those of l-ANAC046 and d-ANAC046. The thermodynamics of RST binding to l-ANAC046 and d-ANAC046 are similar, and in both cases, binding is driven by both entropy and enthalpy, highlighting a near independence from stereochemistry. Assuming that the interaction in these two cases mainly involves the core structure, this on its own has an affinity in the low micromolar range. As the thermodynamic profiles and the affinities for d-DREB2A and d-ANAC013 are similar to those for l-ANAC046 and d-ANAC046, their flanking regions are likely to engage very little with RST in the bound state (Fig. 4e). To assess the apparent graded response to chirality, we calculated for each binding pair the difference in ∆G between l- and d-peptides (∆∆Gd-l) and normalized the ∆∆Gd-l to the strength of the interaction by dividing by the Gibbs free energy of binding the l-enantiomer |∆GL| (Fig. 4f). There is no difference in these parameters between l-H1155–175 and d-H1155–175 interacting with ProTα, whereas interactions requiring backbone structuring show larger ∆∆G values, which increase when more backbone structure is formed. When optimal binding is independent of backbone folding, ambidextrous binding emerges, as for the RST:ANAC046 interaction and is most pronounced for the ProTα:H1 interaction. As more backbone folding occurs, as for the flanking regions of DREB2A and ANAC013 and for PUMA, the dynamics decreases and sensitivity towards chirality increases. Finally, focusing on the folded partner, we assessed the sensitivity to chirality across all systems. We correlated the total CSP induced by the l-enantiomer (CSPtotal–l;  Methods) to the stereosensitivity expressed by the fractional loss of binding energy (∆∆Gd-l/∣ΔGl∣). Although the CSPtotal_l is a crude measure that depends, for example, on the size of the protein, the number of residues involved in binding, resolved peaks, dynamics, allostery and the chemistry, we observe a correlation between the NMR data for l-enantiomer binding and stereosensitivity (Fig. 4g). Thus, the larger the CSPtotal_L of the folded partner, the greater the stereosensitivity.

In conclusion, the analyses of the RST complexes highlight that the propensity for interaction with a d-enantiomer, and thus the sensitivity towards chirality, depends on the extent of disorder and dynamics in the complex. Extending to the full set of five protein pairs, we find that CSPs of the binding partner could provide a method for predicting stereosensitivity.

Discussion

It is counterintuitive that an l-protein should be able to interact with a d-version of its natural partner. However, when considered further, we interrogated whether this applies to a truly disordered interaction. We find that protein complexes that are fully disordered (for example, ProTα:H1) form regardless of chirality. To our surprise, the observation was not limited to completely disordered polyelectrolyte binding partners. Instead, the propensity for interaction between l- and d-proteins exists on a continuum of disorder and order, irrespective of charge and hydrophobicity. The more disordered and dynamic the complex is, the less sensitive it is to chirality (Fig. 4 and Extended Data Table 1). These results, which were obtained from analyses of a set of five different complexes involving IDPs, provide an approach for assessing the degree of disorder within a protein complex directly from its sensitivity to chirality and vice versa. The approach is applicable not only in the case of a fully disordered complex driven by electrostatics, but also for complexes in which critical side chain interactions can be accommodated with adaptable and limited backbone structuring. This continuum of stereosensitivity has translational applications in drug design and implications for our understanding of protein evolution.

Chirality is an important feature for interactions that rely on structure or the formation of structure, but an electrostatic, disordered interaction can proceed regardless. Each of the peptides studied here relies on electrostatics to interact with their respective binding partner24,28,39, therefore, the determining factor appears to be the degree of disorder. PUMA forms an encounter complex with MCL1 owing to long-range electrostatics39, and electrostatics are also integral to the interactions of RST28 (Fig. 4d). For the RST:ANAC046 complex, the interaction is achieved by arranging a few hydrophobic side chains, charges and π-electrons in a way that requires minimal backbone structuring of the core structure providing low sensitivity to chirality. Conversely, interactions of DREB2A and ANAC013 that rely on more extensive backbone folding have higher sensitivity to chirality. Thus, stereosensitivity appears to be rooted in backbone folding. Notably, adaptation of the associating folded partner also contributes to the sensitivity to chirality.

The central core structure, present in all tested RST partners, binds almost independently of chirality in a dynamic state driven by both entropy and enthalpy, while interactions of the flanking regions of the bound DREB2A and ANAC013 impose order, both to the core structure and to the flanking regions. Such strong enthalpy–entropy compensation within the core binding region imposed by dynamic flanking regions was reported previously for the N-terminal flanking region of DREB2A40, and the larger enthalpic contribution for DREB2A binding shows that context matters41.

In this study, we have investigated the disorder–order continuum of protein interactions with similar electrostatic and hydrophobic features (Extended Data Table 1). However, we do not know whether fully disordered hydrophobic complexes exist and whether they are sensitive to chirality, although hydrophobic ligands have been shown to remain disordered in complexes42, and chaperones (such as GroEL/ES) can assist the folding of both l- and d-enantiomers43. Research into highly hydrophobic d-amino acid-based transcriptional coactivators has demonstrated that they can induce transcription to a similar level as their l-enantiomer counterparts44, suggesting that at least some hydrophobic disordered proteins can still be functional regardless of chirality. An implication from our work is that such functional hydrophobic disordered complexes probably rely little on backbone folding. Thus, as IDPs have been historically difficult to target, requiring novel strategies45,46, the presented results have substantial implications for the development of new, stable d-peptide drugs directed against IDPs.

Protein–protein interactions are fundamental for sustaining the information network that separates life from non-living systems. For simplicity, we assume that the quantities of abiotic amino acids of either chirality in the ‘primordial soup’ were equal47. Moreover, peptide bonds may form equally well between l-amino acids and d-amino acids, d-amino acids and d-amino acids, or l-amino acids and l-amino acids48. Therefore, peptide–peptide interactions between heterochiral peptides can be envisaged to have existed before biological systems became homochiral49.

Methods

Synthetic peptides

Synthetic l- and d-peptides of H1155–175, PUMA130–156, ANAC013254–274, ANAC046319–338 and DREB2A255–272 were purchased from Pepscan (now Biosynth) at a minimum purity of 95% and purified by HPLC. The d-peptides contain amino acid residues with a stereoisomeric d-form of each chiral carbon. The peptides were either resuspended in MilliQ H2O or in MilliQ H2O containing 50 mM NH4HCO3 and lyophilized repeatedly to remove leftover trifluoroacetic acid from the last purification step by the manufacturer. Peptides were then either resuspended directly in the buffer used for experiments or in H2O without 50 mM NH4HCO3 to measure the concentration. If no aromatic residue was present in the peptide sequence, the absorbance at 214 nm was used. The extinction coefficient was calculated using Bestsel51.

Expression and purification of proteins

15N-labelled and unlabelled full-length ProTα was expressed and purified as described22. The double-cysteine variant of ProTα (E56C/D110C) used in smFRET experiments was expressed and purified as described24, with some modifications. In brief, ProTα was dialysed against Tris buffer (50 mM Tris, 200 mM NaCl, 2 mM DTT, 1 mM EDTA; pH 8), during which the hexa-histidine tag was cleaved using HRV 3 C protease. Cleaved ProTα was purified further using Ni Sepharose Excel resin (Cytiva, formerly GE Healthcare) and a HiPrep Q FF column (Cytiva) with a gradient from 200 mM to 1 M NaCl. Buffer was exchanged (HiTrap Desalting column (Cytiva)) to labelling buffer potassium phosphate (100 mM, pH 7). 15N-labelled and unlabelled GST–MCL1152–308 was expressed in BL21(DE3)pLysS Escherichia coli in the presence of ampicillin. Cells were grown at 37 °C in LB or M9 minimal medium (for 15N labelling) until OD600 reached 0.6, then induced with IPTG (1 mM final concentration) and collected after 4 h. The cell pellet was resuspended in Tris buffer (20 mM Tris, 100 mM NaCl; pH 8), then lysed by sonication. After pelleting again, the supernatant was applied to GST Sepharose beads (Cytiva), and GST–MCL1152-308 was eluted using Tris-GSH buffer (20 mM Tris, 100 mM NaCl, 10 mM GSH; pH 8). The GST tag was removed using TEV protease (0.7 mg) overnight at room temperature. Final purity was reached using a Superdex 75 26/60 column (Cytiva), equilibrated with 50 mM phosphate buffer (pH 7). 13C,15N-labelled MCL1152–308 was expressed as described52 and purified as above. The expression and purification of 15N-labelled and unlabelled RCD1-RST499–572 were carried out as previously described28 with the lysis buffer changed to 20 mM Tris-HCl, pH 9.0, 20 mM NaCl. The buffer used in the last purification step by size exclusion chromatography on a Superdex 75 10/300 GL column (Cytiva) was the buffer described for the individual methods.

13C,15N-labelled ANAC046319–338 or ANAC013254–274 were expressed with a His6-SUMO fusion tag in BL21(DE3) E. coli in the presence of kanamycin (50 μg ml−1). Cells were grown in LB at 37 °C until OD600 reached 0.6 and the medium was changed to M9 minimal medium, followed by induction with IPTG to 1 mM final concentration and collected after incubation overnight at 16 °C. The cells were resuspended in lysate buffer (50 mM and 20 mM Tris-HCl for ANAC046319–338 and ANAC013254–274, respectively, pH 8.0, 300 mM NaCl) and sonicated. After the centrifugation, the lysate was purified using TALON resin equilibrated in the buffers just described. The fusion peptides were eluted with an equivalent buffer containing 250 mM imidazole. After a dialysis step into 20 mM Tris-HCl pH 8.0, 100 mM NaCl, the fusion tag was cleaved with ubiquitin-like-specific protease 1 (ULP1) (molar ratio between peptide and protease were 1:320 and 1:500 for ANAC046319–338 and ANAC013254–274, respectively) overnight at 4 °C. A second purification step with TALON resin was performed resulting in the peptides in the flowthrough. The purification of the peptides was finalized by size exclusion chromatography on a Superdex peptide 10/300 GL column (Cytiva) and freeze-dried to be resuspended in the desired buffer.

AlphaFold structure modelling

Protein interaction models of RCD1-RST499–572 in complex with ANAC046319–338 or ANAC013254–274 were generated using AlphaFold330 and analysed in PyMOL (The PyMOL Molecular Graphics System, version 3.0 Schrödinger, LLC.). The five generated models for each complex were assessed manually and compared with the secondary chemical shifts of Cα of the l-ligand recorded using ZZ-exchange or CEST (see NMR spectroscopy method). The structures agreeing with the experimental data were visualized in PyMOL or Chimera X53.

Far-UV CD spectropolarimetry

Far-UV CD spectra of l- and d-peptides of H1155–175, PUMA130–156, ANAC013254–274, ANAC046319–338, and DREB2A255–272 were measured on a Jasco 815 spectropolarimeter with a Jasco Peltier control in the range of 260–190 nm at 20 °C. Concentrations of peptides varied between 10–30 µM in either MilliQ H2O, pH 7.0 (PUMA130–156, H1155–175) or 20 mM NaH2PO4/Na2HPO4, pH 7.0 (ANAC013254–274, ANAC046319–338, DREB2A255–272) with 1 mM TCEP in the samples containing ANAC046 peptides. A quartz cuvette with a 1 mm path length was used and 10 scans were recorded and averaged with a scanning speed of 20 nm min−1 and response time of 2 s. A spectrum of the buffer using identical setting was recorded for each protein and subtracted the sample spectrum.

NMR spectroscopy

All NMR spectra were recorded on Bruker Avance III 600 MHz, 750 MHz or an Avance NEO 800 MHz (for 1H) spectrometers equipped with cryoprobes. Natural abundance 1H,15N and 1H,13C-HSQC spectra were recorded on all peptides at either 10 °C or 25 °C. Peptides (0.5 mM) in sample buffer containing 20 mM Na2HPO4/NaH2PO4 pH 7.0, 100 mM NaCl, 10 % (v/v) D2O, 0.02 % (w/v) NaN3 and 0.7 mM 4,4-dimethyl-4-silapentane-1-sulfonic acid (DSS) for ANAC046319–338, ANAC013254–274 and DREB2A255–272 with the addition of 1 mM DTT in the samples containing ANAC046 peptides. 1H,15N-HSQC spectra were recorded on 50 µM ProTα, with or without 500 µM l- or d-H1155–175 in TBSK (ionic strength 165 mM; pH 7.4). 1H,15N-HSQC spectra were recorded on 50 µM MCL1, with or without 45 µM l- or 2.5 mM d-PUMA130–156, in Tris (50 mM; pH 7.0) to compare at 90% saturation, as calculated from Kd values. Assignments of 13C,15N-MCL1 in complex with l-PUMA130–156 were completed from a series of HNCACB and HNCOCACB 3D spectra as described54, and deposited to Biological Magnetic Resonance Data Bank (BMRB) under accession 52264. 1H,15N-HSQC spectra were recorded on 15N-labelled 100 µM RCD1-RST499–572 in 20 mM Na2HPO4/NaH2PO4 pH 7.0, 100 mM NaCl, 10 % (v/v) D2O, 0.02 % (w/v) NaN3 and 0.7 mM DSS at 25 °C in the absence and presence of each stereoisomeric forms of 0−200 µM ANAC046319–338, ANAC013254–274 and DREB2A255–272 in the following ratios; 1:0, 1:0.2, 1:0.4, 1:0.6, 1:0.8, 1:1 and 1:2. Assignments of free ProTα and free RCD1-RST were taken from BMRB entries 27215 and 50545, respectively22,28.

Amide CSPs were calculated from the 1H,15N-HSQCs in the absence and presence of the highest concentration of peptide used for each interaction using equation (1):

$${\Delta \delta }_{{\rm{NH}}}\,({\rm{ppm}})=\sqrt{{(\Delta {\delta }^{1}{\rm{H}})}^{2}+{(0.154\times \Delta {\delta }^{15}{\rm{N}})}^{2}}$$
(1)

The total protein CSP (CSPtotal_L) induced by the binding of the l-enantiomer peptide was quantified by recording the CSPs of all visible 15N,1HN backbone resonances at >90% saturation (MCL1: 90%, RST (all cases): >99%, ProTα: >98%). The CSP for all visible residues were summed to obtain the total CSP. To adjust for unassigned residues, which include prolines, residues that could not be assigned, or residues not visible in either the bound or unbound states, the total CSP was divided by the fraction of residues for which CSPs were recorded. For instance, if CSPs were obtained for only half of the residues, the calculated total CSP was doubled to estimate the perturbation as if all residues were visible. This adjustment ensured that the total CSP could be compared between interactions, accounting for the lack of data from unassigned or invisible residues. The adjustment does not account for the fact that disappearing residues are likely involved in the interaction and thus also likely to experience larger than average CSPs.

2D NMR lineshape analysis

2D NMR lineshape analyses were performed for interactions of l-and d-peptides with RCD1-RST499–572. The recorded 1H,15N-HSQC spectra were processed using qMDD with exponential weighting functions with 4 Hz and 8 Hz line broadening in the direct and indirect dimensions, respectively. The 2D lineshape analysis was performed using the tool TITAN31 in Matlab (Mathworks) and was based on well-separated spin systems that were easily followed. If the trajectory of spin systems overlapped, the spin systems were grouped during fitting. All titrations were fitted to a two-state binding model, and at least 12 spin systems were picked for each analysis. Due to initial poor fitting for the titrations of the interaction 15N-RCD1-RST499–572 and l- ANAC013254–274, the Kd value was fixed using the values determined from ITC. Errors were determined by a bootstrap analysis using 100 replicas to determine the standard error from the mean. From the lineshape analysis, the fitted Kd and koff values were used to calculate the association rate constant (kon) based on equation (2):

$${K}_{{\rm{d}}}=\frac{{k}_{{\rm{off}}}}{{k}_{{\rm{on}}}}$$
(2)

The differences in activation free energies for binding between d- and l-peptides were estimated from the ratios of the association rate constants for both stereoisomers, \({k}_{{\rm{on}}}^{{\rm{D}}}\) and \({k}_{{\rm{on}}}^{{\rm{L}}}\), based on equation (3):

$${\Delta \Delta G}_{{\rm{unbound}}-\ddagger ,{\rm{D}}-{\rm{L}}}=RT{\rm{ln}}\left(\frac{{k}_{{\rm{on}}}^{{\rm{L}}}}{{k}_{{\rm{on}}}^{{\rm{D}}}}\right),$$
(3)

which was rewritten from Fersht (equation 18.22 in ref. 55).

CEST NMR

CEST experiments were recorded for the l-peptide of ANAC046319–338 to determine the chemical shift of its bound state with RCD1-RST499–572. All experiments were recorded on a Bruker Avance Neo 800 spectrometer with a cryoprobe. A sample of 1 mM 13C,15N-labelled l-ANAC046319–338 was prepared with 5% molar ratio of RCD1-RST499–572 in 20 mM Na2HPO4/NaH2PO4 pH 6.5, 100 mM NaCl, 10 % (v/v) D2O, 0.02 % (w/v) NaN3, 0.7 mM DSS and 5 mM DTT. 15N-CEST data was acquired using pulse sequences as previously described56 at 25 °C using three different B1 field strengths: 6.25, 12.5 and 25 Hz. 13C-CEST data were acquired using special pulse sequences57,58 (provided by L. Kay) as done in ref. 59 at 25 °C with a B1 field strength of 25 Hz. The free induction decays were transformed using NMRPipe60 and peak intensities were extracted from each specific peak position. The intensities were analysed using ChemEx61 by fitting to a global two-state model implemented in the program. The fits reported on the change in chemical shifts for peaks experiencing CEST-transfer which directly reflects the chemical shift of the bound state of the peptide. The chemical shifts were extracted for the Cα and compared to a reference set62.

ZZ-exchange

For the complex between RST and 15N-ANAC013254–274, identification of residues and their assignments were resolved by 3D heteronuclear NMR experiments with additional ZZ-exchange63 NMR spectra recorded on a 50% saturated sample of 100 µM 13C, 15N-ANAC013254–274 with 50 µM RCD1-RST499–572 in20 mM Na2HPO4/NaH2PO4 pH 6.5, 200 mM NaCl, 10 % (v/v) D2O, 0.02 % (w/v) NaN3, and 0.7 mM DSS. The ZZ-exchange connections made it possible to manually track the assignment from the 1H,15N-HSQC spectrum of the unbound 15N-ANAC013254–274 to the RST-bound 15N-ANAC013254–274. For the assignments of carbon resonances of ANAC013, two samples were prepared: 13C, 15N-ANAC013254–274 (650 µM) w/wo RCD1-RST499–572 (800 µM) in 20 mM Na2HPO4/NaH2PO4 pH 6.5, 200 mM NaCl, 10 % (v/v) D2O, 0.02 % (w/v) NaN3, and 0.7 mM DSS. Backbone resonances for the unbound peptide were manually assigned from analysis of 15N-HSQC, HNCA, HNCO and HNCACB experiments. All NMR spectra were acquired at 25 °C on a Bruker Avance III 750 MHz, except for ZZ-exchange which was on Bruker Avance III 600 MHz. All 3D experiments were recorded using non-uniform sampling.

Secondary chemical shifts

SCSs were calculated using the POTENCI62 web tool.

Transverse relaxation

To determine the dynamics of l-ANAC046319–338 and l-ANAC013254–274 w/wo RCD1-RST499–572, the sample from ANAC013254–274 assignment was reused whereas a new for ANAC046319–338 was made: 75 µM 13C, 15N-ANAC046319–338 with 180 µM RCD1-RST499–572 in 20 mM Na2HPO4/NaH2PO4 pH 6.5, 100 mM NaCl, 10 % (v/v) D2O, 0.02 % (w/v) NaN3, 0.7 mM DSS and 5 mM DTT. The transverse relaxation rates, R2 values, were acquired on a Bruker Avance Neo 800 spectrometer with the following relaxation delays: 33.8 ms, 67.6 ms, 101.4 ms, 169.0 ms, 236.6 ms, 270.4 ms, 338.0 ms and 405.6 ms (all triplicates), and a recycle delay of 2 s. Data were fitted to a one phase decay function.

Isothermal titration calorimetry

Prior to ITC, all samples were spun down at 17,000g for 10 min at the experimental temperature. ITC experiments involving ProTα and MCL1152–308 as interaction partners were recorded on MicroCal PEAQ-ITC microcalorimeter (Malvern Panalytical). ProTα (7.1 µM) was placed in the cell and either l- or d-H1155–175 (99.1 µM) in the syringe, in TBSK (165 µM ionic strength) at 20 °C. Each injection was 2 µl, with a total of 19 injections at an interval of 150 s between each. Data were fit using a fixed number of binding sites (fixed to one) so that fits could be standardized. For the MCL1152-308 interactions, MCL1152-308 (10 µM) was placed in the cell, with either l- or d-PUMA130–156 (100 µM) in the syringe, in Tris (50 mM; pH 7.0) at 25 °C. Each of the 35 injections was 1 µl, with an interval of 150 s between each. The experiment was repeated for MCL1:d-PUMA130–156, increasing the concentrations to 70 and 700 µM, respectively, while keeping the remaining experimental conditions identical. ITC experiments involving RCD1-RST499–572 as interaction partner were recorded on a MicroCal ITC200 microcalorimeter (MicroCal Instruments) at 25 °C in 50 mM Na2HPO4/NaH2PO4 pH 7.0, 100 mM NaCl. TCEP (1 mM) was added the sample buffer for interactions involving ANAC046 peptides. Concentrations of RCD1-RST499–572 varied between 10–100 µM in the cell and 100-1000 µM of the ANAC046, ANAC013 or DREB2A peptides in the syringe. The first injection was 0.5 µl followed by 18 repetitions of 2 µl injections separated by 180 seconds. These experiments were processed using the Origin7 software package supplied by the manufacturer. The last 18 injections of each experiment were fitted to a one set of sites binding model. Triplicates were recorded for each interaction.

A salt titration was performed measuring the interaction between RCD1-RST499-572 and the l-peptides of ANAC046319–338 and ANAC013254–274 by ITC, varying the NaCl concentration in the experimental buffer. Experiments were recorded on a MicroCal PEAQ-ITC microcalorimeter or a MicroCal ITC200 microcalorimeter at 25 °C. A 50 mM Na2HPO4/NaH2PO4 pH 7.0, 1 mM TCEP buffer was used with NaCl concentrations at 0, 50, 150 and 200 mM, with data at 100 mM NaCl recorded prior to and included in the analysis. Protein and peptide concentrations varied from 10–30 µM in the cell (RCD1-RST) and 100–300 µM in the syringe (peptides). A replica of each experiment was produced, and the isotherm were fitted as described above.

Fluorophore labelling for smFRET

ProTα was labelled by incubating it with Alexa Fluor 488 (0.7:1 dye to protein molar ratio) for 1 h at room temperature and sequentially with Alexa Fluor 594 (1.5:1 dye to protein molar ratio) overnight at 4 °C. Labelled protein was purified using a HiTrap Desalting column and reversed-phase high-performance liquid chromatography (RP-HPLC) on a SunFire C18 column (Waters Corporation) with an elution gradient from 20% acetonitrile and 0.1% trifluoroacetic acid in aqueous solution to 37% acetonitrile. ProTα-containing fractions were lyophilized and dissolved in buffer (10 mM Tris, 200 mM KCl, 1 mM EDTA; pH 7.4).

Single-molecule FRET measurements and analysis

Single-molecule fluorescence experiments were conducted using either a custom-built confocal microscope or a MicroTime 200 confocal microscope (PicoQuant) equipped with a 485-nm diode laser and an Olympus UplanApo 60×/1.20 W objective. Microscope and filter setup were used as previously described24. The 485-nm diode laser was set to an average power of 100 μW (measured at the back aperture of the objective), either in continuous-wave or pulsed mode with alternating excitation of the dyes, achieved using pulsed interleaved excitation (PIE)64. The wavelength range used for acceptor excitation in PIE mode was selected with a z582/15 band pass filter (Chroma) from the emission of a supercontinuum laser (EXW-12 SuperK Extreme, NKT Photonics) driven at 20 MHz, which triggers interleaved pulses from the 485-nm diode laser used for donor excitation. In our experiments, photon bursts (at least 3000 bursts) were selected against the background mean fluorescence counts and, in case of PIE, by having a stoichiometry ratio S of \(0.2 < S < 0.75\), each originating from an individual molecule diffusing through the confocal volume. Transfer efficiencies were quantified according to \(E={n}_{{\rm{A}}}/({n}_{{\rm{A}}}+{n}_{{\rm{D}}})\), where \({n}_{{\rm{D}}}\) and \({n}_{{\rm{A}}}\) are the numbers of donor and acceptor photons in each burst, respectively, corrected for background, channel crosstalk, acceptor direct excitation, differences in quantum yields of the dyes, and detection efficiencies. All smFRET experiments were performed in µ-Slide sample chambers (Ibidi) at 22 °C in TEK buffer with an ionic strength of 165 mM fixed with KCl; 140 mM 2-mercaptoethanol and 0.01% (v/v) Tween-20 were added for photoprotection and for minimizing surface adhesion, respectively. Single-molecule data were analysed using the Mathematica (Wolfram Research) package Fretica (https://schuler.bioc.uzh.ch/programs). For quantifying binding affinities, transfer efficiency histograms were constructed from single-molecule photon bursts identified as described above. Each histogram was normalized to an area of 1 and fit with a Gaussian peak function to extract its mean transfer efficiency \(\langle E\rangle \). The mean transfer efficiency as a function of increasing concentration of d/l-H1155–175, \(\langle E\rangle ({C}_{{\rm{D/L-H1}}})\), was fit with:

$$\begin{array}{l}\langle E\rangle ({C}_{{\rm{D}}/{\rm{L}}-{\rm{H}}1}^{{\rm{t}}{\rm{o}}{\rm{t}}})\,=\Delta {\langle E\rangle }^{{\rm{s}}{\rm{a}}{\rm{t}}}\\ \times \frac{{C}_{{\rm{D}}/{\rm{L}}-{\rm{H}}1}^{{\rm{t}}{\rm{o}}{\rm{t}}}+{K}_{{\rm{d}}}+{C}_{{\rm{P}}{\rm{r}}{\rm{o}}{\rm{T}}\alpha }^{{\rm{t}}{\rm{o}}{\rm{t}}}-\sqrt{{({C}_{{\rm{D}}/{\rm{L}}-{\rm{H}}1}^{{\rm{t}}{\rm{o}}{\rm{t}}}+{K}_{{\rm{d}}}+{C}_{{\rm{P}}{\rm{r}}{\rm{o}}{\rm{T}}\alpha }^{{\rm{t}}{\rm{o}}{\rm{t}}})}^{2}-4{C}_{{\rm{D}}/{\rm{L}}-{\rm{H}}1}^{{\rm{t}}{\rm{o}}{\rm{t}}}{C}_{{\rm{P}}{\rm{r}}{\rm{o}}{\rm{T}}\alpha }^{{\rm{t}}{\rm{o}}{\rm{t}}}}}{2{C}_{{\rm{P}}{\rm{r}}{\rm{o}}{\rm{T}}\alpha }^{{\rm{t}}{\rm{o}}{\rm{t}}}}+{\langle E\rangle }_{0}\end{array}$$
(4)

Here, \({C}_{{\rm{D/L-H1}}}^{{\rm{tot}}}\) and \({C}_{{\rm{ProT\alpha }}}^{{\rm{tot}}}\) are the total concentration of d/l-H1155–175 and ProTα, respectively, \({\langle E\rangle }_{0}\) is the mean transfer efficiency of free ProTα, and \({\Delta \langle E\rangle }^{{\rm{sat}}}\) is the increase in transfer efficiency from free ProTα to ProTα saturated with d/l-H1155–175, while \({K}_{{\rm{d}}}\) is the equilibrium dissociation constant.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.