Protein yoga: Conformational versatility of the Hemolysin II C‐terminal domain detailed by NMR structures for multiple states

Abstract

The C‐terminal domain of Bacillus cereus hemolysin II (HlyIIC), stabilizes the trans‐membrane‐pore formed by the HlyII toxin and may aid in target cell recognition. Initial efforts to determine the NMR structure of HlyIIC were hampered by cis/trans isomerization about the single proline at position 405 that leads to doubling of NMR resonances. We used the mutant P405M‐HlyIIC that eliminates the cis proline to determine the NMR structure of the domain, which revealed a novel fold. Here, we extend earlier studies to the NMR structure determination of the cis and trans states of WT‐HlyIIC that exist simultaneously in solution. The primary structural differences between the cis and trans states are in the loop that contains P405, and structurally adjacent loops. Thermodynamic linkage analysis shows that at 25 C the cis proline, which already has a large fraction of 20% in the unfolded protein, increases to 50% in the folded state due to coupling with the global stability of the domain. The P405M or P405A substitutions eliminate heterogeneity due to proline isomerization but lead to the formation of a new dimeric species. The NMR structure of the dimer shows that it is formed through domain‐swapping of strand β5, the last segment of secondary structure following P405. The presence of P405 in WT‐HlyIIC strongly disfavors the dimer compared to the P405M‐HlyIIC or P405A‐HlyIIC mutants. The WT proline may thus act as a “gatekeeper,” warding off aggregative misfolding.

Short abstract

PDB Code(s): 6D53, 6D5Z and 6WA1;

Abbreviations

CD

circular dichroism

H‐bond

hydrogen‐bond

HlyII

hemolysin II

HlyIIC

C‐terminal domain corresponding to residues D319‐I412 of hemolysin II from B. cereus

HSQC

heteronuclear single‐quantum correlation

NMR

nuclear magnetic resonance

PFG

pulsed field gradient

RMSD

root‐mean‐square deviation

SEM

standard error of the mean

SN

staphylococcal nuclease

WT

wild type

β‐PFT

β‐pore‐forming toxin

1. INTRODUCTION

Bacillus cereus is a pathogenic, gram‐positive, soil‐dwelling bacterium that plays roles in food poisoning ¹ , ² and other types of infections including necrotizing fasciitis, endophthalmitis, septicemia, meningitis, endocarditis, and pericarditis. ² , ³ B. cereus bacteria secrete a variety of toxins including phospholipases, enterotoxins, and hemolysins. ⁴ Among the secreted toxins, hemolysin II (HlyII) is not associated with a particular infection type but has been shown to be toxic to a variety of cell types including erythrocytes, monocytes, dendritic cells, and colon carcinoma cells. ⁵ , ⁶ , ⁷ In addition, HlyII can induce apoptosis in macrophages in vitro and in vivo, ⁷ suggesting possible functions in circumventing host innate immunity.

HlyII is classified as a β‐pore‐forming toxin (β‐PFT) due to its sequence similarity to related toxins such as CytK from B. cereus, and β‐toxin from C. perfringens. The closest sequence homolog with 31% sequence identity is α‐hemolysin from S. aureus. ⁶ , ⁸ , ⁹ , ¹⁰ Like its β‐PFT homologs, HlyII utilizes a pore‐forming mechanism to permeabilize cell membranes. After synthesis, cleavage of an N‐terminal signal peptide releases a water‐soluble 412 a.a. toxin that can bind target cells as a monomer or oligomer. ¹¹ Once on the target cell surface, the HlyII subunits oligomerize to form a pre‐pore complex that subsequently transition to a heptameric lytic trans‐membrane β‐barrel pore. ¹⁰ , ¹² There are no known cell surface receptors specific for HlyII but since its cytotoxicity extends toward a variety of cell types, there may be a ubiquitous moiety that facilitates its oligomerization on a variety of target cell surfaces.

The HlyII from B. cereus has a unique C‐terminal domain (HlyIIC) that shows no sequence or structural homology to other proteins in the Protein Data Bank. ⁶ , ¹³ , ¹⁴ To characterize this domain, we used a 94 amino acid (10.8 KDa) HlyIIC construct, corresponding to residues D319‐I412 of the full length HlyII toxin. ¹³ We use the numbering scheme of the full‐length HlyII toxin throughout this manuscript. The principal function of the HlyIIC domain remains unknown. While HlyIIC is not required for trans‐membrane pore formation, the absence of the domain leads to a eight‐fold reduction in cytotoxicity. ⁶ Deletion of the HlyIIC domain also lowers the thermal stability of trans‐membrane pore oligomers by 4 C compared to wild type (WT). Based on modeling of HlyIIC in the context of full‐length HlyII, we postulated the C‐terminal domain could stabilize the oligomeric pore complex through interactions with membrane lipids on the surfaces of target cells. ¹⁴

The HlyIIC domain is unusual in having a high degree of flexibility and conformational plasticity. ¹⁴ Evolutionary pressures have selected most globular proteins to have a single native state structure. ¹⁵ Proteins that visit multiple conformational states are rare because multiple structures can lead to a variety of functions, some of which could result in disease. Perhaps because HlyIIC is part of a secreted toxin rather than a protein essential for B. cereus viability, it may have had less constraints on its structural promiscuity. NMR spectra of WT‐HlyIIC show a doubling of crosspeaks due to cis/trans peptide bond isomerization of the sole proline in the sequence at position 405. ¹³ To reduce spectral complexity, we initially used a P405M mutant to eliminate the conformational heterogeneity caused by proline cis/trans isomerization. ¹⁴ The NMR structure of the P405M mutant showed that the HlyIIC domain has a novel fold, comprised of two α‐helices and five β‐strands arranged in a pseudo‐barrel structure. ¹⁴ Here, we extend the initial work on the P405M‐HlyIIC mutant by determining the structure and dynamics of the trans and cis forms of WT‐HlyIIC. We characterize the thermodynamics of interconversion between the cis and trans forms, and determine the extent of thermodynamic coupling between proline isomerization and protein folding. The P405M‐HlyIIC mutant, with a trans peptide bond at position 405, is subject to a monomer‐dimer equilibrium. We characterize the dimeric state of the P405M‐HlyIIC mutant using ¹⁵N transverse relaxation rates and pulse field gradient (PFG) diffusion experiments. ¹³C‐filtered NOESY experiments in conjunction with samples of mixed ¹³C and ¹²C‐labeled chains, show that the dimer is formed through domain‐swapping of the C‐terminal strand β5, which immediately follows the P405‐harboring loop. We determine dissociation constants for the dimer at a series of temperatures to characterize the thermodynamics of oligomerization, and show that the dimer is strongly disfavored in WT‐HlyIIC compared to the P405M‐HlyIIC mutant.

2. RESULTS

2.1. The cis and trans state structures differ primarily in the proline‐bearing loop and have similar dynamics

The NMR structures of the cis and trans states of WT‐HlyIIC based on the experimental restraints detailed in Table 1 are shown in Figure 1. About half the residues in the WT‐HlyIIC domain show doubled crosspeaks in ¹H‐¹⁵N HSQC NMR spectra due to cis/trans isomerization of the G404‐P405 peptide bond (Figure S1). ¹³ , ¹⁴ That separate NMR signals are observed for the cis and trans conformers of WT‐HlyIIC, indicates they are in slow exchange on the NMR chemical shift timescale. ZZ‐exchange NMR experiments ¹⁶ performed on WT‐HlyIIC failed to detect magnetization transfer between cis and trans conformers with mixing times as long as 630 ms. (Figure S2). The lack of exchange crosspeaks signifies an interconversion rate slower than about 1 s⁻¹, and that the two species are also in slow exchange on the T1 NMR timescale. Consequently, WT‐HlyIIC protons show different NOE patterns, and different ¹⁵N relaxation values for the two forms. This allowed us to calculate non‐averaged NMR structures for the cis and trans states, as well as to determine distinct dynamic properties.

TABLE 1.

Statistics for the 25 lowest‐energy NMR structures of various HlyIIC states

NMR structure	trans WT‐HlyIIC	cis WT‐HlyIIC	Dimer P405M‐HlyIIC
Experimental restraints
NMR restraints (total)	1761	1790	1694 × 2 a
Distance (total)	1566	1603	1554 × 2
Intraresidue NOEs	657	729	676 × 2
Sequential NOEs	236	234	417 × 2
Medium‐range NOEs (|i–j| < 5)	180	132	150 × 2
Long‐range NOEs ((|i–j ≥ 5)	405	428	199 × 2
Inter‐chain NOEs	0	0	12 × 2
Hydrogen bonds × 2 b	44	40	50 × 2 c
Dihedral (ϕ, ψ, χ1)	195	187	140 × 2
Residual restraint violations d
NOE (Å)	0.0414 ± 0.0003	0.0644 ± 0.0002	0.0777 ± 0.0004
Dihedral (o)	0.62 ± 0.02	0.88 ± 0.02	0.93 ± 0.03
RMSD from ideal geometry
Bonds (Å)	0.005 ± 3E‐5	0.007 ± 3E‐5	0.008 ± 2E‐4
Angles (o)	0.648 ± 0.004	0.830 ± 0.004	0.851 ± 0.005
Impropers (o)	2.09 ± 0.02	2.06 ± 0.02	0.68 ± 0.05
Ramachandran statistics e
Most favored (%)	81.4	76.6	71.3
Allowed (%)	17.7	20.0	25.3
Generously allowed (%)	1.0	2.8	3.3
Disallowed (%)	0	0.7	0
Coordinate RMSD
Backbone (residues 319–412)	1.29 ± 0.27	1.30 ± 0.27	1.53 ± 0.42 (2.81 ± 0.85) g
Heavy (residues 319–412)	1.76 ± 0.24	1.72 ± 0.25	1.93 ± 0.42 (3.12 ± 0.87) g
Backbone (residues 330–412) f	0.65 ± 0.15	0.51 ± 0.08	1.25 ± 0.26 (2.68 ± 0.87) g
Heavy (residues 330–412) f	1.09 ± 0.14	0.93 ± 0.06	1.61 ± 0.26 (3.08 ± 0.92) g

Abbreviations: HlyIIC, C‐terminal domain corresponding to residues D319‐I412 of the hemolysin II from B. cereus; WT, wild type.

^aAll NMR restraints were duplicated for chains A and B of the dimer to maintain two‐fold symmetry.

^bEach restraint was specified by both N—O and NH—O distance bounds to maintain H‐bond linearity.

^cIncludes three (×2) interchain H‐bonds between strand β5 and strand β3 of the neighboring monomer, consistent with the domain‐swapping of strand β5 evinced by the inter‐chain NOESY data.

^dValues ± SEM. Structures have no NOE violations >0.5 Å or dihedral violations >5°.

^eRamachandran plot summary for selected residues from Procheck calculated using the PSVS validation server. ⁵⁴

^fThe first 11 residues of HlyIIC were e×cluded because they are more disordered than the rest of the structure and show low‐S² order parameters (Figure S5A).

^gThe first number gives the coordinate RMSDs of the protomers and describes the structural precision of the subunits, the second number in parentheses is the overall RMSD for the dimer.

FIGURE 1.

NMR structures of the cis and trans states of WT‐HlyIIC. (a) Superposition of the 25 lowest energy NMR structures for the cis (orange) and trans (cyan) states. The position in the structure of the unique proline at position 405 is shown. (b) Precisions of the cis (orange) and trans (cyan) NMR ensembles, calculated as the mean RMSD from the ensemble coordinate average. Also shown in the plot is the RMSD between the cis and trans NMR structures (black). Four regions where the structural differences (in black) are larger than the uncertainty in each structure bundle (blue or orange) are indicated with the Roman numerals I–IV. (c) Structural differences and backbone dihedral angles between the cis (orange) and trans (cyan) states for the proline‐bearing loop L _αBβ5. (d) Illustration of selected differences in side‐chain interactions between the cis and trans states. HlyIIC, C‐terminal domain corresponding to residues D319‐I412 of the hemolysin II from B. cereus; WT, wild type

The NMR structures of the cis and trans states (Figure 1(a)) are close to each other (1.6 Å backbone root‐mean‐square deviation (RMSD)) and to the previously reported NMR structure for the P405M‐HlyIIC mutant ¹⁴ (backbone RMSD 2.7 Å for trans, 2.8 Å for cis). To better quantify differences between the WT‐HlyIIC cis and trans state structures we did an analysis of the sequence dependence of the backbone RMSD within the cis (orange) and trans (blue) NMR ensembles (Figure 1(b)), compared to the RMSD between the cis and trans structure means (black in Figure 1(b)). There are four regions (labeled I–IV in Figure 1(a),(b)) where we can be confident of genuine differences in structure, since the RMSD between the cis and trans state structures (black line) is larger than the RMSD due to the NMR precision of the cis and trans ensembles (orange and blue lines, respectively). The largest differences occur in region IV ‐ the proline‐harboring loop L _αBβ5, and region II the structurally adjacent hairpin between strands β1 and β2, L _β1β2. Smaller differences are also seen for the structurally adjacent region I comprised of helix αA and region III corresponding to the end of helix αB.

The proline‐harboring loop L _αBβ5 that shows the largest RMSD, has different backbone dihedral angles in the cis and trans state structures, as summarized in Figure 1(c). The residues most strongly affected are G404 and Y406 that flank the unique proline P405. The differences in ϕ angles between the two states lead to conformations that are near mirror images of each other (Figure 1(c)). The dihedral angles in this region are less energetically favorable for the cis state, suggesting it is a strained conformation compared to the trans state. Other significant structural differences include that in the trans state the positively charged side chain amino group from K403 near the proline is poised to interact with the negative macrodipole at the C‐terminus of helix αB (Figure 1(d)). In the cis state this interaction is lost, and the amino group of K403 switches to a H‐bonding interaction with the backbone carbonyl of N352 from the adjacent loop L _β1β2 (Figure 1(d)). These changes could account for structural differences in L _αBβ5 , L _β1β2, and helix αB, three of the four most prominent features in the RMSD difference map of Figure 1(b). In the cis state strand β5 begins one residue earlier than in the trans state at Y406 due to a β‐sheet backbone hydrogen between the amide of Y406 and the carbonyl of N377. This hydrogen bond is supported by an extra ³J_NC’ H‐bond coupling for the cis state in a long‐range HNCO experiment. Smaller structural differences, manifested by amide proton NMR resonance splitting throughout the domain, are probably transmitted from P405 to the rest of the structure through the three‐layer hydrophobic core that supports the pseudo‐barrel folding topology of HlyIIC. ¹⁴

We performed the same type of RMSD analysis on the structural differences between the trans state of WT‐HlyIIC and the previously published ¹⁴ P405M‐HlyIIC mutant (Figure S3). The trans and cis states of WT‐HlyIIC exist simultaneously in solution and consequently have structures closer to each other (1.6 Å RMSD) than to the P405M‐HlyIIC mutant (2.6 Å RMSD) that had to be studied at a lower temperature of 15 C to suppress dimerization. ¹³ A primary difference between the WT and the mutant, is that in the latter the proline‐loop and the last strand β5 are splayed out away from the center of the pseudo‐barrel fold compared to either the cis or trans WT structures (Figure S3A). The structural changes associated with the proline‐loop and β5 give rise to the largest chemical shift differences between the mutant and WT, ¹³ and may contribute to the increased proclivity of the mutant to dimerize. ¹³

We next compared the backbone dynamics of the cis and trans states of WT‐HlyIIC using ¹⁵N relaxation data (Figure S4). The data were analyzed in terms of the Model Free formalism, ¹⁷ , ¹⁸ to obtain S ² order parameters that describe the amplitudes of bond vector motions on the ps to ns timescale (Figure S5). S ² order parameters are nearly invariant between the cis and trans states (Figure S5). Exceptions where the cis state appears to be slightly more flexible include G404‐I409 corresponding to the proline‐bearing loop L _αBβ5 and F375‐E379 in the structurally adjacent loop L _β3β4. The differences are small on the order of ∆S ² ≤ 0.1, however, indicating that overall the backbone dynamics of the cis and trans states are very similar.

2.2. The cis state is substantially populated in urea‐unfolded WT‐HlyIIC

We next used NMR to characterize the cis/trans equilibrium in WT‐HlyIIC unfolded by 6 M urea (Figure S6A). As with folded WT‐HlyIIC, the unfolded state shows minor form resonances due to the cis isomer of the only proline in the sequence, P405. Partial NMR assignments for residues close in the sequence to P405 were obtained from separate sequential walks for the cis and trans states using 3D HNCACB data, with the assumption that smaller intensity resonances correspond to the cis state. Using the signals from G404, which shows well‐resolved signals for both the major trans and minor cis conformations, we were able to determine an equilibrium constant K_Uc/Ut = 0.25 ± 0.06 for cis/trans isomerization in urea‐unfolded HlyIIC at 25 C (Table 2). Interestingly, proline isomerization in the urea‐denatured state has a nonzero ∆H of 4.4 ± 1.0 kcal/mol comparable to that of the folded state (Table 2). A recent study of proline isomerization in unfolded proteins found that the fraction of cis isomers calculated as [cis]/([cis] + [trans]) is on the order of 3%–10%, much smaller than values of 10%–20% previously reported in the literature. ²⁰ In the present case, the fraction of cis isomer for HlyIIC unfolded in 6 M urea is ~20%. As an alternative, we looked at the hexapeptide Ac‐KGPYIE‐NH₂, to model P405 in the unfolded state of WT‐HlyIIC in the absence of denaturants. Complete NMR assignmens were obtained for the cis and trans states of the peptide (Table S1), and characteristic dαα and dαδ X‐Pro NOEs unequivocally established that the more populated state corresponds to the trans isomer (Figure S7). The equilibrium constant K_Uc/Ut = 0.33 ± 0.03 for proline isomerization in the peptide (Figure S7) is very similar to that obtained for the urea‐denatured protein, and much larger than in the recent literature report. ²⁰

TABLE 2.

Thermodynamic parameters (kcal/mol) for conformational changes in the HlyIIC domain at a temperature of 298 K

Global unfolding of WT‐HlyIIC
∆GU/F	2.95 ± 0.19 a
Cis/trans isomerization of WT‐HlyIIC
∆GFt/Fc	0.00 ± 0.10 b
∆HFt/Fc	3.46 ± 0.30 c
T∆SFt/Fc	3.54 ± 0.31 c
∆GUt/Ft	2.67 ± 0.12 d
∆HUt/Ft	N.D. e
T∆SUt/Ft	N.D. e
∆GUc/Ut	0.82 ± 0.14 b
∆HUc/Ut	4.39 ± 1.05 c
T∆SUc/Ut	3.65 ± 1.05 c
∆GFc/Uc	−3.49 ± 0.06 d
∆HFc/Uc	N.D. e
T∆SFc/Uc	N.D. e
Dimerization of HlyIIC‐P405M
∆Gd	3.94 ± 0.08 f
∆Hd	−14.3 ± 1.7 g
T∆Sd	−18.2 ± 1.7 g

Abbreviations: HlyIIC, C‐terminal domain corresponding to residues D319‐I412 of the hemolysin II from B. cereus; WT, wild type.

^aObtained from a urea denaturation experiment of WT‐HlyIIC measured by circular dichroism ellipticity at 220 nm.

^bFrom the equilibrium constant for trans/cis isomerization determined by NMR peak volumes of the cis and trans states.

^cFrom a van 't Hoff analysis of the temperature dependence for the trans/cis equilibrium constant.

^dFrom a thermodynamic linkage analysis relating the four states N_c, N_t, U_c, U_t (Figure S8) and K_TS = (Ut + Uc)/(Ft + Fc), as described previously. ²²

^eNot determined.

^fChange in free energy upon dissociation, calculated using the K_d of (1.3 ± 0.1) x 10⁻³ M determined from NMR at 298 K.

^gFrom a van 't Hoff analysis of the temperature‐dependence of the K_d for dimerization, excluding points at 15 and 35 C (Figure 5(b)). When all points are included the values are: ∆H_d = −15 and T∆S_d = −19 kcal/mol.

We note that in the study where low fractions of cis isomers were found in unfolded states, none of the proteins studied (α‐synuclein, pressure‐unfolded eHSP27, and ubiquitin) have a cis proline in their folded states. ²⁰ According to published PDB‐based preferences ²¹ the cis proline conformation in the HlyIIC segment Gly404‐P405‐Y406, should be strongly favored by both the succeeding tyrosine (p = 1.75) and by the preceding glycine (p = 1.46). The glycine in particular should more easily accommodate a strained cis proline due to its lack of a bulky sidechain. This suggests that the high fraction of 20%–25% cis conformers in unfolded HlyIIC is largely driven by local interactions involving residues neighboring P405. This is supported by the similarities in ∆H and T∆S for proline isomerization between the folded and unfolded states in WT‐HlyIIC (Table 2), together with the NMR structures that show that the differences between the cis and trans states of folded WT‐HlyIIC involve primarily the proline‐harboring loop (Figure 1).

2.3. Energetics of cis/trans isomerization are weakly coupled to the unfolding stability of WT‐HlyIIC

Isomerization of the unique proline at position 405 of WT‐HlyIIC, gives rise to four coupled states: Fc = folded cis, Ft = folded trans, Uc = unfolded cis, and Ut = unfolded trans, connected by the thermodynamic linkage cycle in Figure S8. ²² The free energy changes for transitions between the states are connected by:

$${\text{∆G}}_{\text{Ft}/\mathrm{Fc}}+{\text{∆G}}_{\mathrm{Ut}/\text{Ft}}+{\text{∆G}}_{\mathrm{UC}/\mathrm{Ut}}+{\text{∆G}}_{\mathrm{Fc}/\mathrm{UC}}=0,$$ (1)

and the associated equilibrium constants give the product:

$${\mathrm{K}}_1{\mathrm{K}}_2{\mathrm{K}}_3{\mathrm{K}}_4=1.$$ (2)

The equilibrium constant for global unfolding of the protein, measured in equilibrium denaturation experiments (Figure 2(a)), contains unresolved contributions from both cis and trans configurations of the unique proline at position 405.

$${\mathrm{K}}_{\mathrm{U}/\mathrm{F}}=\left(\left[\mathrm{Ut}\right]+\left[\mathrm{Uc}\right]\right)/\left(\left[\text{Ft}\right]+\left[\mathrm{Fc}\right]\right).$$ (3)

FIGURE 2.

Thermodynamics of the WT‐HlyIIC domain. (a) Urea denaturation of WT‐HlyIIC at 25 C followed by CD mean residue ellipticity (MRE) at 220 nm. (b) Portions of ¹H‐¹⁵N HSQC spectra of folded WT‐HlyIIC showing that the signal for the trans form of T383 increases at the expense of the cis form with increasing temperature. (c) van 't Hoff plot for the [trans]/[cis] equilibrium in the folded state of WT‐HlyIIC. The van 't Hoff analysis was done using the peak heights for the cis and trans ¹H‐¹⁵N crosspeaks of residue T383 from (b). Similar analyses for 12 additional residues that give resolved NMR crosspeaks for the cis and trans states in folded WT‐HlyIIC gave consistent results: ∆H_Ft/Fc = 3.4 ± 0.2 kcal/mol, T∆S_Ft/Fc = 3.4 ± 0.3 kcal/mol. HlyIIC, C‐terminal domain corresponding to residues D319‐I412 of the hemolysin II from B. cereus; WT, wild type

From the definitions K₁ = [Ft]/[Fc], K₂ = [Ut]/[Ft], K₃ = [Uc]/[Ut] and K₄ = [Fc]/[Uc] (Figure S8), the equilibrium constant for global unfolding (Equation 3) can be alternatively expressed as ²² :

$$1/{\mathrm{K}}_{\mathrm{U}/\mathrm{F}}=\left({\mathrm{K}}_4{\mathrm{K}}_3+\left[1/{\mathrm{K}}_2\right]\right)/\left({\mathrm{K}}_3+1\right).$$ (4)

Experimentally, K₁ can be measured from the ratio of trans/cis NMR peak intensities in the folded state. K₃ can be measured from the ratio of cis/trans NMR peak intensities in the unfolded state. With K_U/F determined from equilibrium circular dichroism (CD) denaturation measurements, this leaves a solvable system of two equations (Equations (2) and (4)) with two unknowns (K₄ and K₂). Thermodynamic parameters for the transitions between the four states of WT‐HlyIIC (Fc, Ft, Uc, and Ut) were thus calculated from Equations (1), (2), (3), (4)), and are given in Table 2.

The global stability of WT‐HlyIIC was determined from an equilibrium urea denaturation titration monitored by CD (Figure 2(a)). The change in free energy of unfolding, ∆G_U/F = 2.95 kcal/mol, indicates that the domain is marginally stable. Consistently, the thermal stability of the protein is also relatively low with a midpoint of thermal unfolding at 52°C. ¹⁴ The m‐value of 1.33 kcal/mol•M, obtained from urea denaturation, reflects the change in accessible surface area between the folded and unfolded states, and is consistent with that calculated from the NMR structure of HlyIIC using an m‐value prediction server http://best.bio.jhu.edu/mvalue/. ²³

Under standard conditions (298 K or 25 C) the concentrations for the cis and trans states in folded WT‐HlyIIC are nearly equal (Figure 2(b)). Hence ∆G_Ft/Fc, the free energy change between the isomers in the folded state is zero (Table 2). With increasing temperature, the population of the trans state increases slightly at the expense of the cis state (Figure 2(b)). A van 't Hoff analysis of the NMR data for the temperature‐dependence of the equilibrium relating the cis and trans states in folded WT‐HlyIIC (Figure 2(c)) gave ∆H_Ft/Fc = 3.46 kcal/mol and T∆S_Ft/Fc = 3.54 kcal/mol (Table 2). The cis state is thus enthalpically favored, suggesting more favorable bonding interactions occur in the cis compared to the trans state.

The change in entropy for the cis to trans reaction is positive T∆S_Ft/Fc = 3.54 kcal/mol (Figure 2(c), Table 2) indicating that the trans state is entropically favored. The S² order parameter data (Figure S5) indicate that the two states have very similar flexibilities, so differences in protein dynamics are unlikely to account for the change in entropy. Solvation differences are another important factor to consider in determining the entropy change between the two states. Solvent accessibilities for the cis and trans NMR structures were calculated with ProtSA server ²⁴ (http://webapps.bifi.es/ProtSA/) and the results are show in Figure S9. A distinct feature of the plots is that the proline‐harboring loop has more buried nonpolar surface area in the trans than in the cis state (Figure S9B). As the proline‐harboring loop moves away from the core of the structure in the cis state (Figure 1(c)), it causes the nonpolar segment P405‐Y406‐I407 to become more solvent exposed. The trans state appears to be entropically favored because it buries a greater amount of hydrophobic surface area, consequently increasing solvent disorder.

The cis proline of HlyIIC is as an intrinsically strained conformation, maintained by the totality of interactions involving solvent and protein, that stabilize the tertiary structure of the domain. When the stability of the folded structure is lowered, for example at high temperature, the bonding forces maintaining the strained cis conformation are relaxed to allow a higher proportion of the trans state. Similar coupling of cis/trans proline isomerization to global stability, was previously reported for staphylococcal nuclease, where the dominant (80%) strained cis conformation ²⁵ is compensated by stabilizing interactions from the rest of the protein. ²² While the folded state of HlyIIC stabilizes the cis form of P405 to a fraction of ~50%, the effect is small since the cis form accounts for ~25% of the unfolded state. If the recent finding that typical cis prolines in unfolded proteins occur at a fraction of ~5% ²⁶ is general, we can conclude that about half of the stability of the cis isomer in HlyIIC is due to the folding of the protein and half is due to localized effects from neighboring residues in the sequence, since the cis form is already substantially populated in the unfolded state.

2.4. A minor dimeric state in the P405M and P405A mutants

The initial NMR structure of HlyIIC was solved using the P405M‐HlyIIC mutant, since it decreased spectral complexity by eliminating the cis form. ¹⁴ Nevertheless, at high‐protein concentrations, we observed a second minor form in NMR spectra of P405M‐HlyIIC and P405A‐HlyIIC, mutants designed to eliminate the cis state. Studies as a function of protein concentration and temperature showed that this second minor form is a dimer. The initial NMR structure of P405M‐HlyIIC was thus determined at low temperature and protein concentration to suppress the dimeric form. ¹³ , ¹⁴ In this work, working at higher temperature and protein concentration, partial NMR assignments were obtained for ~25% of the dimer signals in P405M‐HlyIIC, as shown in Figure 3(a). We did not detect chemical exchange between the monomer and dimer NMR resonances in ZZ‐exchange spectroscopy experiments (not shown). Therefore, the monomer and dimer states of P405M‐HlyIIC are in slow exchange on both the chemical shift and T1 NMR timescales, and have distinct magnetic properties that are not averaged by exchange. Using the assigned monomer and dimer¹H‐¹⁵N HSQC crosspeaks (Figure 3(a)), we measured¹⁵N R2 relaxation data for the major and minor form of P405M‐HlyIIC (Figure 3(b)). The R2 value of 11.6 s⁻¹ for the major form is consistent with isotropic rotational diffusion of a monomer with a comparable size to P405M‐HlyIIC (10.8 KDa). The R2 value for the minor form of 18.5 is consistent with a dimer of P405M‐HlyIIC. To further characterize the two forms, we performed BPPSTE‐HSQC NMR pulse‐field gradient (PFG) experiments ²⁷ to investigate the translational diffusion of the two states (Figure 3(c)). Under the conditions of the experiment, the diffusion constant of the minor form is ~1.5x slower compared to the major form. From an empirical equation relating the diffusion of folded proteins to their sizes published by Wilkins et al. (Equation (4) of Wilkins et al., ²⁸ ), we can estimate that a doubling in size of P405M‐HlyIIC from 94 amino acids (10.8 KDa) to 188 amino acids (21.6 KDa), should decrease the translational diffusion rate by a factor of 1.2x, close to the values observed in Figure 3(c). Thus, the¹⁵N R2 relaxation data, the translational diffusion data, and the concentration dependence of intensities for minor and major form NMR signals, are all consistent with a dimeric oligomerization state for the minor form of P405M‐HlyIIC. Since we detected only one set of resonances for each residue in the P405MA‐HlyIIC minor form, the minor state corresponds to a dimer with C2 symmetry.

FIGURE 3.

Dimerization of P405M‐HlyIIC by NMR (a) ¹H‐¹⁵N HSQC spectra of P405M showing assigned correlations form the monomer (green) and dimer (purple) states. The spectrum was obtained for a 0.8 mM P405M‐HlyIIC sample in 20 mM NaH₂PO₄ buffer at pH 6.6, and a temperature of 33 C. The blue arrow indicates the well‐resolved monomer and dimer signals for residue T383. (b) Comparison of ¹⁵N transverse relaxation (R2) rates for the HlyIIC‐P405M monomer and dimer. (c) Comparison of translational diffusion constants determined by NMR PFG techniques (see methods) for the HlyIIC‐P405M monomer and dimer. HlyIIC, C‐terminal domain corresponding to residues D319‐I412 of the hemolysin II from B. cereus; PFG, pulsed field gradient

2.5. The NMR structure of the P405M‐HlyIIC minor form is a strand‐β5 domain‐swapped dimer

The largest chemical shift differences between the monomer and dimer occur in the P405‐bearing loop between helix αB and strand β5 (Figure S10) suggesting that the largest structural differences between the monomer and the dimer involve residues near position 405, which is mutated in P405M‐HlyIIC. To better characterize the structural properties of the dimer we collected a 3D¹³C‐F1‐filtered, ¹³C‐F3‐edited NOESY‐HSQC experiment ²⁹ recorded on a P405M‐HlyIIC sample comprised of a 1:1 mixture of ¹²C¹³:C protomers (Figure S11). In this experiment, the asymmetric labeling ensures that NOE distance contacts are only seen between monomers. ²⁹ Most of the 12 interchain NOEs that we were able to assign in this experiment involved contacts between W372‐V376 from strand β3 and I407‐I412 from strand β5. In the previously determined P405M‐HlyIIC NMR structure, strands β3 and β5 form a nonsequential hydrogen‐bonded β‐sheet interaction in the monomers. ¹⁴ The interchain NOE contacts observed in the¹³C‐filtered NOESY data for P405M‐HlyIIC at high‐protein concentration, suggests that strand β5 is domain‐swapped in the dimer.

A structural model of the P405M‐HlyIIC dimer based on dihedral angle constraints as well as intrachain and interchain NOE constraints (Table 1) is shown in Figure 4. Compared to the P405M‐HlyIIC monomer structure (Figure 4(a)), strand β5 extends out to β‐sheet pair with strand β3 from the neighboring protomer (Figure 4(b)). While the protomers are well defined in the NMR model of the dimer (Figure 4(c)), the relative orientation of the protomers is less precise (Figure 4(d), Table 1) since these are constrained by only 12 assigned interchain NOEs.

FIGURE 4.

NMR structure of the P405M‐HlyIIC dimer. (a) NMR structure closest to the ensemble average for the previously reported P405M‐HlyIIC monomer structure. ¹⁴ (b) NMR structure closest to average for the P405M‐HlyIIC domain‐swapped dimer structure described in this work (Table 1). Strand β5 (red in the A protomer, purple in the B protomer) is labeled, and the positions of M405, which is substituted for the WT proline in the P405M‐HlyIIC mutant is indicated by white spheres at the Cα carbons. (c) Global superposition of the 25 lowest‐energy NMR conformers to illustrate the precision of the dimer structure. (d) Superposition on the A protomers of the dimer show that while the protomer structures are well defined, the overall dimer structure is less precise due to a limited number of interchain NOEs. The first 11 disordered residues in each structure are not shown. HlyIIC, C‐terminal domain corresponding to residues D319‐I412 of the hemolysin II from B. cereus

A key difference between the monomer and dimer states involves the P405‐bearing loop L _αBβ5. In the monomer, the loop undergoes a sharp hairpin‐like chain reversal near the substituted residue M405 to allow intramolecular antiparallel β‐sheet pairing of strands β3 and β5 (Figure 4(a)). By contrast in the dimer, the loop extends outwards to allow swapping of strand β5 between monomers (Figure 4(b)). The backbone dihedral angles of M405 are different, falling near the α‐helical region of Ramachandran space in the monomer but near the extended β‐sheet region in the dimer. Domain‐swapped dimers often occur due to changes in the hinge regions that connect the interchanged domains. ³⁰ , ³¹ The differences in the L _αBβ5 loop between monomer and dimer, suggest that P405, the site of cis/trans isomerization is connected to dimerization.

2.6. P405 in the sequence of WT‐HlyIIC strongly disfavors the domain‐swapped dimer

The P405M‐HlyIIC dimer is favored at high temperature and protein concentration. To characterize dimerization thermodynamics, we used NMR to follow the dilution of P405M‐HlyIIC at six temperatures ranging between 15 and 35 C. Monomer and dimer concentrations under each set of conditions were calculated ³² , ³³ from the peak intensities of the well‐resolved ¹H‐¹⁵N HSQC crosspeaks for residue T383 (indicated by the blue arrow in Figure 3(a)). The concentration dependence of the monomer and dimer peak intensities was used to calculate the K_d value at each temperature (Figure 5(a)) as described in Materials and Methods. A van 't Hoff analysis of the temperature‐dependence of K_d (Figure 5(b)) gave the changes in enthalpy and entropy associated with the dissociation of dimers into monomers (Table 2). At a temperature of 25°C the K_d for P405M‐HlyIIC is 1.3 mM, and the associated change in free energy for dissociation is ∆G_d = 3.9 kcal/mol. The K_d decreased with increasing temperature as the fraction of dimer increases at higher temperature, similar to two other literature examples. ³⁴ , ³⁵ Dissociation gives a decrease in entropy T∆S = −19 kcal/mol (unfavorable), and a decrease in enthalpy ∆H_d = −15 kcal/mol (favorable). The van ‘t Hoff plot has apparent curvature (Figure 5(b)), typical of a reaction with a nonzero negative heat capacity term ³⁵ that is characteristic of dimerization reactions dominated by the hydrophobic effect. ³⁵ , ³⁶ , ³⁷ In our analysis we used a linear fit of the data due to the potential errors for the points at the temperature extremes. At 15°C precise measurements of the K_d are difficult as the dimeric state is suppressed by low temperature. At 35 C the thermal denaturation transition of the mutant is approached. ¹⁴

FIGURE 5.

Thermodynamics of HlyIIC dimerization. (a) Plots of dimer versus monomer concentrations for P405M‐HlyIIC determined from the NMR crosspeaks of residue T383 according to Equation (7). The slopes of the data were used to obtain K_d values at each temperature according to Equation (8). (b) van 't Hoff analysis of the temperature‐dependence of K_d using the data from (A). Only the linear part of the data were used, excluding the points at 15 and 35 C. The resulting ∆H and ∆S values are close to those when all temperatures were used (Table 2). (c) Portions of ¹H‐¹⁵N HSQC spectra of P405M, P405A, and WT‐HlyIIC. Peaks from monomers are labeled green, those from dimer purple, the cis state orange, the trans state cyan, and peaks where conformational states cannot be resolved are labeled black. The downfield peak D* is assigned to the dimer based on its disappearance at low protein concentration. Although a site‐specific assignment for the D* peak is not available, it is well resolved and provides a useful handle of the amount of protein in the dimer state. While the spectra for the P405M and P405A mutants were obtained for 0.8 mM protein concentrations in about 20 min, the spectrum of WT‐HlyIIC was recorded at a 9 mM concentration using a total acquisition time of 13 h. The weak relative intensity of the D* peak attests to the diminished propensity of HlyIIC to dimerize when the WT proline occurs at position 405. HlyIIC, C‐terminal domain corresponding to residues D319‐I412 of the hemolysin II from B. cereus; WT, wild type

P405A‐HlyIIC, a mutant where the proline is changed to a smaller alanine sidechain, has very similar structural and stability properties to P405M‐HlyIIC. ¹⁴ The alanine substitution also eliminates the cis form, and shows a minor dimer form at high‐protein concentrations and temperatures (Figure 5(c)). By contrast, we reported that WT‐HlyIIC, where a proline occurs at position 405, shows no NMR signals characteristic of the dimer at concentrations as high as 2 mM. ¹⁴ The NMR results are qualitatively consistent with size exclusion chromatography data that show partially resolved monomer and dimer peaks for the mutants but only a monomer peak for WT‐HlyIIC at 0.5 mM protein concentrations and a temperature of 28 C (Figure S12).

If only the trans state were capable of dimerizing, one would expect a 50% increase in the K_d of WT‐HlyIIC compared to P405M‐HlyIIC because 50% of the WT protein is in cis form and thus unable to form the dimer. The 50% cis population would reduce the effective concentration of WT‐HlyIIC by half. In fact, dimerization of WT‐HlyIIC is suppressed much more strongly. To examine if WT‐HlyIIC can form a dimer, we collected NMR data at an extremely high‐total protein concentration of 9 mM. After acquiring a ¹H‐¹⁵N HSQC spectrum for 13 h (compared to 20 min for P405A‐HlyIIC and P405M‐HlyIIC) we detected a crosspeak marked D*, that while not specifically assigned to a residue in the HlyIIC sequence, occurs in a highly characteristic position at about 10.6 ppm (¹H) and 130 ppm (¹⁵N). The peak can be attributed to the dimer based on its disappearance at low protein concentrations or temperature in the mutant spectra. From the intensities of the dimer and monomer peaks we calculate K_d using: ³²

$${\mathrm{K}}_{\mathrm{d}}={\left[\mathrm{M}\right]}^2/\left[\mathrm{D}\right]=\left({\text{2P}}_{\mathrm{t}}{\left({\mathrm{ƒ}}_{\mathrm{m}}\right)}^2\right)/{\mathrm{ƒ}}_{\mathrm{d}}=\left({\text{2P}}_{\mathrm{t}}{\left({\mathrm{ƒ}}_{\mathrm{m}}\right)}^2\right)/\left(1-{\mathrm{ƒ}}_{\mathrm{m}}\right),$$ (5)

where [M] is the monomer concentration, [D] is the dimer concentration, [P_t] is the total protein concentration, and ƒ_m and ƒ_d are the fractions of monomer and dimer states, respectively. We thus obtained the following K_d values at 25 C: 0.9 mM for P405M‐HlyIIC, 1.4 mM for P405A‐HlyIIC, and 43.3 mM for WT‐HlyIIC. Therefore, the K_d for WT‐HlyIIC is raised at least 30‐fold compared to mutants that substitute another sidechain for the proline. We view this K_d value for WT‐HlyIIC as a lower estimate. If the D* peak is mis‐assigned, the K_d of the WT could be larger, or the WT protein may not form a dimer at all.

3. DISCUSSION

Domain‐swapped oligomerization is usually driven by changes in the hinges between domains, while the structure of the swapped segments is maintained compared to the monomers. ³⁰ , ³¹ Changes promoting domain‐swapped oligomer formation can include lengthening or shortening of hinge regions, changes in electrostatic interactions, or conformational changes induced by binding of metals. ³⁰ Hinge loops have been proposed to act like “molecular springs,” releasing strain by changing conformation within the context of an otherwise unchanged structure. ³⁸ Prolines are enriched in the sequences of domain‐swapped dimer hinge regions, ³⁹ and mutations of these residues has been shown to affect the dimerization equilibrium of a number of domain‐swapped proteins. ³¹ , ³⁸ , ³⁹ , ⁴⁰ , ⁴¹

In some cases, the presence of proline residues rigidify the chain and favor more extended hinge conformations that can lend themselves to domain swapped oligomers. ³⁹ Conversely, substitution of the WT proline by a more flexible alanine in the P90A mutation of the protein suc1 raises the K_d, favoring the monomer. ³⁸ A second example where replacing a proline in a domain‐swapped hinge with a more flexible residue favors the monomer is P51G in cyanovirin‐N. ⁴⁰ In other cases, such as the P92A mutation in the same hinge region of the protein suc1, ³⁸ the P114G and P114A mutations of the cis proline in ribonuclease A, ⁴¹ and the P405M and P405A mutants of HlyIIC in the present work, replacement of a WT proline by another residue favors a domain‐swapped dimer state. A possible explanation for these cases, is that replacement of proline by another residues allows greater flexibility for the hinge, enabling it to “reach” to the neighboring monomer to form the domain‐swapped structure.

In examples where domain‐swapped dimer formation is disfavored by a proline, the residue acts as a “gatekeeper” to avoid aggregative misfolding of the protein. ⁴¹ The presence of P405 in HlyIIC and the conformational heterogeneity associated with isomerization between its cis and trans states may therefore be a consequence of the residue ¹ stabilizing the protein (the WT has a T_m of 52°C compared to 47°C for either of the P405M and P405 mutants), ^{14 , 2} and preventing aggregative misfolding via a domain‐swapped dimer structure. A Protein BLAST ⁴² search of the HlyIIC sequence against the nonredundant protein sequence database indicates proline 405 is conserved in homologous toxins from other Bacillus species.

4. MATERIALS AND METHODS

4.1. NMR spectroscopy

Samples of the 94 residue (10.8 kDa) HlyIIC domain (corresponding to residues D319‐I412 of HlyII), isotopically enriched with¹³C and/or¹⁵N were prepared as previously described. ¹³ , ¹⁴ Unless otherwise noted, NMR samples contained 1.0 mM WT‐HlyIIC or 0.4 mM P405M‐HlyIIC in 20 mM sodium phosphate buffer at pH 6.0, with 1 mM EDTA, 1 mM PMSF, and 0.05% wt/vol sodium azide. Data were collected on 800 and 600 MHz Varian INOVA NMR instruments with cryogenic probes. NMR assignments for WT‐HlyIIC were obtained by performing separate sequential walks for the cis and trans state using heteronuclear 3D NMR spectra. ¹³ NMR assignments of the minor dimeric form of P405M‐HlyIIC were obtained on a 0.8 mM sample of P405M‐HlyIIC at 30°C to favor dimerization. Heteronuclear NMR experiments used for dimer assignments included 3D HNCACB, HNCO, ¹⁵N‐TOCSY (60 ms mixing time), and ¹⁵N‐NOESY (125 ms mixing time).

4.2. ZZ‐exchange spectroscopy to characterize WT‐HlyIIC proline cis/trans isomerization

ZZ‐exchange spectroscopy ¹⁷ was used in an attempt to measure the kinetics of cis/trans isomerization for folded WT‐HlyIIC (Figure S2). To test the experiment, we collected data on a 1.1 mM sample of a L25G mutant staphylococcal nuclease, at 30°C, where the resonances corresponding to the cis and trans forms of P117 are about equal. ²² As expected, the experiment showed clear magnetization transfer crosspeaks mediated by chemical exchange between the cis and trans forms of P117 in staphylococcal nuclease. ²² The same experiment performed on a 0.85 mM sample of ¹⁵N labeled WT‐HlyIIC in 20 mM sodium phosphate buffer at 37°C using mixing times of 0.42, 0.63, 0.84 s showed no exchange, indicating that cis/trans isomerization is slower than the mixing times. To increase the rate of cis/trans proline interconversion for WT‐HlyIIC, we used a 20 μM concentration of human cyclophilin A (Novus Biologicals, Littleton Colorado), a peptidyl prolyl isomerase known to catalyze proline isomerization. ⁴³ ZZ‐exchange experiments with mixing times of 0.42 s or 0.63 with WT‐HlyIIC in the presence of 20 μM cyclophilin A, also showed no exchange further attesting to slow proline isomerization in folded WT‐HlyIIC. Similarly, we observed no exchange between monomer and dimer for the P405M‐HlyIIC mutant, indicating slow exchange between these states as well.

4.3. NMR structure determination for cis and trans states of WT‐HlyIIC and the P405M‐HlyIIC dimer

Distance constraints for cis and trans WT‐HlyIIC were obtained from ¹³C and ¹⁵N‐edited NOESY‐HSQC experiments ⁴⁴ collected with 125 ms mixing times. Backbone (φ and ψ), and sidechain (χ₁) dihedral angles were obtained from the programs DANGLE, ⁴⁵ and TALOS‐N ⁴⁶ , ⁴⁷ using assigned ^NH, Hα, N, Cα, Cβ, and C′ chemical shifts. Hydrogen bonds, for cis and trans WT‐HlyIIC were determined from a long‐range HNCO experiment, as described previously. ¹⁴ NMR structures were calculated using Xplor‐NIH ⁴⁸ and were subsequently refined in an explicit water shell using ARIA 2.3 ⁴⁹ on the NMRbox platform. ⁵⁰

For the P405M‐dimer structure we identified interchain distance contacts from a 3D¹³C‐F1‐filtered, ¹³C‐F3‐edited NOESY‐HSQC experiment on a sample composed of a 1:1 mixture of ¹³C‐labeled and unlabeled P405M‐HlyIIC (total concentration 0.8 mM) at a temperature of 37°C. The mixing time for the NOESY experiment was 200 ms. To calculate the dimer structure the inter‐chain distance restraints from the aforementioned experiment were supplemented with intrachain NOESY data from 3D ¹⁵N and ¹³C‐edited NOESY‐HSQC experiments recorded with mixing times of with 150 and 100 ms respectively, ¹⁴ and dihedral angles calculated with the TALOS‐N program from the assigned chemical shifts of the dimer. The structures were calculated with the Xplor‐NIH ⁴⁸ program, including non‐crystallographic and collinear intermolecular restraints to maintain dimer subunit symmetry.

Restraints and statistics for the trans, cis, and dimer NMR structures are summarized in Table 1. For each state, the 25 lowest‐energy nonviolated NMR structures were deposited in the Protein Data Bank with PDB accession codes 6D53 for trans WT‐HlyIIC, 6D5Z for cis WT‐HlyIIC, and 6WA1 for the P405‐HlyIIC dimer. Structure 1 is the closest to the coordinate mean in each NMR ensemble. Corresponding BMRB accession codes are: 19,462, 19,461, and 30,737, respectively.

4.4. NMR relaxation experiments

Backbone dynamics of the cis and trans states of WT‐HlyIIC were investigated using ¹⁵N longitudinal (R₁), transverse (R₂), and cross‐relaxation (¹H‐¹⁵N NOE) experiments (Figure S4). All relaxation data was acquired at a field strength of 800 MHz and 15°C, for comparison with previous values for the P405M‐HlyIIC mutant in its monomer form. ¹⁴ R₁ values were obtained using interleaved relaxation delays of 0.05, 0.13, 0.21, 0.49, 0.57, 0.71, and 0.99 s. R₂ values were determined using interleaved relaxation delays of 0.01, 0.03, 0.05, 0.07, 0.09, 0.11, 0.15, 0.25 s. A 2 s pre‐acquisition delay was used for recovery to thermal equilibrium. The relaxation data were processed as previously described, ⁵¹ and a Model Free analysis ¹⁹ to extract S ² order parameters (Figure S5) was performed with the program TENSOR2. ¹⁸

Experiments to compare ¹⁵ N transverse relaxation rates (R2s) between the monomer and dimer states of P405M‐HlyIIC were performed at 600 MHz field strength on a 1 mM protein sample at 33°C. The high temperature and protein concentration were selected to favor dimerization. ¹³ , ¹⁴ R2 relaxation delays of 0.01, 0.03, 0.05, 0.07, 0.09, 0.11, 0.15, and 0.25 s were sampled.

4.5. Characterization of aggregation in P405M‐HlyIIC by PFG diffusion methods

PFG diffusion data were collected on a 0.67 mM sample of ¹⁵N‐labled P405M‐HlyIIC in 90% H₂O/10% D₂O, at a temperature of 33°C. Translational diffusion was characterized using a ¹⁵N‐edited sensitivity‐enhanced 3D DOSY pulse program called BPPSTE‐HSQC (Bipolar Pulse Pair STimulated Echo–HSQC). ²⁷ The experiment determines the diffusion coefficient (in the third dimension) associated with each crosspeak of a 2D¹H‐¹⁵N HSQC experiment. Diffusion rates were measured using an array of seven gradients with values of 2.2, 22.1, 31.1, 38.1, 44.0, 49.2, and 53.9 G/cm. The gradients had a duration δ = 2 ms, a diffusion delay ∆ = 50 ms, and a delay between bipolar gradient pairs σ = 0.2 ms. The diffusion‐encoded ¹H‐¹⁵N HSQC spectra were collected in an interleaved manner to minimize sample fluctuations during the ~1 day acquisition time. Diffusion coefficients were determined by the Stejskal‐Tanner formula ⁵² modified for the BPPSTE‐HSQC experiment (Equation (2) of Rajagopalan et al. ²⁷ ).

4.6. NMR of urea‐unfolded WT‐HlyIIC

To characterize cis/trans isomerization in unfolded WT‐HlyIIC, a 1.2 mM ¹³C,¹⁵N double‐labeled sample was incubated in 6 M urea at 4°C for 2.5 h prior to NMR experiments. The ¹H‐¹⁵N HSQC spectra confirmed the protein was unfolded, with resonances concentrated in the random coil region of the spectrum between 7.5 and 8.7 ppm. Partial assignments of unfolded WT‐HlyIIC (Figure S6A) were obtained from 3D HNCACB, and HNCO spectra collected at a temperature of 25°C on a 600 MHz NMR machine. A van 't Hoff analysis of the cis/trans isomerization equilibrium in the unfolded state (Figure S6B), was obtained from peak integration of cis and trans state resonances from residue G404 in ¹H‐¹⁵N HSQC spectra of 6 M urea‐denatured WT‐HlyIIC at temperatures of 15, 20, 25, 30, and 37°C.

4.7. NMR determination of K_d and thermodynamic parameters for P405M‐HlyIIC dimerization

Dimerization of P405M‐HlyIIC was characterized by ¹H‐¹⁵N HSQC NMR spectra of five samples containing total protein concentrations between 0.2 and 1.0 mM. The total protein concentration is the sum of the molar concentrations of monomer and dimer:

$$\left[{\mathrm{P}}_{\mathrm{t}}\right]=\left[\mathrm{M}\right]+2\left[\mathrm{D}\right].$$ (6)

We used the NMR crosspeak heights of the well‐resolved monomer and dimer signals from residue T383 (blue arrow in Figure 3(a)) to determine monomer and dimer concentrations ³² , ³³ :

$$\left[\mathrm{M}\right]=\left[{\mathrm{P}}_{\mathrm{t}}\right]{\mathrm{ƒ}}_{\mathrm{m}}/\left({\mathrm{ƒ}}_{\mathrm{m}}+{\mathrm{ƒ}}_{\mathrm{d}}\right)\text{and}\left[\mathrm{D}\right]=\left[{\mathrm{P}}_{\mathrm{t}}\right]{\mathrm{ƒ}}_{\mathrm{d}}/2\left({\mathrm{ƒ}}_{\mathrm{m}}+{\mathrm{ƒ}}_{\mathrm{d}}\right).$$ (7)

Since the dissociation constant is:

$${\mathrm{K}}_{\mathrm{d}}={\left[\mathrm{M}\right]}^2/\left[\mathrm{D}\right],$$ (8)

its value can be determined as the slope in a plot of [D] (x‐axis) versus [M]² y‐axis (Figure 5(a)) as the protein is diluted from 1 to 0.2 mM. ⁴⁰

To determine thermodynamic parameters, we characterized the reversible dimerization equilibrium of P405‐HlyIIC at six temperatures (15, 20, 25, 30, 33, 35 C) using the monomer and dimer signals of residue T383 as described above. A van 't Hoff analysis of the temperature dependence of K_d (Figure 5(b)) was used to obtain ∆H and T∆S values for dimerization (Table 2).

4.8. Equilibrium measurements of the stability of WT‐HlyIIC to unfolding by CD

A Chiroscan V100 instrument from Applied Photophysics (Leatherhead, UK) with a large area avalanche photodiode (LAAPD) detector, was used to collect CD spectra. The signal at 220 nm was averaged over 20 scans for each sample, using a slit width of 2 nm, and a 10 mm cell pathlength. Solutions of 10 μM WT‐HlyIIC in 20 mM sodium phosphate buffer, pH 6.0 were prepared with urea concentrations ranging from 0 to 6 M, in 0.1 M increments. A set of control 0–6 M urea samples in buffer without protein, were prepared in the same fashion for baseline measurements. All samples were incubated overnight at 25°C. Urea concentrations were verified using a refractometer (Carl Zeiss, Germany). ΔG_U/F values were calculated using nonlinear least‐squares fits of the denaturation data (Figure 2(a)), as described previously. ⁵³ , ⁵⁴

AUTHOR CONTRIBUTIONS

Annie R. Kaplan: Conceptualization; data curation; formal analysis; funding acquisition; investigation; methodology; project administration; resources; supervision; validation; visualization; writing‐original draft; writing‐review & editing. Rich Olson: Investigation; resources; supervision; writing‐review & editing. Andrei T. Alexandrescu: Conceptualization; data curation; formal analysis; funding acquisition; investigation; methodology; project administration; resources; supervision; validation; visualization; writing‐original draft; writing‐review & editing.

CONFLICT OF INTEREST

The authors declare no potential conflict of interest.

Supporting information

Appendix S1: Supporting information

Kaplan AR, Olson R, Alexandrescu AT. Protein yoga: Conformational versatility of the Hemolysin II C‐terminal domain detailed by NMR structures for multiple states. Protein Science. 2021;30:990–1005. 10.1002/pro.4066