Peptide backbone circularization enhances antifreeze protein thermostability

Abstract

Antifreeze proteins (AFPs) are a class of ice‐binding proteins that promote survival of a variety of cold‐adapted organisms by decreasing the freezing temperature of bodily fluids. A growing number of biomedical, agricultural, and commercial products, such as organs, foods, and industrial fluids, have benefited from the ability of AFPs to control ice crystal growth and prevent ice recrystallization at subzero temperatures. One limitation of AFP use in these latter contexts is their tendency to denature and irreversibly lose activity at the elevated temperatures of certain industrial processing or large‐scale AFP production. Using the small, thermolabile type III AFP as a model system, we demonstrate that AFP thermostability is dramatically enhanced via split intein‐mediated N‐ and C‐terminal end ligation. To engineer this circular protein, computational modeling and molecular dynamics simulations were applied to identify an extein sequence that would fill the 20‐Å gap separating the free ends of the AFP, yet impose little impact on the structure and entropic properties of its ice‐binding surface. The top candidate was then expressed in bacteria, and the circularized protein was isolated from the intein domains by ice‐affinity purification. This circularized AFP induced bipyramidal ice crystals during ice growth in the hysteresis gap and retained 40% of this activity even after incubation at 100°C for 30 min. NMR analysis implicated enhanced thermostability or refolding capacity of this protein compared to the noncyclized wild‐type AFP. These studies support protein backbone circularization as a means to expand the thermostability and practical applications of AFPs.

Introduction

Antifreeze proteins (AFPs) are produced by many overwintering organisms such as fish and terrestrial insects to protect them from the damages imposed by freezing of body fluids.1, 2 AFPs function by binding to small ice crystals and depressing the freezing temperature of surrounding water to below the melting temperature; a property termed thermal hysteresis (TH). As a result, surrounding water remains liquid at lower temperatures and further ice growth is suppressed. This has led to use of AFPs in the frozen food industry for preserving product texture and improving cold storage3, 4, 5, 6 and for assessing AFPs as kinetic hydrate inhibitors to prevent the build‐up of gas hydrate in oil pipelines.7, 8, 9

A current limitation to AFP use in industrial or biomedical applications is their vulnerability to denaturation upon exposure to elevated temperatures or nonphysiological concentrations of salt, organic solvents, urea, or other chemical agents during commercial processing.10, 11 To mitigate this problem, protein engineers are investigating the ways to improve the robustness of protein to denaturation, such as increasing fold‐stabilizing electrostatic interactions through site‐directed mutations.12, 13 Another approach attracting interest among industry and biotechnology developers is protein circularization.14 The head‐to‐tail or end‐to‐end peptide backbone structure of circular proteins enhances conformational stability and affords resistance to exopeptidases and heat degradation that would otherwise affect proteins with free N and C termini.15

In comparison to the more formidable chemical synthesis of circular proteins, intein‐mediated trans‐splicing has proven to be a relatively straightforward and highly effective means to achieve this as it simply involves fusing N‐intein and C‐intein fragments to the C and N termini, respectively, of the protein to be circularized.16 After translation of this precursor protein, the two intein fragments associate with high affinity to form a functional intein that cleaves itself from the precursor protein, simultaneously ligating the ends of the nested peptide (extein) via a peptide bond.17 The well‐characterized, naturally occurring split‐intein from Nostoc punctiforme DnaE (catalytic subunit α of DNA polymerase III) is now regarded as the “golden standard” trans‐splicing system as it can be expressed in E. coli.18 It requires no external energy or cofactors, allows for variation of the extein sequence, and was recently used to increase thermostability of the plant cell wall‐degrading enzyme, xylanase, presumably by reducing the conformational entropy of the unfolded state of the enzyme.19

To assess the feasibility of backbone circularization and any benefits toward AFP thermostability, we used the Npu DnaE split‐intein to ligate the N‐ and C‐termini of the small (7 kDa), moderately active type III AFP from the North‐Atlantic ocean pout Macrozoarces americanus (isoform QAE1 HPLC12) as a model system (Fig. 1).20, 21 This AFP is able to depress the freezing point of a solution of ice and prevent ice recrystallization. Based on the latter property, it is a common additive in frozen food products such as ice cream and frozen dough.3 The structure of type III AFP isoform QAE1 HPLC12 has been solved by both X‐ray crystallography22, 23 and NMR.20, 21 It exhibits a compact globular structure composed of nine short β‐strands and three 3₁₀‐helices that are connected by numerous β‐turns and coils to form two internal tandem motifs. Arrangement of these motifs about a pseudo‐dyad symmetry gives this AFP a “pretzel” or “beta‐clip” fold.24, 25 The ice‐binding site (IBS) of this type III AFP isoform encompasses residues Thr18, Leu19, and Val20 that bind to the primary prism plane of ice and linearly dispersed residues like Gln9, Asn14, Ala16, and Gln44 that form a patch binding to a pyramidal plane of ice.24, 26 These residues are solvent‐exposed and form two adjacent, relatively flat and hydrophobic surfaces inclined at an angle of about 150° to each other to form a compound IBS. On the opposing side, the N and C‐termini of type III AFP are 19.8 Å apart, allowing circularization via a short linker. Thus, the reasons for choosing this particular AFP for our study are that it is small, globular, structurally well characterized, has its N and C termini close together, can be recombinantly produced in Escherichia coli, and is in widespread use in the food industry. Also, it is possible to efficiently refold type III AFP after solubilization in 8 M guanidine‐HCl.27, 28 The reason for choosing a QAE1 isoform over the QAE2 or SP isoforms is that the latter are not as active in halting ice growth because they lack one of the two compound IBSs.29 A recent protein engineering study showed that it was possible to convert a QAE2 isoform into a fully active QAE1‐like form by four surface mutations that introduced the second (pyramidal) ice‐binding plane.30

Figure 1.

Type III AFP backbone cyclization mechanism using protein trans‐splicing. (A) Cartoon depiction of type III AFP (PDB ID:1AME³²) with the modeled backbone circularization loop shown in magenta, blue, and orange, which correspond to N‐extein, C‐extein, and linker colors, respectively, as shown in (B) and (C). Secondary structure elements are colored red (helix), green (β‐strand), and yellow (coil). (B) Schematic of the noncyclized type III AFP construct embedded in the Npu DnaE split‐intein. To facilitate cyclization, the N‐extein and C‐extein moieties are fused to the C‐ and N‐termini, respectively, of the polypeptide to be cyclized. The N‐intein is fused to the C‐terminal end of the N‐extein, and the C‐intein is located at the N‐terminal end of the C‐extein. (C) Products of trans‐splicing by split inteins. After assembly of the two intein fragments, a splicing reaction takes place, where the intein removes itself from the precursor protein and simultaneously ligates the exteins together via a peptide bond.

Gratifyingly, the type III AFP‐intein fusion protein showed a high level of expression in E. coli and the majority of the protein completed autosplicing prior to, or during, cell lysis. Moreover, the excised AFP could be recovered by ice‐affinity purification and exhibited ice‐shaping activity similar to the non‐cyclized wild‐type type III AFP. TH measurements showed the circularized AFP (cAFP) retained the ability to suppress ice‐growth following heating to 100°C, whereas the noncyclized wild‐type protein had significantly reduced activity after heating to only 37°C, and no activity after heating to 100°C. Based on two‐dimensional NMR spectra observed at several temperatures, the retention of TH activity of cAFP after extensive heating appears to be attributable to improved thermal stability against denaturation and/or the capacity to recover from a partially unfolded state.

Results

Design of a circularized type III AFP construct

To facilitate the circularization of type III AFP, we designed a vector based on the Npu DnaE split‐intein [Fig. 1B,C)].18 This intein has been shown to be most efficient when paired with a non‐native RGKCWE extein scar sequence.31 However, an examination of two of the three‐dimensional structures of QAE1 type III AFP (PDB 1AME and 1HG7^23, 32) suggested that these additional six residues were not enough to span the distance between the N and C termini without causing strain on the protein's ice‐binding surface. Therefore, we modeled the structures of a small series of circular type III AFPs that each included different flexible linkers (AA, GAA, and GGAA) in combination with the RGKCWE extein scar sequence and monitored trajectories of each protein in molecular dynamics (MD) simulations. As can be seen in Figure 2(A), there was little difference in fluctuations on a per residue basis (root‐mean‐square fluctuation [RMSF] analysis), therefore we based our decision of linker selection on the overall protein root‐mean‐square deviation (RMSD). With reference to the RMSD trajectories represented in Figure 2(B), the trajectory of the RGKCWEGAA linker sequence matches more closely to the trajectory of the wild‐type noncyclized protein than those of the other two linker sequences. The RGKCWEGGAA linker sequence has an initial jump in RMSD before lowering near the end of the simulation. The RGKCWEAA linker sequence, on the other hand, has a relatively low RMSD compared to the wild‐type noncyclized type III AFP then increases at the end of the simulation. Therefore, we reasoned the RGKCWEGAA linker sequence was our best candidate for moving forward to in vitro experiments. On this basis, we cloned the Npu DnaE/RGKCWEGAA intein/extein system as a fusion with the type III AFP gene [Fig. 2(C)].

Figure 2.

MD simulations of wild‐type and three cyclized type III AFP constructs. (A) RSMF values (nm) for the α carbons of wild‐type type III AFP (black line) and cAFPs with GGAA (red), GAA (green), and AA (blue) linkers are plotted against residue number. The 20‐ns simulations were performed on energy‐minimized models of each protein using GROMACS. The solid horizontal line below the graph represents the length of the noncyclized AFP with ice‐binding residues shown by cyan blocks. (B) RMSD values (nm) for all atoms of each AFP construct during the 20‐ns simulations in GROMACS are presented, with averaging of 51 frames to generate a smooth curve. The wild‐type noncyclized type III AFP and cAFP with different linkers are represented by the same colours as in (A). (C) Cartoon models comparing noncyclized AFP to cAFP showing the extein scar (RGKCWE) and the additional flexible linker GAA. Secondary structure elements are colored as in Figure 1.

Purification of circularized type III AFP

The recombinant type III AFP‐intein fusion protein was overexpressed in E. coli BL21(DE3) cells in the soluble fraction, which was subjected to ice‐shell affinity purification [Fig. 3(A)]. After 1 h of extraction of the diluted cell lysate at −1.5°C, the rotary ice‐shell had become cloudy due to protein incorporated into the ice fraction [Fig. 3(Aii)]. Assessment of the liquid and ice fractions by SDS‐PAGE showed enrichment of a protein of less than 6.5 kDa in the ice fraction [Fig. 3(B)]. Subjecting the ice fraction to size‐exclusion chromatography generated four major peaks [Fig. 3(C)], which were further investigated by SDS‐PAGE. Peaks 1, 2, and 3 contained many bands that are likely contaminating E. coli proteins, while peak 4 contained a single species that migrated farther by SDS‐PAGE than the noncyclized type III AFP [Fig. 3(D)]. Similar differences in mobility between circularized and noncyclized proteins have been observed previously,33 suggesting ligation of the N‐ and C‐termini of type III AFP had occurred.

Figure 3.

Purification of circularized type III AFP. (A) Rotary ice‐affinity purification of 1:10 diluted cAFP lysate. (i) Starting ice‐shell; (ii) ice‐shell after 1 h incubation at −1.5°C with cAFP lysate. (B) SDS‐PAGE analysis of ice‐shell purification. Molecular weight standards were electrophoresed in the left‐hand lane. (C) Size‐exclusion purification of ice‐shell‐enriched ice fraction. (D) SDS‐PAGE analysis of size‐exclusion purification fractions of the cAFP with purified noncyclized AFP run alongside for comparison.

Confirmation of circularization by mass spectrometry and NMR spectroscopy

The predicted mass of noncyclized type III AFP is 8.12 kDa based on its amino acid sequence with a hexahistidine tag attached. This matches the 8100 Da mass determined by MALDI‐TOF mass spectrometry for the purified noncyclized III AFP sample (Fig. 4). Likewise, the expected molecular mass of 7845 Da for the cAFP (adjusted from 7863 Da to 18.01 Da to account for the water molecule released during peptide bond formation) matches closely to the experimentally determined mass of 7842 Da, an observation consistent with circularization. As a further confirmation, NMR analysis was performed to verify the sequence where the C‐to‐N terminus linkage was made and to assess the folding state of the cAFP. An overlay of ¹H‐¹⁵N HSQC spectra of the circularized and noncyclized type III AFP constructs [Fig. 5(A)] showed that the two constructs adopted very similar structures in solution, indicating that the tertiary structure of type III AFP was not adversely affected by circularization. Standard triple resonance experiments were collected and analyzed to assign the backbone resonances of cAFP. The chemical shifts of residues in the central region of the wild‐type noncyclized AFP and central sequence of the cAFP are very similar. Their peaks overlay well, with chemical shifts of less than 0.1 ppm, again suggesting the overall structure of the protein has not significantly changed (Supporting Information Figs. S1, S2).

Figure 4.

Mass analysis of circularized and noncyclized type III AFPs. MALDI‐TOF analysis of cAFP (black line) and noncyclized AFP (blue line). Masses are displayed in Daltons. The peak at 15,683.31 Da is double the mass of the cAFP peak.

Figure 5.

NMR analysis of circularized and noncyclized type III AFPs. (A) Overlay of ¹H/¹⁵N‐HSQC spectra of uniformly 13C/¹⁵N‐labeled cAFP (black resonances) and uniformly 15N‐labeled non‐cyclized type III AFP (red resonances). Sequence‐specific backbone resonance assignments are shown as one‐letter amino acid code numbered according to their position in the sequence (Supporting Information Fig. S3). (B) Overlaid strip plots of CBCA(CO)NH (green) and HNCACB (Cα, red and Cβ, blue) of the region linking the C‐ and N‐termini of type III AFP. Dashed lines indicate sequential connectivity of 72R‐73G‐74K‐1C‐2W‐3E.

A strip plot displaying the HNCACB and CBCACONH was produced showing all the RGKCWE residues of the extein scar [Fig. 5(B)]. This analysis allows for a sequential “walk” through the polypeptide backbone by iteratively correlating the amide group to the α and β carbons of the same amino acid residue and to those of the preceding residue. Here, the data unambiguously showed the C to N terminus covalent linkage of the Lys and Cys residues of the extein scar, and therefore the cyclization of the type III AFP.

Comparisons of activity between noncyclized wild‐type and circularized type III AFP

To assess the impact of covalently linking the N‐ and C‐termini of type III AFP on TH activity and the preservation of TH activity after heating, we performed TH experiments in triplicate. The cAFP shaped a single ice crystal into a hexagonal bipyramid (not shown) characteristic of type III AFP, indicating the protein binds ice in a similar manner to noncyclized type III AFP. To assess any changes in activity via circularization we examined the concentration‐dependence of TH activity. Comparing cAFP to noncyclized type III AFP on a molar basis shows these proteins possess approximately the same concentration‐dependent TH activity, revealing that circularization had no deleterious effect on ice‐binding activity [Fig. 6(A)]. Next, we examined the amount of TH activity cAFP retained after heating to 35°C, 68°C, and 100°C [Fig. 6(B)]. The TH activity of the heat‐treated cAFP and noncyclized type III AFP samples was markedly different. After heating at 35°C, the cAFP possessed over 90% activity whereas the noncyclized type III AFP was reduced to 65% activity of their respective unheated samples. This difference in activity after heating is more obvious after heating to higher temperatures. The cAFP retained approximately 50% and 40% activity after heating to 68°C or 100°C, respectively. The noncyclized type III AFP displayed only 30% activity after heating to 68°C and had almost no activity after heating to 100°C.

Figure 6.

TH activity of circularized and noncyclized type III AFPs. (A) A molar comparison of TH activity. Black = cAFP; grey = noncyclized AFP. (B) Amount of TH activity after heat treatment at various temperatures. All measurements were performed in triplicate with standard deviation displayed.

Assessment of structural and functional stability of cAFP

To establish if retention of TH activity by cAFP at significantly elevated temperatures is the result of enhanced folding stability, NMR spectroscopy was used examine the effect of temperature on the structures of cAFP and noncyclized type III AFP. We recorded ¹H‐¹⁵N HSQC spectra of both noncyclized AFP and cAFP at 25°C, 35°C, and 45°C (Supporting Information Fig. S4). Spectra of both noncyclized wild‐type and circularized type III AFP displayed well‐dispersed resonances, indicative of folded protein, with only slight temperature‐dependent chemical shift changes. Overlays of spectra for the circularized and noncyclized AFPs recorded at 25°C before and after heating at 68°C are shown in Figure 7(A,B). Upon heating, the cAFP spectrum revealed a large cluster of poorly dispersed peaks in the center of the spectra indicative of an unfolding event. However, many well‐dispersed resonances remain visible suggesting that a population of the cAFP sample remained folded [Fig. 7(A)]. Noncyclized type III AFP, on the other hand, exhibited a spectral pattern symptomatic of entirely misfolded or aggregated protein [Fig. 7(B)]. Consistent with these spectral features significant precipitate was observed in the NMR tube. Taken together, these studies demonstrate that circularization affords the cAFP thermal stability and recovery from thermal shock, but with some loss of protein due to aggregation and precipitation.

Figure 7.

¹H–15N HSQC analysis of circularized and non‐cyclized type III AFPs after heating. (A) Spectral overlay of cAFP collected at 25°C after heating at 68°C for 45 min (red resonances) or without heating (black resonances). (B) Spectral overlay of noncyclized AFP collected at 25°C after heating at 68°C for 45 min (red resonances) or without heating (black resonances).

Discussion

In this study, we circularized the moderately active type III AFP using the natural split‐intein DnaE from Nostoc punctiforme. To facilitate linking of the N and C termini while reducing strain on the structure of the protein and its ice‐binding surface, we used the preferred RGKCWE peptide extein sequence with an additional flexible linker. Initially, three different versions of this flexible linker (AA, GAA, and GGAA) were modeled and examined using MD simulations to determine an N to C linkage that preserved the integrity of the AFP fold in general, and the conformational properties of its ice‐binding face in particular. Based on the RMSD and RMSF of cAFPs harboring each of the three linkers, the RGKCWEGAA linker‐containing cAFP was selected for recombinant protein synthesis. After expression in E. coli, intein splicing and AFP circularization were confirmed by MALDI‐TOF spectrometry, electrophoretic mobility, and 2D NMR spectroscopy. Activity of cAFP was assured based on its ability to be purified by ice affinity and its ability to retard ice growth during TH measurements.

Comparing the noncyclized and cAFPs on a molar basis showed no significant changes in activity due to additional linker residues. The change in mass of the AFP is well below the threshold for heightening of TH activity resulting from increased size.34 As predicted, circularization led to an enhancement of thermal stability compared to the noncyclized type III AFP. NMR analysis showed a population of cAFP remained folded after heating to 68°C, whereas all of the noncyclized type III AFP appeared to be unfolded. Remarkably, when these heat‐treated samples were assessed for TH activity after rapid cooling over a matter of minutes on ice, cAFP had retained 50% activity and the noncyclized type III AFP had 30% activity compared to their respective unheated samples. Qualitative assessment of the NMR spectra suggested that much less than 50% of the cAFP and less than 30% of the noncyclized type III AFP remained properly folded in solution. This discrepancy between activity and amount of folded protein was unexpected and indicates other factors might be at play. One possibility is that we might be observing ice‐induced refolding of the AFP during the assay. Contact with the surface of a growing ice crystal may induce unfolded or misfolded AFP molecules to form their IBS on their surface ligand. Such a phenomenon may be facilitated by a decrease in entropy of the unfolded cAFP (i.e., the number of ways the protein can be unfolded) compared to that of an AFP with free N and C termini. Exposure to ice could further favor the folded state of the AFP, offering an explanation for the higher TH activity than would be expected from the proportion of folded protein observed in NMR spectra of heated samples. Moreover, we noted that circularization reduced aggregation/precipitation of unfolded protein, which in turn allows for the possibility of substrate‐induced refolding. Note also that the mM AFP concentrations used for NMR analysis are more prone to aggregation and precipitation after denaturation than dilute AFP solutions. This principle is at work when refolding type III from inclusion bodies. After dissolving the protein aggregates in 8 M guanidine‐HCl, refolding is favored by extreme dilution such that the polypeptide folds on its own without contact with other partially folded proteins that could lead to aggregation.

In addition to increasing thermal stability and recovery from partial unfolding, intein splicing may be useful for increasing AFP activity. It has been demonstrated that AFP activity can be enhanced by increasing the size of the AFP or by linking together multiple AFPs.34, 35 Therefore, linking multiple AFPs together via inteins may lead to superior activity and stability. Furthermore, we think AFPs can provide a model system for studying intein engineering. AFPs have a diverse range of structures; from single α‐helices to β‐solenoids to complex globular folds, yet can be assessed by the same ice‐binding and TH activity assays.1 Thus, the effect or compatibility of intein‐circularization on different secondary structures or protein folds can be elucidated, leading to the expansion of intein engineering knowledge.

Materials and Methods

Design of circularized type III AFP

PyMOL and PyRosetta were used to generate three different models of type III AFP that had their N‐ and C‐termini joined with unique linkers.36 The linkers consisted of the extein scar RGKCWE with additional alanine and glycine residues appended to span the length of the termini. GROMACS‐compatible coordinate files of each circular AFP construct were then made and energy minimized and subjected to MD simulations to assess linker and ice‐binding surface trajectories.37, 38 All energy‐minimized models were solvated in a 100 ų box of water and run through a 0.1 ns position‐restrained MD simulation at constant volume, constant temperature (272 K), and constant pressure (1 atm), followed by a 20 ns unconstrained MD simulation in GROMACS. RMSD was calculated by the GROMACS g_rms function on the backbone α‐carbons for all cAFP constructs. The RMSD was used to evaluate the overall stability of each circularized protein in comparison to the wild‐type type III AFP. RMSF of the backbone α‐carbons was calculated using the g_rmsf function in GROMACS. Subsequently, the RMSF of the α‐carbon atoms in the ice‐binding residues was used to select a linker with fluctuations similar to that of noncyclized wild‐type AFP. Based on these parameters, the RGKCWEGAA linker was selected for building the circularized type III AFP assessed here.

Circularized type III AFP expression and purification

The gene for circularized type III AFP, including the Npu DnaE intein/extein and the RGKCWEGAA linker sequence, was synthesized by Integrated DNA Technologies (IDT) and cloned into a pET24 plasmid using XbaI and NotI restriction sites (Supporting Information Fig. S3). The resulting plasmid was transformed into BL21 (DE3) cells for protein expression. Similar to the method detailed in Baardsnes et al. for expression of noncyclized type III AFP,39 cells were grown to an optical density (600 nm) of 0.8 and induced overnight at 20°C with 1 mM IPTG. After overexpression, cells were pelleted and resuspended in 50 mL of buffer containing 10 mM Tris‐HCl (pH 7.6) and 150 mM NaCl. Resuspended cells were lysed and the insoluble cell debris was removed by centrifugation. A 10‐mL aliquot of the supernatant was diluted 1:10 to a final volume of 100 mL in distilled water and subjected to ice‐shell affinity purification.40 The ice fraction was concentrated to 5 mL using a 3000 Da molecular weight cutoff centricon centrifugal filter and further purified using a HiLoad 16/60 Superdex 200 prep grade size‐exclusion column (GE Healthcare). Fractions containing the purified cAFP were pooled and concentrated. This process resulted in a yield of approximately 45 mg of soluble protein from 1 L of liquid culture, a modest 10% reduction in pure protein compared to typical expression of the non‐cyclized type III AFP.

Mass determination of circularized AFP

Protein masses were measured using a Sciex DE Pro MALDi‐TOF mass spectrometer. Samples were desalted using C4 zip tips and ionized by sinapinic acid matrix.

Thermal hysteresis and heat shock assays

TH assays were performed as previously described with minor modifications.41 AFP samples were dissolved in 10 mM Tris‐HCl (pH 7.6) and 100 mM NaCl. The solution temperature was cooled at a rate of 0.01°C/min by a Clifton nanoliter osmometer equipped with a Newport 3040 temperature controller and LabView software. Single ice‐crystal images were recorded using a Panasonic WV‐BL200 digital camera at a rate of 30 fps. TH values were measured in triplicate with an average ice crystal size of 20 µm. To assess AFP activity after heat shock, the samples with an AFP concentration of 3 mM, were heated to 35°C, 68°C, and 100°C for 30 min, and then rapidly cooled in wet ice.

NMR spectroscopy studies to determine end ligation and thermostability

An NMR sample was prepared that contained 3 mM uniformly 13C/¹⁵N‐labeled cAFP in 20 mM Tris‐HCl (pH 7.6), 5 mM Dithiothreitol (DTT), 50 mM NaCl, and 10% D₂O. Sequential backbone assignment of cAFP backbone resonances was carried out using HNCO, HN(CA)CO, HNCACB, and CBCA(CO)NH triple resonance experiments collected at 30°C. To assess thermostability, ¹H‐¹⁵N HSQC spectra of uniformly 15N‐labeled noncyclized type III AFP and cAFP were collected at 25°C, 35°C, and 45°C. The NMR samples were then removed from the spectrometer and heated to 68°C for 45 min before recording ¹H‐¹⁵N HSQC spectra at 25°C. All NMR experiments were performed using a Varian INOVA 600 MHz spectrometer equipped with a triple resonance room‐temperature probe. NMR data were processed using NMRPipe version 8.142 and spectra were assigned using CcpNmr Analysis version 2.4.2.43 Chemical shift data have been deposited in the BMRB with accession number 27120.

Supporting information

Supporting Information