βα-Hairpin Clamps Brace βαβ Modules and Can Make Substantive Contributions to the Stability of TIM Barrel Proteins

Non-local hydrogen bonding interactions between main chain amide hydrogen atoms and polar side chain acceptors that bracket consecutive βα or αβ elements of secondary structure in αTS from E. coli, a TIM barrel protein, have previously been found to contribute 4–6 kcal mol−1 to the stability of the native conformation. Experimental analysis of similar βα-hairpin clamps in a homologous pair of TIM barrel proteins of low sequence identity, IGPS from S. solfataricus and E. coli, reveals that this dramatic enhancement of stability is not unique to αTS. A survey of 71 TIM barrel proteins demonstrates a 4-fold symmetry for the placement of βα-hairpin clamps, bracing the fundamental βαβ building block and defining its register in the (βα)8 motif. The preferred sequences and locations of βα-hairpin clamps will enhance structure prediction algorithms and provide a strategy for engineering stability in TIM barrel proteins.


Introduction
The (ba) 8 , TIM barrel is one of the most common folds in biology, supporting a myriad of catalytic functions essential to life [1]. Experimental [2] and bioinformatics [1,3,4] analyses of TIM barrel proteins have led to the conclusion that a pair of adjacent parallel b-strands and the intervening anti-parallel a-helix, i.e., the bab module, serve as the minimal unit of stability. Gene duplication of this elemental bab building block into higher-order structures has been suggested to result in several common barepeat structures, including the TIM barrel, Rossman, flavodoxin and leucine-rich folds [3]. The interactions stabilizing this supersecondary structure include: (1) main chain-main chain (MC-MC) hydrogen bonds (H-bonds) between the b-strands, (2) intra-helical MC-MC H-bonds, (3) hydrophobic interactions between the side chains (SC) protruding from the b-strands and the a-helix, (4) side chain-side chain (SC-SC) H-bonds and salt bridges, (5) dipoledipole interactions between the a-helix and the pair of b-strands on which it docks [5] and (6) main chain-side chain (MC-SC) Hbonds [6][7][8]. The surprising role of a subset of non-local MC-SC H-bond interactions in structure and stability is the subject of this communication.
A majority of MC-SC interactions in proteins are local in sequence, usually within six residues [7,8], and are often involved in capping either the N-or the C-termini of a-helices [9,10]. Mutational analysis has shown that these non-covalent interactions usually contribute modestly, typically in the range of 1-2 kcal mol 21 , to the stability of their resident proteins [11][12][13][14].
In contrast, the removal of three non-local MC-SC H-bond interactions each reduce the stability of the alpha subunit of tryptophan synthase (aTS), a TIM barrel protein, by 4-6 kcal mol 21 , and disrupt the complete formation of the TIM barrel motif [6]. These three interactions in aTS, between MC amide H-bond donors and SC H-bond acceptors, connect the Nterminus of one element of secondary structure, either b-strand or a-helix, to the C-terminus of the subsequent element of structure, either a-helix or b-strand, respectively. These non-local MC-SC interactions were designated as ba-hairpin clamps and ab-hairpin clamps, respectively [6]. The significant contribution to structure and stability by three such clamps in aTS [6] raises the possibility that potent baand ab-hairpin clamps may be an important general feature of TIM barrel proteins.
A two-pronged approach was taken to probe the significance of ba-hairpin clamps in TIM barrel proteins. First, mutational analysis of two representative TIM barrel proteins, indole-3glycerol phosphate synthase (IGPS) from S. solfataricus (sIGPS) and E. coli (eIGPS), shows that a subset of their ba-hairpin clamps make significant contributions to protein stability. Second, a survey of 71 TIM barrel proteins [15] explored the frequency, location and sequence preferences of all ba-hairpin clamps. The observed preferences for location and sequence for the bahairpin clamps and their contribution to the structure and stability of TIM barrel proteins suggest that the recognition of these interactions can enhance protein structure prediction algorithms and provide a strategy for engineering stability in TIM barrel proteins.

Experimental analysis of ba-hairpin clamp interactions in two TIM barrel proteins
The generality of the potent hairpin clamps found in aTS was tested by mutational analysis of ba-hairpin clamps in two homologous TIM barrel proteins with low sequence identity (,30%) to aTS and to each other. sIGPS ( Figure 1A) and eIGPS ( Figure 1B), each contain three ba-hairpin clamps ( Figure 1C and 1D), some of which are conserved in location with those in aTS and others between sIGPS and eIGPS. Figure 1 displays the distances between the donor and acceptor atoms of the ba-hairpin clamps interactions observed in sIGPS ( Figure 1C) and eIGPS ( Figure 1D). The solvent-exposed surface area of the H-bond acceptor atoms ranges from 0.2 Å 2 (0%) to 11.8 Å 2 (,25%), while the MC H-bond donor amide is typically completely excluded from solvent ( Figure 1C and 1D). The b1a1 clamp is observed in aTS and eIGPS (aTS-b1a1-F19 NH R Od2 D46, eIGPS-b1a1-F50 NH R Oc S82), the b2a2 clamp only appears in sIGPS (sIGPS-b2a2-S104 NH R Oe1 E74), the b3a3 clamp is observed in all three proteins (aTS-b3a3-I97 NH R Od2 D124, sIGPS-b3a3-I107 NH R Od1 D128 and eIGPS-b3a3-I111 NH R Od2 D132), and the b7a7 clamp is observed in sIGPS and eIGPS (sIGPS-b7a7-K207 NH R Od2 N228 and eIGPS-b7a7-V211 NH R Od1 N231).
Perturbation of the secondary and tertiary structure by clamp deletion in sIGPS and eIGPS. The contribution of each ba-clamp interaction to the structure of the TIM barrel proteins, sIGPS and eIGPS, was assessed by replacing the H-bond acceptor SC with alanine and monitoring the effects on the secondary and tertiary structure by far-UV and near-UV circular dichroism (CD) spectroscopy. The far-UV CD spectra for the wild-type (WT) and clamp-deletion variants of sIGPS (sIGPS-WT, sIGPS-Db2a2-E74A, sIGPS-Db3a3-D128A and sIGPS-Db7a7-N228A) and eIGPS (eIGPS-WT, eIGPS-Db1a1-S82A, eIGPS-Db3a3-D132A and eIGPS-Db7a7-N231A) are shown in Figure 2A and 2B, and the near-UV CD spectra are shown in Figure 2C and 2D. Relatively small changes in the far-UV and near-UV CD spectra are observed for sIGPS-Db3a3-D128A, eIGPS-Db1a1-S82A and eIGPS-Db3a3-D132A compared to their respective WT sequences. However, the significant changes in the near-UV CD spectra for the sIGPS-Db2a2-E74A, sIGPS-Db7a7-N228A and eIGPS-Db7a7-N231A variants imply that the deletion of the b7a7 clamps in both proteins and the b2a2 clamp in sIGPS result in altered aromatic side chain packing. (C) and (D) display the intervening elements of secondary structures between the residues forming the clamps for sIGPS: sIGPS-b2a2-S104 NH R Oe1 E74; sIGPS-b3a3-I107 NH R Od1 D128; and sIGPS-b7a7-K207 NH R Od2 N228 and for eIGPS: eIGPS-b1a1-F50 NH R Oc S82; eIGPS-b3a3-I111 NH R Od2 D132; and eIGPS-b7a7-V211 NH R Od1 N231. The SCs involved in the clamp interactions are highlighted with the H-bond donor and acceptor atoms shown in blue and red, respectively. The distances between the donor and acceptor atoms are indicated. The solvent exposed surface areas of the H-bond donor and acceptor atoms is shown in parenthesis. The H-bonds and their corresponding distances were determined by using the program HBPLUS [45]. The structures were generated using PyMOL v 0.99 [46], and the PDB codes are 2C3Z for sIGPS [21] and 1PII for eIGPS [22]. doi:10.1371/journal.pone.0007179.g001 Perturbation of stability by clamp deletion in sIGPS and eIGPS. The effect of ba-hairpin clamp deletion on the stability of sIGPS and eIGPS was determined by urea denaturation. As for aTS [16], both sIGPS [17] and eIGPS [18] unfold via a highly populated intermediate, and their unfolding titration curves are well described by a three-state model, Ns Is U. With the exception of eIGPS-Db7a7-N231A, the urea-induced unfolding transition for each of the remaining five clamp-deletion variants is also well-described by this three-state model ( Figure 3A and 3B). Because a distinct transition between the native state (N) and the intermediate state (I) is not observed during the urea induced denaturation of eIGPS-Db7a7-N231A ( Figure 3B), kinetic unfolding experiments were performed to verify the existence of I and measure the free energy difference between N and I [6].
The presence of I in eIGPS-Db7a7-N231A is verified by the observation of a slow kinetic unfolding phase, whose relaxation times decrease with increasing final denaturant concentration [19], when eIGPS is subjected to an unfolding jump from 0 to 3 M urea where I is expected to be highly populated. Because the amplitude of the unfolding phase is proportional to the population of N from which the reaction initiates, the decrease in the amplitude as a function of increasing initial urea concentrations ( Figure S1) can be fit to a two-state model, N s I, to extract the stability, DGu NI , and the urea dependence of the stability, m NI (Dataset S1). These parameters are used to fit the CD unfolding transition for eIGPS-Db7a7-N231A ( Figure 3B) and to extract DGu IU and m IU (Methods S1).
The stabilities of N and I for the clamp-deletion variants and the WT parent sequences are illustrated graphically in Figure 3C and 3D for sIGPS and eIGPS, respectively. The free energy differences between N and I, DGu NI , and between I and the unfolded state, U, DGu IU , as well as the m-values, are tabulated in Table 1. The deletion of the b2a2 clamp in sIGPS, sIGPS-Db2a2-E74A, only reduces the stability of N by 1.08 kcal mol 21 , and the deletion of the b3a3 clamp, sIGPS-Db3a3-D128A, has no significant effect on its stability. By striking contrast, the elimination of the b7a7 clamp, sIGPS-Db7a7-N228A, reduces the stability of N by 4.30 kcal mol 21 . Consistent with the absence of these clamps in I for all of these variants, the free energy differences between I and U for the clamp-deletion variants are comparable to the corresponding value for sIGPS-WT ( Figure 3C and Table 1). Similar results are obtained for eIGPS. Only eIGPS-Db7a7-N231A decreases the stability of N significantly, DDGu = 4.32 kcal mol 21 . eIGPS-Db1a1-S82A and eIGPS-Db3a3-D132A have no significant effect on the stability of N, and none of the clampdeletion variants perturb the stability of I relative to U (Figure 3D and Table 1). Thus, while the elimination of either the b1a1, b2a2 or b3a3 clamps has only marginal effects on sIGPS and eIGPS, the b7a7 clamps in both proteins contribute significantly to both the structure and the stability of the native states for their resident TIM barrel protein.

Survey of ba-hairpin clamps in the TIM barrel proteins
The observation that ba-hairpin clamps can have a significant effect on structure and stability in three TIM barrel proteins motivated a survey of the prevalence of such non-local MC-SC Hbonds in the TIM barrel fold. This analysis was carried out for a structural database of 71 TIM barrel domains, previously reported as a non-redundant representation of the TIM barrel fold [15]. Hbonds between main chain amide hydrogens and polar side chains (MC NH R SC) that serve as ba-hairpin clamps in the TIM barrel domains were identified (Materials and Methods) for a direct comparison with experimental results.
In the 71 TIM barrel proteins examined, there are 131 MC NH R SC ba-hairpin clamps. As can be seen in Table S1, there is a very significant preference, .42% of the clamps (x 2 Yates = 592.49, n = 131, d = 3, p-value = 4.26610 2128 ), for aspartic acid SCs forming H-bonds with the MC amide hydrogen of isoleucine, leucine and valine residues. Inspection of the location of the donor and acceptor residues in the b-strands reveals that every bahairpin clamp secures the N-terminus of one b-strand to the loop preceding the subsequent b-strand in the barrel.
The locations of the entire group of 131 MC NH R SC bahairpin clamps are displayed in Figure 4A, with each clamp interaction represented as a bridge across two adjacent b-strands. A very strong preference (77%) is seen for b1a1, b3a3, b5a5 and b7a7 clamps, where the SC acceptor is C-terminal to the MC Hbond donor. With the exception of 13 b8a8 clamps, the paucity of b2a2, b4a4 and b6a6 clamps is distinct from their odd b-strand counterparts. The relatively large number of clamps for the b8b1 interface may reflect the necessity for securing the N-and Cterminal b-strands. Far fewer ba-hairpin clamps, in which the SC acceptor is N-terminal to the MC H-bond donor, are observed. Highlighting the significance of this distribution pattern, the 55 Ile, Leu and Val (I/L/V) MC R SC Asp (D) sub-group of ba-hairpin clamps always have their MC H-bond donor I/L/V located in the odd-numbered stands, b1, b3, b5 or b7, and their SC acceptor, D, is always located before the succeeding even-numbered b-strands, b2, b4, b6 and b8. There is also a strong preference for the I/L/V residue to occupy the 2 nd position in the odd-numbered b-strand and for the D residue to occupy the position immediately preceding the even-numbered b-strand ( Figure 4B). This positional preference braces two consecutive and adjacent b-strands, along with the intervening helix, and reinforces the b-strand register required for the canonical TIM barrel architecture [20].

Discussion
Experimental analysis of ba-hairpin clamps between MC Hbond donors and SC H-bond acceptors in three TIM barrel proteins, aTS [6], sIGPS and eIGPS, has shown that a subset of these non-covalent interactions make substantive contributions to stability. Comparisons of the potency of the ba-hairpin clamps in these three proteins shows no correlation between the contributions of these clamps to stability and either the location of the clamps in the structure, their contributing residues or their relative exposure (0-25%) to the solvent. The observation of potent clamps formed by the neutral N228 in sIGPS and N231 in eIGPS, the b7a7 clamps, also shows that the formal negative charge on the aspartic acid H-bond acceptors in the remaining two potent bahairpin clamps is not determinative of the strength of the clamp interaction. An examination of the crystal structures of the three proteins, however, suggests that the length of the H-bond in each structure differentiates between the clamps that make major or minor contributions to stability (Table 1). Although the nominal resolutions of the crystal structures of these proteins, 2.0 to 2.8 Å [21][22][23], dictate that the correlation between H-bond length and the clamp contribution to protein stability be viewed as tentative, it appears ba-hairpin clamps whose H-bonds are less than 2.8 Å in length are those, which when replaced with alanine, reduce the stability of the native state by 4-6 kcal mol 21 . The apparent correlation provides a logical and testable hypothesis for future experiments on ba-hairpin clamps in other TIM barrel proteins.
The assay for the contribution of the ba-hairpin clamps to the stability of three TIM barrel proteins involves the replacement of the polar side chain H-bond acceptors, asparagine, aspartic acid, glutamic acid and serine, with alanine. The absence of the H-bond acceptor moiety is accompanied by the introduction of a potential void for these buried side chains, reflecting the absence of chemical mass as the side chain is truncated to the b-carbon. The perturbations in the secondary and/or tertiary structures induced by the mutations (Figure 2C and 2D) show that the loss of the clamp is propagated to numerous other non-covalent interactions via the global cooperativity of the native conformation.
The absence of the ba-hairpin clamps in the I states of all three TIM barrel proteins demonstrates that the potent effects of these clamps only appear as the N state appears [6]. Kinetic folding studies on aTS revealed further that each clamp is crucial for accessing the transition state ensemble required to reach the properly-folded structure [6]. Although the local connectivity of the bab modules might have been expected to enable the clamp to have a role in the early stages of the folding reaction, the primary role of the potent set of clamps is to drive the final stage of the reaction to completion and fully develop global cooperativity.
The 4-fold symmetry of the preferred ba-hairpin clamps is mirrored, not only in the symmetry of the bab modules, but also in the packing of the side chains in the interior of the b-barrel. A residue oriented towards the inside of the b-barrel from an oddnumbered b-strand is at the same level as corresponding residues from the three remaining odd-numbered b-strands. The next layer is comprised of the four side chains from the even-numbered bstrands ( Figure 5); the third, and usually final layer, is comprised again of side chains from the odd-numbered b-strands [20]. The layering of side chains inside the barrel has its origin in the tilt of the b-strands (35u) with respect to the central axis of the b-barrel [24]. The resulting S = 8 shear [20,24] provides a favorable orientation for the H-bonding network between adjacent parallel b-strands and provides opportunities for MC-SC ba-hairpin clamp interactions. Together, these non-covalent interactions and others stabilize the (ba) 8 , TIM barrel fold ( Figure 5). The observation of similarly placed non-local MC-SC interactions in a limited survey of flavodoxin fold proteins (data not shown) suggests that ba-hairpin clamps are a common structural feature of ba-repeat proteins. The chemical origin for the asymmetry between odd-and evennumbered b-strands is apparent from an inspection of the residue preference (.15%) at positions preceding the N-terminus of each b-strand ( Figure 5). The conserved proline just before oddnumbered b-strands provides a kink in the backbone that marks the beginning of a b-strand [5]. The preferred sequence pattern of the tight turn connecting the a-helix and the subsequent evennumbered b-strand ( Figure 5), GAD, has been reported previously [1]. The positive Q angle allowed by glycine and the hydrophobic nature of alanine immediately following the a-helix enables a Schellman motif for the C-terminal capping of the helix [10] and a tight turn to the next b-strand. The aspartic acid just prior to the beginning of even-numbered b-strands forms the ba-hairpin clamp and braces the bab module. This N-terminal cap for the odd-numbered b-strand is very often complemented by a MC-MC H-bond, with the amide group of the aspartic acid acting as the donor to the MC carbonyl oxygen of the partner residue. While other SC acceptors are observed (Table S1), the length of the aspartic acid side chain appears to be optimal for the reinforcement of the MC-SC H-bond with the MC-MC H-bond, providing a plausible explanation for its higher frequency in bahairpin clamps.
The preference for I/L/V residues at the MC NH H-bond donor position may reflect, in part, the .40% occurrence of these residues in parallel b-strands of TIM barrel proteins [20]. Further, along with alanine and glycine, I/L/V are the only amino acids that do not partition favorably from the vapor phase to water [25]. As such, these large aliphatic side chains are especially effective at excluding water from MC-SC H-bonds in the ba-hairpin clamps. The exclusion of water, that is apparent from the limited access to solvent for the H-bond donor and acceptor atoms of potent clamp interactions in aTS, sIGPS and eIGPS ( Figure 1C and 1D), is expected to strengthen these H-bonds and make them more resistant to exchange with solvent, as observed previously for aTS [26][27][28]. This presumption is supported by the conclusions of Gao et al. [29], who recently reported that the strength of a MC-MC H-bond is inversely related to the polarity of its local environment. Valine more effectively screened an underlying b-sheet MC-MC H-bond from solvent than alanine in a Pin WW domain, increasing the strength of the H-bond by up to 1.2 kcal mol 21 .
The occurrence of the bab motif in a large number of protein families [3,20] suggests that the N-terminal capping of b-strands by ba-hairpin clamps, akin to the analogous N-capping of ahelices [9,10], may be a useful property for the refinement of protein fold prediction and for engineering stability in ba-repeat proteins. ba-repeat proteins are readily recognized from their sequences and the predicted alternating patterns of a-helices and b-strands [30]. The refinement of the 3D structures predicted from knowledge-based potentials [31], threading [32] and homology modeling [33] of these protein sequences, could be enhanced by screening for ba-hairpin clamps between the MC amide hydrogens at favored positions near the N-terminus of a b-strand and H-bond acceptor SC in the loop before the subsequent bstrand (,25 residues apart in sequence). These clamps would establish the register of the pair of b-strands, and, with the very short loop linking the intervening a-helix to the second b-strand, it might be possible to establish the register of the a-helix on the bstrand pair in the bab module. Although TIM barrel proteins typically contain only a few ba-hairpin clamps, defining the spatial relationships of the components of a subset of bab modules might increase the probability of predicting the packing of adjacent barepeats in the structures. The effect of accurately predicting the structure of one bab module might, therefore, propagate throughout the TIM barrel protein.
The TIM barrel architecture provides a scaffold that is capable of a very diverse set of enzymatic functions [1], and this property has enabled TIM barrel enzymes to be re-engineered in order to accommodate alternative substrates [34][35][36][37] and even to catalyze non-biological reactions [38]. Because the active sites of TIM barrel enzymes are invariably comprised of the loops protruding from the C-termini of the b-strands, engineering ba-hairpin clamps at the N-termini of the b-strands offer a unique opportunity to enhance the stability of TIM barrel proteins without compromising function.

Clamp-deletion variants
The plasmid encoding a truncated version of sIGPS, in which the non-canonical additional a-helix (a00) at the N-terminus was deleted to eliminate aggregation during folding, pTNI4 [17], was obtained from Dr. K. Kirschner (University of Basel, Switzerland). The plasmid coding for eIGPS, pJB122 [39], was obtained from Dr. J. M. Blackburn (University of the Western Cape, South Africa). The eIGPS, with an additional Ala residue after the start codon and a C-terminal FLAG peptide sequence (GSDYKDDDDK), is fully folded and catalytically active [39]. Oligonucleotides for mutagenesis were purchased from Eurofins MWG Operon (Huntsville, AL), and the Quickchange TM sitedirected mutagenesis kit was obtained from Stratagene (La Jolla, CA). The site-directed mutations were confirmed by DNA sequence analysis (Genewiz Inc, NJ).

Protein expression and purification
The sIGPS protein and its variants were expressed in BL21/ DE3 cells and purified as described previously [17]. The expression and purification of eIGPS and its variants followed the same protocol, with the exception that the procedures were conducted at pH 7.0. The purity (.95%) was demonstrated by the appearance of a single band Coomassie blue stained PAGE and confirmed using electrospray mass spectrometry at the

Circular dichroism
Far-and near-UV CD spectroscopy was employed to monitor the secondary and the tertiary structure near aromatic side chains, respectively. Spectra were obtained on a Jasco Model J-810 spectropolarimeter equipped with a thermoelectric cell holder. Far-UV CD data were collected from 280 nm to 185 nm at a scan rate of 50 nm/min and at 1 nm intervals using a 0.1 cm pathlength cell, a bandwidth of 2.5 nm, with an averaging time of 8 s. Three replicate spectra were collected and averaged. The protein concentration was 5 mM. Near-UV CD data were collected from 350 nm to 250 nm at 5 nm/min using a 0.5 cm path length cell, and the protein concentration was 50-150 mM. The temperature was maintained at 25uC with a computercontrolled Peltier system.

Thermodynamic measurements
The stability of the IGPS clamp-deletion variants was measured by urea denaturation as described previously [17] in a buffer containing 10 mM potassium phosphate, pH 7.8 for sIGPS and pH 7.0 for eIGPS, 0.2 mM K2EDTA, and 1 mM bME. A Hamilton 540B automatic titrator was used to prepare the samples containing 0 to 8 M urea at concentration increments of 0.2 M urea to enhance the precision of the measurements. The samples were incubated overnight at 25uC to ensure equilibration.

Data analysis
Equilibrium CD data at 222 nm were fit to a three-state model, N s I s U, as described previously [40]. All thermodynamic folding data were fit using Savuka version 6.2, an in-house, nonlinear, least-squares program [40].

Survey of TIM barrel proteins
A database of 71 TIM barrel proteins has been previously developed (http://www.cbrc.jp/,gromiha/tim/proteinlist.html [15]) from the SCOP [41] and HOMSTRAD [42] databases, with a pair-wise sequence homology of ,25%. The highest resolution structure for each domain was chosen from the Protein Data Bank [43]. The secondary structure was calculated using the DSSP program [44] and the H-bond interaction parameters were calculated using default settings of the HBPLUS program [45].

Definitions of ba-hairpin clamp interactions
For each protein, the 8 canonical b-strands and a-helices in the context of the TIM barrel architecture were identified and labeled accordingly. H-bonding partners identified using the HBPLUS program [45], were subjected to the following filters: 1) the Hbonds must be between a MC amide donor and a SC acceptor, 2) the amino acid chain length between the donor and acceptor must be $15 residues thereby eliminating shorter-range helix-capping interactions [9,10] and 3) the chain must include exactly one bstrand and one a-helix identified in the context of the TIM barrel architecture. For the case of the b8a8-hairpin clamps, MC NH R SC H-bonds between the residues prior to b1 and b8 were included. The H-bonds that passed each stage of the filtering process were exported to a PyM0L [46] script in color-coded fashion for manual confirmation.
Statistical significance of residue preference for bahairpin clamps.
The frequency of MC NH R SC H-bonds in the 71 TIM barrel proteins, where the donor and acceptor residues were at least 15 amino acids apart and were not involved in ba-hairpin clamp interactions, was determined. This frequency was used to calculate the expected frequency of H-bonding between any two types of residues and compared to that observed in ba-hairpin clamps. Four categories, Ile MC R SC D, Leu MC R SC D, Val MC R SC D, Other MC R SC Other, were used to determine the x 2 distribution probabilities, with Yates correction [47], of observed ba-hairpin clamps.

Supporting Information
Dataset S1 Thermodynamic analysis of the eIGPS Db7a7