Single-Molecule Experiments Reveal the Flexibility of a Per-ARNT-Sim Domain and the Kinetic Partitioning in the Unfolding Pathway Under Force

Biophysical Journal Volume 102 May 2012 2149–2157 2149

Single-Molecule Experiments Reveal the Flexibility of a Per-ARNT-Sim Domain and the Kinetic Partitioning in the Unfolding Pathway under Force

Xiang Gao, Meng Qin, Puguang Yin, Junyi Liang, Jun Wang, Yi Cao,* and Wei Wang* National Laboratory of Solid State Microstructure and Department of Physics, Nanjing University, Nanjing, People’s Republic of China

ABSTRACT Per-ARNT-Sim (PAS) domains serve as versatile binding motifs in many signal-transduction proteins and are able to respond to a wide spectrum of chemical or physical signals. Despite their diverse functions, PAS domains share a conserved structure. It has been suggested that the structure of PAS domains is flexible and thus adaptable to many binding partners. However, direct measurement of the flexibility of PAS domains has not yet been provided. Here, we quantitatively measure the mechanical unfolding of a PAS domain, ARNT PAS-B, using single-molecule atomic force microscopy. Our force spectroscopy results indicate that the structure of ARNT PAS-B can be unraveled under mechanical forces as low as ~30 pN due to its broad potential well for the mechanical unfolding transition of ~2 nm. This allows the PAS-B domain to extend by up to 75% of its resting end-to-end distance without unfolding. Moreover, we found that the ARNT PAS-B domain unfolds in two distinct pathways via a kinetic partitioning mechanism. Sixty-seven percent of ARNT PAS-B unfolds through a simple two-state pathway, whereas the other 33% unfolds with a well-defined intermediate state in which the C-terminal b-hairpin is detached. We propose that the structural flexibility and force-induced partial unfolding of PAS-B domains may provide a unique mechanism for them to recruit diverse binding partners and lower the free-energy barrier for the formation of the binding interface.

INTRODUCTION Per-ARNT-Sim (PAS) domains are ubiquitous in signal- the conformational dynamics of proteins, such as the exis- transduction proteins and serve as versatile interaction tence of mechanical unfolding intermediate states (26–33), hubs (1–4). They sense and respond to various chemical the presence of parallel unfolding pathways (34–38), the and physical stimuli and regulate the activities of many structural changes upon ligand binding or protein-protein downstream effector domains. In contrast to their functional interactions (39–45), and the capture of rare misfolding diversity, PAS domains have a conserved structure com- events (46,47), most of which are difficult or impossible posed of a five-stranded antiparallel b-sheet and several to study at the ensemble level. Previous single-molecule a-helices (5–9)(Fig. S1 in the Supporting Material). The AFM studies on photoactive yellow protein (48,49), a repre- b-sheet is essential for the function of PAS domains, as it sentative PAS domain, revealed that the PAS domain is can bind to myriad protein targets via the solvent-exposed extended by ~3 nm and mechanically destabilized by surface and modulate the binding through conformational ~30% in the light-activated state, indicating partial unfold- changes triggered by environmental signals or cofactor ing of photoactive yellow protein upon photoactivation occupancy at the other part of the domain (8,10–16). (48). However, the conformational dynamics of the native Because of their universal binding capability, it is suggested PAS domain is not fully addressed. Moreover, in this that the b-sheet of PAS domains is flexible and adaptable to single-molecule AFM experiment (48), the PAS domain different binding targets (1). However, a direct and quantita- was not stretched along the N- and C-termini but along tive measurement of the flexibility of the PAS domain is the axis defined by two engineered cysteine residues. Such lacking. Here, we applied atomic-force-microscope (AFM)- force direction may not be physiologically relevant to the based single-molecule force spectroscopy to directly probe signal transduction processes mediated by PAS domains. the conformational change of a PAS domain, ARNT Here, we use single-molecule AFM to study conforma- PAS-B, upon elongation force. tional dynamics of the ARNT PAS-B domain by applying Single-molecule force spectroscopy has evolved into a force along the N- and C-termini, following its native powerful tool to study the conformational dynamics of connection in proteins. ARNT PAS-B domain is primarily proteins, especially protein folding/unfolding dynamics involved in the binding of HIF-2a for the formation of a (17–22). It allows force to be applied to a protein to trigger dimeric transcriptional regulator complex through an anti- the conformational change and quantitatively measure the parallel b-sheet-b-sheet packing (8).The two terminal free-energy landscape along the force direction (23–25). b-strands are part of the b-sheet and are proposed to be Single-molecule AFM has revealed rich information on responsible for the flexibility of the b-sheet (1,7). Our single-molecule AFM results indicate that despite its high thermodynamic stability, the structure of ARNT PAS-B Submitted December 28, 2011, and accepted for publication March 20, 2012. can be disrupted by mechanical forces as low as ~30 pN, *Correspondence: [email protected] or [email protected] which is lower than the force required to disrupt many proteins studied so far. Detailed kinetic studies reveal that Editor: Peter Hinterdorfer. Ó 2012 by the Biophysical Society 0006-3495/12/05/2149/9 $2.00 doi: 10.1016/j.bpj.2012.03.042 2150 Gao et al. the low unfolding forces of the ARNT PAS-B domain are the kn0 is set at two times that for the two-state pathway to reflect the due to its long unfolding distance (the distance between different probability of these two pathways based on Boltzmann distribu- the transition state and the native state along the force direction. Thus, when a PAS domain unfolds, it has a 33% probability of unfolding via an intermediate state and a 67% probability of proceeding directly to tion) of ~2 nm (50). Such a long unfolding distance allows the unfolded state. In the three-state pathway, the unfolding of the interme- ARNT PAS-B to adjust its conformation upon hydrody- diate state follows a similar equation: namic motion for the binding of different partners without FDx unfolding. Moreover, we find that the ARNT PAS-B domain ¼ i ; ki ki0 exp k T can unfold through two distinct pathways via a kinetic par- B titioning mechanism (51). Sixty-seven percent of ARNT where k is the unfolding rate constant and k is the spontaneous unfolding PAS-B unfolds in a two-state fashion, whereas 33% of the i i0 rate at zero force for the intermediate state. Dxi is the distance between the unfolding events involve a well-defined intermediate state intermediate state and the second unfolding transition state. in which the C-terminal b-hairpin is unstructured. We In the Monte Carlo simulation, we assume that (PAS-B)8 is elongated in propose that the structural flexibility and the kinetic parti- discrete time steps of dt (typically set to 10 4 s). The contour length of the tioning in the unfolding pathway may be important for protein can be calculated from the number of folded states, Nf, the number of intermediate states, Ni, the number of unfolding states, Nu, and the spacer PAS-B domains to interact with multiple binding partners lengths, L. with low sequence identity. Lc ¼ Lg þ Nf Lf þ Ni Li þ Nu Lu;

MATERIALS AND METHOD where Lc is the contour length of the PAS domain, and Lf,Li, and Lu are the lengths of a PAS domain in the folded state, the intermediate state, and the Protein engineering unfolding state, respectively. The extension of the polyprotein can be calculated as v multiplied by t, where v is the pulling speed and t is time. The Plasmid encoding wild-type ARNT PAS-B was synthesized by Genscript stretching force on poly-PAS can be calculated by the wormlike chain (Nanjing, China). The (PAS-B) polyprotein gene was constructed by iter- 8 (WLC) model atively cloning monomer into monomer, dimer into dimer, and tetramer into 0 1 tetramer using a previously described method based on the identical sticky ends generated by the BamHI and BglII restriction enzymes (New England k T B1 1 1 C B B v tC Biolabs, Ipswich, MA) (52). The polyprotein was expressed in BL21 and F ¼ @ þ A; p 4 v t 4 L purified by Ni-NTA resin (GE Healthcare Bio-Sciences, Little Chalfont, 1 c UK). Then the polyprotein was kept at 4C in solutions of 50 mM Tris Lc (pH 7.5), 15 mM imidazole, 47 mM NaCl, and 5 mM dithiothreitol. where p is the persistence length and the other parameters are defined above. Single-molecule AFM experiments The unfolding probability, P, can be calculated as k dt << 0, where k is kn or ki, depending on the corresponding unfolding event. Then P is Single-molecule AFM experiments were carried out on a JPK AFM compared with a random number between 0 and 1 to decide whether the (ForceRobot 300, JPK Instruments, Berlin, Germany). In all force spectros- protein unfolds and by which pathway it unfolds. In the two-state pathway, copy experiments, we deposited polyprotein onto a mica plate and allowed this leads to Nf ¼ Nf 1 and Nu ¼ Nu þ 1. In the three-state pathway, this it to adsorb for 10–20 min. Then, the fluid chamber was filled in z1mL leads to Nf ¼ Nf 1 and Ni ¼ 1. Then, after the unfolding of the interme- buffer containing 50 mM Tris, 17 mM NaCl, and 5 mM dithiothreitol diate state, Nu ¼ Nu þ 1 and Ni ¼ 0. The procedure repeats till both Nf and (pH 7.5). The AFM experiments were carried out after allowing the mixture Ni reach zero. to equilibrate for z30 min. Cantilevers (Biolever, Olympus, Melville, NY) with a typical spring constant of 6 pN nm1 were used for all experiments and calibrated using the equipartition theorem before each experiment. RESULTS The force-extension traces were recorded using JPK software and analyzed using a home-written protocol in Igor 6.0 (Wavemetrics, Lake Oswego, OR). Forced unfolding of ARNT PAS-B is characterized by a long unfolding distance to the transition state Monte Carlo simulation To study the mechanical unfolding of PAS-B domain by In the Monte Carlo simulation, the mechanical unfolding of the PAS-B using AFM-based single-molecule force spectroscopy, we domain is modeled as a bifurcation of a three-state and a two-state process. The unfolding rate constants of both native state and intermediate state are followed a standard polyprotein-based approach (Fig. 1 a) dependent on force. The unfolding rate for the native state in both pathways (52). We engineered a polyprotein, (PAS-B)8, that is made can be described as: of eight tandem repeats of PAS-B. Stretching (PAS-B)8 at 1 D a constant pulling speed of 400 nm s yields sawtoothlike ¼ F xn ; kn kn0 exp k T unfolding patterns (Fig. 1 b) in which each individual peak, B except the last one, corresponds to the unfolding of a PAS-B domain in the polyprotein. The last peak corresponds to the where k is the Boltzmann constant, T is the temperature in Kelvin, k is the B n detachment of the polyprotein from either the cantilever tip unfolding rate constant at a stretching force of F, kn0 corresponds to the unfolding rate constant at zero force, and Dxn is the distance between the or the substrate. Because the polyprotein is picked up by the native state and the transition state (52,53). For the three-state pathway, cantilever tip from the substrate surface randomly along its

Biophysical Journal 102(9) 2149–2157 Mechanical Unfolding of a PAS Domain 2151

FIGURE 1 Single-molecule force spectroscopy on PAS-B. (a) Schematic of the force spectroscopy

experiment on (PAS-B)8 using AFM. (b) Stretching polyprotein (PAS-B)8 results in sawtoothlike force- extension curves, in which each individual force peak corresponds to the mechanical unfolding of a PAS domain. Fitting consecutive unfolding events using the WLC model of polymer elasticity gives rise to a contour-length increment of DLc ¼ 39.3 nm. (c) The unfolding-force histogram of PAS-B (gray bars) centers at 32.6 5 6.3 pN. (d) The mechanical unfolding forces of PAS-B show weak dependency on the pulling speed, indicating a long unfolding distance. The error bars correspond to standard deviation. An unfolding rate 5 1 constant at zero force of k0 ¼ 3.0 10 s and an unfolding distance of Dxu ¼ 2.0 nm can be used to reproduce both the unfolding-force histogram and the pulling-speed-dependent experiments using a standard Monte Carlo simulation procedure (continuous lines in c and d).

contour, each trace has various numbers of unfolding events of a protein at its mechanical unfolding transition state from one to eight. Fitting consecutive unfolding events (Fig. S2)). The narrower the distribution, the longer is the using the WLC model for polymer elasticity yields a unfolding distance. Using a standard Monte Carlo simula- contour-length increment (DLc) of ~39 nm and a persistence tion procedure, we estimate the unfolding distance of length (p) of 0.5–1.0 nm. Since the PAS-B domain is PAS-B to be 2.0 nm and the spontaneous unfolding rate of comprised of 115 amino acids and the distance between PAS-B at zero force to be 3.0 105 s1. Generally, the the N- and C- termini of a folded PAS-B domain is fitting error using the Monte Carlo method is less than one 2.4 nm (PDB id 1X0O), stretching a PAS-B domain from order of magnitude for the spontaneous unfolding rate the folded state to the unfolded state should give rise constant and <0.3 nm for unfolding distance (Fig. S3). It to a contour-length increment of 39 nm (0.365 nm/aa is worthy of note that such a long unfolding distance of 114 aa 2.4 nm), in good agreement with our experimen- PAS-B is rare for b-sheet proteins, as the unfolding distance tally measured value. Therefore, it is certain that the force for b-sheet proteins is typically in the range 0.1–0.5 nm peaks shown in Fig. 1 b are indeed from the unfolding of (20,54). Our results indicate that the distance between the individual PAS-B domains. The amplitude of these unfold- two ends of PAS-B can extend by as much as 2.0 nm before ing force peaks is 32.6 5 6.3 pN (mean 5 SD, n ¼ 373), it is unfolded. Given that the measured distance between the indicating that the unfolding force of PAS-B domains is N- and C-termini based on the NMR structure is only rather low. This value is much lower than those for many 2.4 nm, such an extension accounts for 75% of its original proteins with b-sheet structures and even some a-helical length, indicating the high flexibility of PAS-B domain proteins (17,54) and is comparable to the force that can be upon stretching. Such a long unfolding distance is also generated by biological motors (55). The unfolding-force confirmed by pulling-speed-dependent experiments. The histogram of PAS-B is shown in Fig. 1 c. Besides the low average unfolding force of PAS-B increases from 29 pN at unfolding force, another interesting feature for the unfolding 50 nm s1 to 35 pN at 3600 nm s1 (Fig. 1 d). Using the force distribution is the narrow distribution. The standard same unfolding distance of 2.0 nm and unfolding rate deviation of the unfolding force is as small as 6.3 pN, which constant at zero force of 3.0 105 s1, we are able to is much smaller than that for protein domains containing reproduce such pulling-speed dependency using Monte b-sheets, such as fibronectin, immunoglobulin, and b-grasp Carlo simulations. Such a weak dependency of unfolding domains. The distribution of the unfolding force is directly force on pulling speed provides additional evidence for related to the unfolding distance (Dxu), the distance between the long unfolding distance of PAS-B. native state and unfolding transition state along the force Since the two ends of PAS-B need to extend as long as direction (the length gain between the two stretching points 2.0 nm to reach the mechanical unfolding transition state,

Biophysical Journal 102(9) 2149–2157 2152 Gao et al. we would infer that the two terminal b-strands are partially separated under force before the entire protein is completely unfolded. A schematic drawing of the structure of the native state and transition state of PAS-B domain is shown in Fig. S2. Closely examining the NMR structures of PAS-B reveals that the ﬁrst two hydrogen bonds between strand- A and strand-I (Thr361.O-Asn463.H and Phe363.H-Asn461.O) are 2.52 A˚ and 2.19 A˚ in length, respectively, i.e., longer than typical hydrogen bonds in antiparallel b-strands (~2 A˚ )(7). Such long hydrogen bonds indicate that they are prone to break by stretching. Hence, it is quite possible that the major barrier for mechanical unfolding of PAS-B is located in the middle part of the two terminal b-strands. However, to pinpoint the exact location of the mechanical unfolding barrier, detailed molecular dynamics simulations are required.

ARNT PAS-B unfolds through two distinct pathways By close examination of the force-extension curves of PAS-B, we found that it unfolds through two distinct pathways (Fig. 2 a). The majority of the unfolding events show two-state unfolding behavior without any detectable mechanical unfolding intermediate states (Fig. 2 b). The contour-length increment for two-state unfolding events is ~39.3 5 1.0 nm (mean 5 SD, n ¼ 249), corresponding to full unfolding of PAS-B domains (Fig. 2 c). However, ~33% of the unfolding events show a transient mechanical unfolding intermediate state (Fig. 2 d). It is clear that the positions of the intermediate state are superimposable to FIGURE 2 Unfolding of PAS-B shows two different scenarios. (a)A typical force-extension curve of (PAS-B)8, in which two unfolding events each other, indicating that the intermediate state has a follow a two-state unfolding scenario (blue) and three follow a three-state un- well-defined structure. WLC fitting measures the contour- folding scenario (from native state to intermediate state (green) and from length increment between the native state and the inter- intermediate state to unfolded state (cyan)). (b) Superposition of four two- state unfolding events. (c) The histogram of DLc for the two-state unfolding mediate state (DLcN-I) at 10.5 5 2.1 nm (mean 5 SD, n ¼ 124) and the contour-length increment between the events centers at 39.3 nm. (d) Superposition of four three-state unfolding events. (e) DLc from the native state to the intermediate state (DLcN-I) and intermediate state and the unfolded state (DLcI-U)at from the intermediate state to the unfolding state (DLcI-U) are ~10.5 nm 28.7 5 2.5 nm (mean 5 SD, n ¼ 124) (Fig. 2 e). Summing (green), and ~28.7 nm (cyan), respectively. The total DLc for three-state up DLcN-I and DLcI-U yields the total contour-length incre- unfolding is almost identical to that for two-state unfolding. ment for mechanical unfolding of 39.2 5 1.2 nm for the three-state unfolding pathway, similar to that for the two- state unfolding scenario shown in Fig. 2 c. This indicates effect (Fig. 3 a). The unfolding forces for the intermediate that the three-state unfolding events are indeed from the un- states are 26.8 5 6.6 pN (mean 5 SD), which is slightly folding of the PAS-B domain. lower than the unfolding forces of the native state (Fig. 3 b). The average lifetime for the intermediate state is only ~5 ms. Within such a short timescale, the unfolding of the The mechanical unfolding intermediate state is intermediate state can be considered quasistatic, and the transient force is approximated as constant in the unfolding process Subsequently, we characterized the mechanical unfolding (Fig. 3 a). Therefore, the unfolding kinetics for the interme- intermediate state. As the position of the intermediate state diate state can be extracted from the dwell-time distribution in the force-extension curves overlaps with the cantilever of the intermediate state in Fig. 3 c. Exponential fitting relaxation phase after the unfolding of the native PAS-B of the distribution yields an unfolding rate constant of structure, we plotted force against time to more precisely 192 5 13 s1. Using this unfolding rate constant as a capture the unfolding kinetics of the mechanical unfolding constraint, we extracted the unfolding distance of 0.15 nm intermediate state and to avoid the cantilever overdamping for the intermediate state and the spontaneous unfolding

Biophysical Journal 102(9) 2149–2157 Mechanical Unfolding of a PAS Domain 2153

ing intermediate structures are referred as PAS-B(-N) and PAS-B(-C), respectively (Fig. S5). Another possible structure for the mechanical unfolding intermediate state is one in which both the N-terminal and C-terminal b-strands A and I are disrupted. However, this would lead to a contour-length increment of ~16 nm, longer than the measured DLcN-I of 10.5 nm. Moreover, this process requires the breaking of interactions between the A-B hairpin and the H-I hairpin, the simultaneous occurrence of which is very unlikely, if not impossible. To unequivocally distinguish the structure of the intermediate state between PAS-B(-N) and PAS-B(-C), we resorted to a more elaborate approach, the glycine-insertion method (29,56). Five glycine residues were inserted at two different positions of the PAS-B domain. This leads to a change in the contour-length increment of the mechanical unfolding events. If the inserted glycine residues are within the mechanical unfolding intermediate state, it will increase DLcI-U without affecting DLcN-I, and vice versa. We have engineered what to our knowledge are two new proteins with five glycine residues inserted into the turn between A and B b-strands after FIGURE 3 Transient intermediate state in the three-state unfolding residue isoleucine 359 (PAS-B-I359/E360-5G) and the pathway. (a) A representative intermediate state shown in the force-time turn between H and I b-strands after residue tyrosine 450 plot (the red line is smoothed by moving average). The unfolding force is (PAS-B-Y450/S451-5G) of PAS-B. However, PAS-B-I359/ close to constant during the unfolding process, and the dwell time of the E360-5G cannot be expressed in Escherichia coli, probably intermediate state is ~8 ms. (b) The unfolding-force histogram of the intermediate state centers at 26.8 5 6.6 pN. (c) The dwell time of the interme- because the glycine insertion affects its proper folding (57). diate state. Exponential fitting to the dwell-time distribution yields an PAS-B-Y450/S451-5G is thermodynamically stable, allow- unfolding rate constant of 192 s1. Based on Monte Carlo simulation, the ing us to make the polyprotein and use single-molecule 1 unfolding rate constant at zero force, k0,is80s and the unfolding AFM to measure the effect of glycine insertion on the con- D distance, xu, is 0.15 nm for the mechanical unfolding of the intermediate tour-length increment (Fig. 4 a). Glycine insertion increases state. the total contour-length increment of the PAS-B domain by ~1.7 nm (Fig. S6) but does not affect the occurrence of the rate constant at zero force of 80 s1 using the Monte Carlo intermediate states. Around 35% of the unfolding events simulation (Fig. S4). The error for k0 is less than a half order follow the three-state unfolding pathway. Superpositioning of magnitude and that for Dxu is <0.15 nm (Fig. S4). The the three-state unfolding events of PAS-B-Y450/S451-5G unfolding forces of the intermediate state at different pulling and wild-type PAS-B shows that the DLcN-I increases by speeds are also shown in Fig. S4. It is worth noting that the ~1.6 nm, whereas the DLcI-U remains the same (Fig. 4 b). unfolding distance of the intermediate state is much shorter This is further confirmed by the contour-length increment than that of the native state, suggesting that the intermediate histogram in Fig. 4 c. These results indicate that the location state is prone to disruption upon stretching. Moreover, the at which the five glycine residues are inserted is outside the fast spontaneous unfolding rate of 80 s1 indicates that structure of the intermediate state. Therefore, the structure the lifetime of the mechanical unfolding intermediate state of the mechanical unfolding intermediate state should be is only ~0.0125 s at zero force. Therefore, this mechanical PAS-B(-C), the cyan parts shown in Fig. 4 a, in which the unfolding intermediate state of PAS-B domain is transient H and I b strands are detached. and cannot survive for long even without a stretching force. Is such an intermediate state thermodynamically stable if H and I b-strands are allowed to remain detached for a long time? To address this question, we engineer a trun- The structure of the mechanical unfolding cated mutation of PAS-B corresponding to the PAS-B(-C) intermediate state structure. However, the circular dichroism spectrum of Contour-length increment can be used as a sensitive indi- PAS-B(-C) shows that it mainly adopts a random-coil struc- cator to map the structure of the mechanical unfolding inter- ture instead of the expected a þ b structure, distinct from mediate state. A DLcN-I of 10.5 nm corresponds to the wild-type PAS-B (Fig. S7). This indicates that PAS-B is detachment of either 34 amino acids from the N-terminus not thermodynamically stable in the absence of C-terminal or 37 amino acids from the C-terminus of PAS-B to reach b-strands H and I. Our data are consistent with the chemical the intermediate state (see Supporting Material). The result- denaturation measurement of PAS-B, in which it was found

Biophysical Journal 102(9) 2149–2157 2154 Gao et al.

unfolding pathways have been reported for a few proteins in mechanical unfolding experiments. The detailed molecular mechanism and physiological impact of the kinetic partitioning merit further detailed study.

DISCUSSION The PAS-B domain is mechanically labile Mechanical force is involved in many biological processes, including muscle contraction, cell division, cell movement, protein translocation, protein degradation, and signal transduction (55). For many functions, such as muscle contraction, high mechanical stability of proteins is essential (22). However, for signal transduction, high mechanical stability may not be required. PAS domains are important building units of signal-transduction proteins. Here, we show that PAS domains unfold at low mechanical force, distinct from many other proteins studied so far by single-molecule AFM. Such a low unfolding force may be beneficial to its function, because very little external work is required to trigger the conformational change of PAS domains. Such external force may be obtained from external stimuli such FIGURE 4 Mapping the structure of the unfolding intermediate state as light, chemical signal, or even relative movement of using glycine insertion. (a) Schematic of the structure of PAS-B-Y450/ different domains. Therefore, we propose that the low S451-5G. Five glycine residues are inserted into the loop linking b-strands H and I. (b) Superimposing the three-state unfolding peaks for PAS-B- mechanical stability may be common to many PAS domains Y450/S451-5G and wild-type PAS-B. Insertion of five glycines leads to and other signal-transduction proteins. an increase in contour-length increment from the native state to the interme- The low mechanical stability of the PAS domain origi- diate state without affecting the contour-length increment from the interme- nates from its unique three-dimensional structure. Theoret- b diate state to the unfolded state. This suggests that the H and I -strands are ical and experimental efforts have demonstrated that not involved in the mechanical-unfolding intermediate state. (c) Histograms of the detailed contour-length increment of the glycine-insertion mutant, proteins with shear topology, in which two terminal PAS-B-Y450/S451-5G (open bars) in comparison with wild-type PAS-B b-strands are antiparallel, are generally mechanically more (solid bars). stable than proteins without this structural motif (59–61). This is because in such a structural arrangement, the that the folding and unfolding of PAS-B are both two-state hydrogen bonds that bond the two force-bearing termini processes without any detectable intermediate state (58). break concomitantly during the mechanical unfolding process. However, the two termini of PAS domain do not bear the shear topology. Instead, they are arranged in parallel The free-energy landscape for the mechanical and the hydrogen bonds between them break sequentially unfolding of PAS-B upon mechanical stretching. Typically, proteins with such Based on our single-molecule AFM data, the mechanical- a structural motif are mechanically labile and the ARNT unfolding free-energy landscape of PAS-B is summarized PAS-B reported here is a representative example. Moreover, in Fig. 5. Upon stretching, the N- and C-termini of PAS-B the major mechanical unfolding barrier resides in the middle b move against each other for ~2.0 nm, leading to the shearing of the N- and C-terminal strands. Such a long unfolding of two terminal b-strands, A and I. The free-energy barrier distance not only gives rise to high flexibility of the native structure of PAS-B but also lowers the unfolding force. for this transition is ~24 kBT, as estimated from the unfolding rate constant (Supporting Material). After surmounting this barrier, unfolding can proceed in two distinct pathways. Kinetic partitioning in the mechanical unfolding In the first pathway, the PAS-B domain completely unfolds of ARNT PAS-B without navigating any observable barriers. In the second pathway, the PAS-B domain reaches an intermediate state It is now widely accepted that proteins can fold to their corresponding to a structure with b-strands H and I native structure via multiple pathways by a kinetic partition- detached, which is ~10.5 nm away from the native state. ing mechanism (51,62), despite the fact that experimental Further unfolding requires it to overcome a barrier of evidence supporting this mechanism is still limited (34–38). 9.4 kBT to reach the completely unfolded state. Such parallel The mechanical unfolding of PAS-B provides new, to our

Biophysical Journal 102(9) 2149–2157 Mechanical Unfolding of a PAS Domain 2155

FIGURE 5 Free-energy landscape for the mechanical unfolding of PAS-B. The mechanical unfolding of PAS-B follows two distinct pathways, of which one involves a transient intermediate state, whereas the other does not. N, I, U, and EU correspond to native, intermediate, unfolded, and extended unfolded states, respectively. The detailed free-energy barriers and locations of each unfolding pathway are depicted at left. The possible structure of each state and the three-dimensional landscape are illustrated at right. The width of each arrow reflects the relative flux of each unfolding pathway. knowledge, evidence for this mechanism. We showed that free-energy landscape but not in every cross section. the mechanical unfolding could proceed through two Single-molecule AFM provides unique opportunities to distinct pathways. In the first path, the C-terminus of the study different cross sections of the free-energy landscape PAS domain unfolds first, leading to a marginally stable un- of protein folding. Therefore, many surprising folding/unfolding intermediate state in which the C-terminal b hairpin folding behaviors can be found from these studies, enriching is detached from the rest of the domain. In the second path, our understanding of the protein folding mechanism and the unfolding of the PAS domain does not show any detect- making the free-energy landscape accessible in multiple able intermediate states. The two pathways account for 33% dimensions experimentally. and 67%, respectively, of the unfolding events. Thus, the free-energy barrier for the three-state unfolding pathway is Physiological consequences of the flexible native only ~0.7 k T higher than that for the two-state unfolding B structure pathway. Based on the unfolding distance for the native state, such a difference in free-energy barrier will lead It has been proposed that structural flexibility is important to <1 pN difference in the unfolding force, below the for ligand binding or protein-protein interactions, as it detection limit of our AFM setup. There are two possible helps proteins to recruit binding partners and lowers the mechanisms that may explain such kinetic partitioning free-energy barrier for the formation of the binding inter- behavior. For the first one, there are two subpopulations in face. A recent fly-casting mechanism proposed by Wolynes the native-state ensemble with similar thermodynamic and co-workers indicates that proteins can even partially stability. The observed two distinct unfolding pathways unfold to facilitate the binding process (63–66). As the result from the unfolding of these two subpopulations. For major function of PAS-B domains is to bind different the second possible mechanism, unfolding can occur ligands and proteins for the signal-transduction process, through two different pathways, in which the transition we suggest that PAS-B domains are able to utilize this states are of similar stability but diverse structure. unique mechanism for binding in their physiological condi- Recent single-molecule AFM studies have provided tions. Our results show that PAS-B can extend up to 75% of increasing evidence for the kinetic partitioning mechanism its original end-to-end distance in its native conformation of protein unfolding, distinct from the results obtained by and unfold at a marginal force of <30 pN. Furthermore, other experimental methods (34–38). This is probably the unfolding can occur via a kinetic partitioning mecha- because of the unique unfolding/folding pathway probed nism and reach a transient intermediate state in one of the in single-molecule AFM experiments: the pathway along unfolding pathways. Such high structural flexibility makes the direction defined by the stretching force. The folding/ PAS-B domains ideal candidates for protein-binding unfolding of proteins may be cooperative in the globular motifs. Despite the high flexibility of PAS-B domains

Biophysical Journal 102(9) 2149–2157 2156 Gao et al. upon mechanical stretching, they are thermodynamically DNA binding activity and preparation of a DNA complex. J. Biochem. extremely stable (>9 kcal/mol) (58). Thus, the flexibility 134:83–90. of PAS-B domains is limited to particular directions. Since 10. Salomon, M., J. M. Christie, ., W. R. Briggs. 2000. Photochemical and mutational analysis of the FMN-binding domains of the plant the N- and C-termini of PAS-B domains are among the most blue light receptor, phototropin. Biochemistry. 39:9401–9410. flexible parts in the structure and are actively involved in 11. Harper, S. M., L. C. Neil, and K. H. Gardner. 2003. Structural basis of the binding process, we believe that such a structural and a phototropin light switch. Science. 301:1541–1544. topological arrangement of PAS-B domains is of biological 12. Zoltowski, B. D., C. Schwerdtfeger, ., B. R. Crane. 2007. Conforma- importance. However, it is still largely unknown whether tional switching in the fungal light sensor Vivid. Science. 316:1054– PAS-B domains are indeed subject to tensile force in their 1057. physiological condition. Therefore, further experimental 13. Evans, M. R., P. B. Card, and K. H. Gardner. 2009. ARNT PAS-B has a fragile native state structure with an alternative b-sheet register evidence is required to conclude that the flexibility and nearby in sequence space. Proc. Natl. Acad. Sci. USA. 106:2617–2622. function of PAS-B domains are indeed regulated by force 14. Mo¨glich, A., R. A. Ayers, and K. Moffat. 2010. Addition at the molec- in vivo. Nonetheless, the studies presented here open the ular level: signal integration in designed Per-ARNT-Sim receptor door for the study of anisotropic conformational flexibility proteins. J. Mol. Biol. 400:477–486. of proteins and its biological consequences. 15. Partch, C. L., and K. H. Gardner. 2011. Coactivators necessary for transcriptional output of the hypoxia inducible factor, HIF, are directly recruited by ARNT PAS-B. Proc. Natl. Acad. Sci. USA. 108:7739– SUPPORTING MATERIAL 7744. 16. Becker, D. F., W. Zhu, and M. A. Moxley. 2011. Flavin redox switching Protein sequence, estimation of the free energy barrier, calculation of of protein functions. Antioxid. Redox Signal. 14:1079–1091. the number of residues detached from protein at the intermediate state, 17. Carrion-Vazquez, M., A. Oberhauser, ., A. Martinez-Martin. 2006. and seven figures are available at http://www.biophysj.org/biophysj/ Protein nanomechanics as studied by AFM single-molecule force spec- supplemental/S0006-3495(12)00383-9. troscopy. In Advanced Techniques in Biophysics. J. Arrondo and A. Alonso, editors. Springer, New York. 163–236. We thank Dr. D. Thirumalai for penetrating discussions and critical 18. Puchner, E. M., and H. E. Gaub. 2009. Force and function: probing comments on the manuscript. proteins with AFM-based force spectroscopy. Curr. Opin. Struct. This work is supported by the National Natural Science Foundation of Biol. 19:605–614. China under grants 11074115, 81121062, 10834002, 10904064, and 19. Borgia, A., P. M. Williams, and J. Clarke. 2008. Single-molecule 31170813, the Natural Science Foundation of Jiangsu Province under grant studies of protein folding. Annu. Rev. Biochem. 77:101–125. BK2009008, the program for New Century Excellent Talents in University, 20. Crampton, N., and D. J. Brockwell. 2010. Unravelling the design prin- and the Priority Academic Program Development of Jiangsu Higher Educa- ciples for single protein mechanical strength. Curr. Opin. Struct. Biol. tion Institutions. 20:508–517. 21. Galera-Prat, A., A. Go´mez-Sicilia, ., M. Carrioń-Va´zquez. 2010. Understanding biology by stretching proteins: recent progress. Curr. REFERENCES Opin. Struct. Biol. 20:63–69. 22. Li, H., and Y. Cao. 2010. Protein mechanics: from single molecules to 1. Mo¨glich, A., R. A. Ayers, and K. Moffat. 2009. Structure and signaling functional biomaterials. Acc. Chem. Res. 43:1331–1341. mechanism of Per-ARNT-Sim domains. Structure. 17:1282–1294. 23. Carrion-Vazquez, M., H. Li, ., J. M. Fernandez. 2003. The mechan- 2. McIntosh, B. E., J. B. Hogenesch, and C. A. Bradfield. 2010. Mamma- ical stability of ubiquitin is linkage dependent. Nat. Struct. Biol. lian Per-Arnt-Sim proteins in environmental adaptation. Annu. Rev. 10:738–743. Physiol. 72:625–645. 24. Brockwell, D. J., E. Paci, ., S. E. Radford. 2003. Pulling geometry 3. Gilles-Gonzalez, M. A., and G. Gonzalez. 2004. Signal transduction by defines the mechanical resistance of a b-sheet protein. Nat. Struct. heme-containing PAS-domain proteins. J. Appl. Physiol. 96:774–783. Biol. 10:731–737. 4. Taylor, B. L., and I. B. Zhulin. 1999. PAS domains: internal sensors 25. Dietz, H., F. Berkemeier, ., M. Rief. 2006. Anisotropic deformation of oxygen, redox potential, and light. Microbiol. Mol. Biol. Rev. response of single protein molecules. Proc. Natl. Acad. Sci. USA. 63:479–506. 103:12724–12728. . 5. Nambu, J. R., J. O. Lewis, , S. T. Crews. 1991. The Drosophila 26. Marszalek, P. E., H. Lu, ., J. M. Fernandez. 1999. Mechanical unfold- single-minded gene encodes a helix-loop-helix protein that acts as ing intermediates in titin modules. Nature. 402:100–103. a master regulator of CNS midline development. Cell. 67:1157–1167. 27. Li, L., H. H. Huang, ., J. M. Fernandez. 2005. Mechanical unfolding 6. Huang, Z. J., I. Edery, and M. Rosbash. 1993. PAS is a dimerization intermediates observed by single-molecule force spectroscopy in domain common to Drosophila period and several transcription factors. a fibronectin type III module. J. Mol. Biol. 345:817–826. Nature. 364:259–262. 28. Bertz, M., and M. Rief. 2008. Mechanical unfoldons as building blocks 7. Card, P. B., P. J. Erbel, and K. H. Gardner. 2005. Structural basis of of maltose-binding protein. J. Mol. Biol. 378:447–458. ARNT PAS-B dimerization: use of a common b-sheet interface for hetero- and homodimerization. J. Mol. Biol. 353:664–677. 29. Schwaiger, I., A. Kardinal, ., M. Rief. 2004. A mechanical unfolding intermediate in an actin-crosslinking protein. Nat. Struct. Mol. Biol. 8. Erbel, P. J., P. B. Card, ., K. H. Gardner. 2003. Structural basis for 11:81–85. PAS domain heterodimerization in the basic helix-loop-helix-PAS transcription factor hypoxia-inducible factor. Proc. Natl. Acad. Sci. USA. 30. Puchner, E. M., A. Alexandrovich, ., M. Gautel. 2008. Mechanoenzy- 100:15504–15509. matics of titin kinase. Proc. Natl. Acad. Sci. USA. 105:13385–13390. 9. Kikuchi, Y., S. Ohsawa, ., Y. Fujii-Kuriyama. 2003. Heterodimers of 31. Garcia-Manyes, S., L. Dougan, ., J. M. Fernańdez. 2009. Direct bHLH-PAS protein fragments derived from AhR, AhRR, and Arnt observation of an ensemble of stable collapsed states in the mechanical prepared by co-expression in Escherichia coli: characterization of their folding of ubiquitin. Proc. Natl. Acad. Sci. USA. 106:10534–10539.

Biophysical Journal 102(9) 2149–2157 Mechanical Unfolding of a PAS Domain 2157

32. Nunes, J. M., U. Hensen, ., D. J. Muller. 2010. A ‘‘force buffer’’ pro- single-molecule experiments, analysis, and simulation. Proc. Natl. tecting immunoglobulin titin. Angew. Chem. Int. Ed. Engl. 49:3528– Acad. Sci. USA. 104:20799–20804. 3531. 50. Rief, M., J. Pascual, ., H. E. Gaub. 1999. Single molecule force spec- 33. Damaghi, M., S. Ko¨ster, .,D.J.Mu¨ller. 2011. One b hairpin fol- troscopy of spectrin repeats: low unfolding forces in helix bundles. lows the other: exploring refolding pathways and kinetics of the trans- J. Mol. Biol. 286:553–561. b membrane -barrel protein OmpG. Angew. Chem. Int. Ed. Engl. 51. Guo, Z. Y., and D. Thirumalai. 1995. Kinetics of protein-folding: 50:7422–7424. nucleation mechanism, time scales, and pathways. Biopolymers. 34. Peng, Q., and H. Li. 2008. Atomic force microscopy reveals parallel 36:83–102. mechanical unfolding pathways of T4 lysozyme: evidence for a kinetic 52. Carrion-Vazquez, M., A. F. Oberhauser, ., J. M. Fernandez. 1999. partitioning mechanism. Proc. Natl. Acad. Sci. USA. 105:1885–1890. Mechanical and chemical unfolding of a single protein: a comparison. 35. Mickler, M., R. I. Dima, ., M. Rief. 2007. Revealing the bifurcation in Proc. Natl. Acad. Sci. USA. 96:3694–3699. the unfolding pathways of GFP by using single-molecule experiments 53. Rief, M., M. Gautel, ., H. E. Gaub. 1997. Reversible unfolding of and simulations. Proc. Natl. Acad. Sci. USA. 104:20268–20273. individual titin immunoglobulin domains by AFM. Science. 276: 36. Bertz, M., A. Kunfermann, and M. Rief. 2008. Navigating the folding 1109–1112. energy landscape of green fluorescent protein. Angew. Chem. Int. Ed. 54. Li, H. B. 2008. ‘‘Mechanical engineering’’ of elastomeric proteins: Engl. 47:8192–8195. toward designing new protein building blocks for biomaterials. Adv. 37. Aggarwal, V., S. R. Kulothungan, ., S. R. Ainavarapu. 2011. Ligand- Funct. Mater. 18:2643–2657. modulated parallel mechanical unfolding pathways of maltose-binding 55. Bustamante, C., Y. R. Chemla, ., D. Izhaky. 2004. Mechanical proteins. J. Biol. Chem. 286:28056–28065. processes in biochemistry. Annu. Rev. Biochem. 73:705–748. 38. Peng, Q., J. Fang, ., H. Li. 2011. Kinetic partitioning mechanism 56. Carrion-Vazquez, M., P. E. Marszalek, ., J. M. Fernandez. 1999. governs the folding of the third FnIII domain of tenascin-C: evidence Atomic force microscopy captures length phenotypes in single at the single-molecule level. J. Mol. Biol. 412:698–709. proteins. Proc. Natl. Acad. Sci. USA. 96:11288–11292. 39. Ainavarapu, S. R., L. Li, ., J. M. Fernandez. 2005. Ligand binding 57. Philip, A. F., M. Kumauchi, and W. D. Hoff. 2010. Robustness and modulates the mechanical stability of dihydrofolate reductase. evolvability in the functional anatomy of a PER-ARNT-SIM (PAS) Biophys. J. 89:3337–3344. domain. Proc. Natl. Acad. Sci. USA. 107:17986–17991. 40. Bertz, M., and M. Rief. 2009. Ligand binding mechanics of maltose 58. Evans, M. R., and K. H. Gardner. 2009. Slow transition between two binding protein. J. Mol. Biol. 393:1097–1105. b-strand registers is dictated by protein unfolding. J. Am. Chem. Soc. 131:11306–11307. 41. Junker, J. P., F. Ziegler, and M. Rief. 2009. Ligand-dependent equilib- rium fluctuations of single calmodulin molecules. Science. 323: 59. Klimov, D. K., and D. Thirumalai. 2000. Native topology determines 633–637. force-induced unfolding pathways in globular proteins. Proc. Natl. Acad. Sci. USA. 97:7254–7259. 42. Cao, Y., M. M. Balamurali, ., H. Li. 2007. A functional single-mole- . cule binding assay via force spectroscopy. Proc. Natl. Acad. Sci. USA. 60. West, D. K., D. J. Brockwell, , E. Paci. 2006. Mechanical resistance 104:15677–15681. of proteins explained using simple molecular models. Biophys. J. 90: 287–297. 43. Cao, Y., T. Yoo, ., H. Li. 2008. Protein-protein interaction regulates . proteins’ mechanical stability. J. Mol. Biol. 378:1132–1141. 61. Cao, Y., C. Lam, , H. Li. 2006. Nonmechanical protein can have significant mechanical stability. Angew. Chem. Int. Ed. Engl. 45: . 44. Wang, C. C., T. Y. Tsong, , P. E. Marszalek. 2011. Inhibitor binding 642–645. increases the mechanical stability of staphylococcal nuclease. Biophys. J. 100:1094–1099. 62. Thirumalai, D., D. K. Klimov, and S. A. Woodson. 1997. Kinetic partitioning mechanism as a unifying theme in the folding of biomol- 45. Oroz, J., A. Valbuena, ., M. Carrioń-Va´zquez. 2011. Nanomechanics ecules. Theor. Chem. Acc. 96:14–22. of the cadherin ectodomain: ‘‘canalization’’ by Ca2þ binding results in a new mechanical element. J. Biol. Chem. 286:9405–9418. 63. Shoemaker, B. A., J. J. Portman, and P. G. Wolynes. 2000. Speeding molecular recognition by using the folding funnel: the fly-casting 46. Oberhauser, A. F., P. E. Marszalek, ., J. M. Fernandez. 1999. Single mechanism. Proc. Natl. Acad. Sci. USA. 97:8868–8873. protein misfolding events captured by atomic force microscopy. Nat. 64. Perham, M., M. Chen, ., P. Wittung-Stafshede. 2005. Unfolding of Struct. Biol. 6:1025–1028. heptameric co-chaperonin protein follows ‘‘fly casting’’ mechanism: 47. Jollymore, A., C. Lethias, ., H. Li. 2009. Nanomechanical properties observation of transient nonnative heptamer. J. Am. Chem. Soc. of tenascin-X revealed by single-molecule force spectroscopy. J. Mol. 127:16402–16403. Biol. 385:1277–1286. 65. Levy, Y., J. N. Onuchic, and P. G. Wolynes. 2007. Fly-casting in 48. Zhao, J. M., H. Lee, ., W. D. Hoff. 2006. Single-molecule detection of protein-DNA binding: frustration between protein folding and electro- structural changes during Per-Arnt-Sim (PAS) domain activation. Proc. statics facilitates target recognition. J. Am. Chem. Soc. 129:738–739. Natl. Acad. Sci. USA. 103:11561–11566. 66. Sugase, K., H. J. Dyson, and P. E. Wright. 2007. Mechanism of coupled 49. Nome, R. A., J. M. Zhao, ., N. F. Scherer. 2007. Axis-dependent folding and binding of an intrinsically disordered protein. Nature. anisotropy in protein unfolding from integrated nonequilibrium 447:1021–1025.

Biophysical Journal 102(9) 2149–2157