|Home | About | Journals | Submit | Contact Us | Français|
Conceived and designed the experiments: SL YZ ML. Performed the experiments: SL. Analyzed the data: SL YZ ML. Wrote the paper: SL YZ ML.
Allostery of P-selectin lectin (Lec) domain followed by an epithelial growth factor (EGF)-like domain is essential for its biological functionality, but the underlying pathways have not been well understood. Here the molecular dynamics simulations were performed on the crystallized structures to visualize the dynamic conformational change for state 1 (S1) or state 2 (S2) Lec domain with respective bent (B) or extended (E) EGF orientation. Simulations illustrated that both S1 and S2 conformations were unable to switch from one to another directly. Instead, a novel S1' conformation was observed from S1 when crystallized B-S1 or reconstructed “E-S1” structure was employed, which was superposed well with that of equilibrated S1 Lec domain alone. It was also indicated that the corresponding allosteric pathway from S1 to S1' conformation started with the separation between residues Q30 and K67 and terminated with the release of residue N87 from residue C109. These results provided an insight into understanding the structural transition and the structure-function relationship of P-selectin allostery.
Selectin-ligand interactions play a crucial role in inflammatory and immune responses via mediating leukocyte tethering to and rolling on vascular surfaces. Selectin is characterized by an extracellular C-type lectin (Lec) domain, a single epidermal growth factor (EGF)-like domain, and a variable number of short consensus repeat (SCR) units homologous to complement regulatory proteins, a transmembrane region, and a cytoplasmic tail . EGF domain is well-known to be functional in ligand recognition and cell adhesion , since Lec domain alone does support leukocyte adhesion but is insufficient for maximal binding whereas the concurrence of Lec and EGF (LE) domains constitutes the optimal recognition unit for leukocyte binding . Presence of SCR domains enhances the leukocyte adhesion via presenting sufficient length and flexibility for a P-selectin molecule to bind to its counterpart ligand, P-selectin glycoprotein ligand 1 (PSGL-1) . Particularly, an effective binding unit is formed by LE domains and SGP-3 peptide, a 19 N-terminal sulfoglycopeptide of PSGL-1 composed of three tyrosine sulfate residues Y605, Y607, and Y610, and a sLex-modified glycan at T616 . Despite the biological significance of LE domains is well defined , the structural bases of both Lec and EGF domains in regulating cell adhesion need to be further investigated.
Orientation and conformation of Lec and EGF domains are important to possess their functions. Comparison of crystallized structures of unliganded P-LE domains with liganded P-LE-SGP-3 or P-LE-sLeX (sialic acid X) complex make it possible to elucidate the structure-function relationship of P-selectin-PSGL-1 interactions at atomic level . For example, two distinct conformations of Lec domain with different orientations of EGF domain are visualized, one is so-called state 1 Lec domain (denoted as S1) followed by bent EGF domain (denoted as B) with a closed angle from EGF to Lec domain, and the other is state 2 Lec domain (denoted as S2) followed by extended EGF domain (denoted as E) with an open angle (cf. blue and silver newcartoons in Fig. 1A). The conformational difference between S1 and S2 Lec domain mainly lies in four specific regions (cf. H, R1, R2 and R3 highlighted as red circles in Fig. 1A). For example, R3 loop of S1 Lec domain (P81-D89) involved directly in ligand binding site parallels the binding interface and points away from the major binding sites while that of S2 Lec domain orients vertically outwards the binding interface and points toward those sites. Noting that B-S1 conformation appears in the unliganded P-LE or liganded P-LE with sLex and E-S2 conformation is found in the liganded P-LE-SGP-3 complex, it is evident that the conformation of Lec domain is associated with the orientation of EGF domain and that the allosteric pathways of conformational transition is expected. The structural analysis is also crucial to the binding capacity of a selectin molecule to PSGL-1 ligand under blood flow, since the conformational change of Lec domain is usually accompanied with the action of mechanical force .
Allostery is a common feature for biomolecular function observed in such the proteins as FimH , integrin , , von Willebrand Factor (vWF) , and myosin . P-selectin allostery was observed via introducing a glycan wedge between Lec and EGF domains  or mutating A28H residue near the hinge of Lec and EGF interface , which in turn enhanced the binding affinity to PSGL-1 ligand. An allosteric model was proposed based on crystallized structures to predict the pathway of P-LE domains from a low-affinity (the same as S1) to the high-affinity conformation (the same as S2)  where the conformational transition started with the destroy of interaction network among those residues locating at the interface between Lec and EGF domains and was terminated by the redistribution of hydrogen bonds (Hbond) between “switch2” and “switch3” (corresponding to respective R2 and R3 regions in the current study; cf. Fig. 1A). It still remains unknown, however, whether the allostery from S1 to S2 takes place directly and what are the underlying structural bases. Here we explored the allosteric dynamics of P-selectin Lec domain at atomic level to understand the structural evolution of P-selectin allostery. Impacts of Lec domain stability, EGF domain orientation, and SGP-3 ligation on conformational change of Lec domain were tested using molecular dynamic (MD) simulations.
The crystallized B-S1 structures of unliganded P-LE domains (PDB code: 1G1Q) and of sLeX-liganded E-LE domains (PDB code: 1G1T), and the E-S2 structure of SGP-3-liganded P-LE domains (PDB code: 1G1S) were employed as initial structures. Here a strontium ion Sr2+ was replaced by a calcium ion Ca2+ for the E-S2 structure. Every simulation system was built by solvating the target molecule into a rectangular water box and neutralized with ~100 mM Na+ and Cl− ions added to mimic the physiological ionic concentration, and was then equilibrated no less than 5 nanosecond (ns) using a NAMD program , a CHARMM22 all-atom force field for protein , and a self-built force field for six sugar residues and a tyrosine sulfate residue in SGP-3 ligand . An integration time step of 1 femtosecond (fs) and the periodic boundary conditions were applied in the simulations. Prior to equilibration process, energy minimization was initiated with 10000 steps by fixing backbone atoms of protein or heavy atoms of sugar followed by additional 10000 steps with all atoms free. System heating was then performed from 0 to 300 K at 30 K increment every 5 picosecond (ps). A smooth (10–12 Å) cutoff and the Particle Mesh Ewald (PME) method were employed to calculate van der Waals forces and full electrostatics, respectively. The 300 K heat bath was manipulated under Langevin thermostat, and the 1 atm pressure was controlled by Nosé-Hoover Langevin piston method.
Steered molecular dynamics (SMD) simulations were also conducted in some cases to test the impact of external force on conformational stability of Lec domain during P-LE unfolding or P-LE-SGP-3 complex dissociation. Here B-S1 structure of P-LE domains was forced to unfold using cf-SMD algorithm with a constant force of 100 pN along the vector from the fixed C-terminal atom L116-Cα of Lec domain to the pulled C-terminal atom D158-Cα of EGF domain. P-LE-SGP-3 complexes with different conformations were forced to dissociate using cv-SMD algorithm where C-terminal atom P618-Cα of SGP-3 ligand peptide was pulled via a spring with a spring constant of 70 pN/Å at a constant speed of 0.01 Å/ps along the vector from fixed C-terminal atom D158-Cα of EGF domain to the pulled end. Every system was equilibrated no less than 2 ns before SMD simulations.
Three types of structural analyses were conducted to illustrate the conformational characteristics of molecule of interest. The first was to quantify the orientation from EGF to Lec domain using the angle θ between two vectors connecting respectively the geometric center of heavy atoms of Lec-EGF interface hinge (residues A120 and S121) to that of Lec domain (residues W1 to T119) and to that of main EGF domain (residues C122 to T141) (magenta circles and lines in Fig. 1A). The next was to figure out the conformational change of Lec domain using the root of mean standard deviation (RMSD) of entire Lec domain or of specific regions, or the geometric center displacement of a residue by aligning the target to reference Lec domain, as well as the distance between specific residues. The final was to determine the interactions between Lec domain and SGP-3 ligand or protein and water molecules or among different regions of P-LE domains using the number of hydrogen bond (Hbond) with a donor-acceptor distance <3.5 Å and a donor-hydrogen-acceptor angle <45°. The system built-up and data analyses were performed using VMD program .
The B-S1 conformation of P-selectin in the absence of ligand or soaked with sLeX presented four distinct regions of Lec domain from those of E-S2 conformation with SGP-3 ligand , that is, the hinge region (denoted as H) of Lec and EGF interface involving two terminals of Lec domain (W1-Y3 and T119-A120), the loop region (denoted as R1) right ahead alpha helix 2 (I29-Q30), the anti-parallel beta sheets β4 and β5 and followed loop R54-N75 (denoted as R2), and the loop P81-D89 involving in ligand binding site (denoted as R3) (red circles in Fig. 1A). Here we focused on the conformational changes of the four regions and denoted the other regions of Lec domain as the rigid regions. The difference between S1 and S2 conformations was further quantified using the residue displacement by aligning S1 to S2 conformation, which was found to be >1 Å in all four regions (grey stripes in Fig. 1B).
The aforementioned differences between S1 and S2 conformations may lie in the following line of reasoning: the first is attributed to the conformational difference of Lec domain, the second is owing to the presence of SGP-3 ligand since the B-S1 structure is unliganded or sLeX-liganded and the E-S2 structure is SGP-3-liganded, and the third is referred to the orientation difference of EGF domain. To test the possibilities, five sets of simulations over eighteen runs were done upon crystallized unliganded and SGP-3-liganded P-LE and sLeX-liganded E-LE structures (Table 1), as described below.
Set I simulations were conducted to compare the conformational difference of Lec domain alone between the two structures (Table 1). Here EGF domain was deleted from crystallized B-S1 or E-S2 structure, and the two resulted structures (denoted as reference S1 and S2) were equilibrated for 16 and 20 ns, respectively (Fig. 2). To quantify the conformational changes, we first calculated the RMSD of Lec domain (Fig. 2A, C) and the displacement of heavy atom geometry center of each residue (Fig. 2B, D) via aligning the equilibrated snapshots to reference S1 (black) or S2 (red). As exemplified in Fig. 2A, the RMSD of equilibrated S1 structure exhibited a transition phase with ~1.5 Å at <6 ns followed by ~2.2 Å at >8 ns when aligning with reference S1 (black), while it yielded much high RMSD value ranging from ~2.8 Å at <6 ns to ~3.5 Å at >8 ns when aligning with alternative reference S2 (red). Averaged residue displacement calculated from last 2 ns equilibration presented the significant conformational change in the loop of R3 region, as exemplified by a maximal value of 10.1 and 17.4 Å for residue R85 (arrow), respectively (Fig. 2B). The slightly-high displacement was also found for the loop N56-K58 (arrow) between anti-parallel β4 and β5 sheets of R2 region. These data suggested that S1 Lec domain alone is not stable.
This was further tested by aligning the equilibrated S2 Lec domain to reference S1 or S2 for 20 ns simulation. Again, both RMSDs reached a high equilibrium value of ~2.8 Å for last 12 ns (black and red lines in Fig. 2C) with the large displacement of the residues at the R3 loop and R2 (Fig. 2D), supporting the above observation. Taken together, these results indicated that either S1 or S2 Lec domain is an unstable structure with the instability of R3 for the former or of both R3 and R2 for the latter. The conformational interchange between S1 and S2 was not observed in the simulations.
Next we tested the impact of presence of SGP-3 ligand on conformational difference of Lec domain. Set II simulations (Table 1) were done for 5 ns and the equilibrated conformation of S2-SGP-3 structure isolated from crystallized E-S2-SGP-3 complex was compared with that of S1-SGP-3 structure reconstructed via replacing S2 by S1 Lec domain. It was found that S1 Lec domain retained its original conformation with an equilibrated RMSD of ~1.3 Å aligned to reference S1 (black) but high RMSD of ~2.6 Å to reference S2 (red) (Fig. 3A). Combined with the observation of much smaller displacement for the former (the maximum of 2.4 Å; black line) than that for the latter (the maximum of 10.0 Å; red) (Fig. 3B), these data indicated that the presence of SGP-3 stabilized the conformation of S1 Lec domain (especially for R3 loop). Similarly, S2 Lec domain was also stabilized by SGP-3 ligation, as seen in an equilibrated RMSD of ~2.4 Å and ~1.3 Å to reference S1 (black) and S2 (red), respectively (Fig. 3D). Corresponding displacement profiles illustrated the same trend with small value to reference S2 (red) and large value to reference S1 (black) (Fig. 3E). Noting that the R3 is the most flexible region and that the residue R85 at the apex of R3 loop is the major binding site for SGP-3 ligand, our simulations demonstrated that a stable salt bridge between R85 and E617 of SGP-3 peptide was formed in S1-SGP3 complex (Fig. 3C) and that the interaction between R85 and sulfated tyrosine T610 of SGP-3 peptide was stably reserved in S2-SGP-3 complex (Fig. 3F). These results indicated that it is the binding of R85 with SGP-3 peptide that reduced the flexibility of R3 loop and stabilized its conformation. By contrast, the current simulations did not propose the allostery from S1 to S2 Lec domain when SGP-3 ligand was presented, suggesting that SGP-3 ligation is not directed to induce the allosteric transition of Lec domain. It was further noted that the structural stability induced by SGP-3 ligation was also observed when extending the simulations from 5 to 10 ns or even performing the 25-ns equilibrations for B-S1-SGP-3 and E-S2-SGP-3 complexes (data not shown), indicating these observations from the simulations were reliable.
EGF domain with residues S121-E158 distant from Lec-SGP-3 interface is necessary for optimal interaction between selectin and PSGL-1 ligand . Previous reports of the distinct EGF orientations in crystallized B-S1 and E-S2 structures  and the conformational transition of Lec domain by the presence of EGF domain  implied the potential impact of EGF orientation on conformation of Lec domain. To test the hypothesis, two systems of “E-S1” and “B-S2” reconstructed by interchanging B-EGF domain with E-EGF domain between two P-LE structures were equilibrated for 10 ns with constrained backbone atoms of interchanged EGF domain (Set III in Table 1). As exemplified in Fig. 4, EGF orientation was enhanced to ~149.0±2.9° for “E-S1” (Fig. 4A) but reduced to ~127.3±3.0° for “B-S2” Lec domain (Fig. 4C). Conformational analysis indicated that S1 Lec domain was no longer stable with dramatic structural change of R3 (Fig. 4B) and that the conformational change of S2 Lec domain locating at R3 and the ending loop of R2 (red) was intermediate (Fig. 4D). These results indicated that the re-orientation of EGF domain induced the conformational change of Lec domain but not the direct interchange between S1 and S2 conformations, since the equilibrated conformation of S1 or S2 Lec domain obtained from “E-S1” or “B-S2” system was evidently different from that of crystallized S1 or S2 Lec domain (Fig. 4).
Moreover, the equilibrated structures from either S1 Lec domain alone or “E-S1” with extended EGF domain exhibited the dramatic change of Lec domain conformation, as seen in the RMSD and displacement profiles (Figs. 2A–B and and4B),4B), which proposed the existence of a potential novel structure. To test this, the equilibrated structures were compared by superposing the equilibrated Lec domain from S1 alone (Fig. 5A, cyan) or “E-S1” (Fig. 5B, pink) with S1 (blue) and S2 (silver) references (Figs. 5A–B). The sharp difference in R3 loop demonstrated that the equilibrated structures were distinct from the references. Interestingly, the conformation of equilibrated Lec domain from “E-S1” structure (pink) was superposed well with that from S1 Lec domain alone (cyan) (Fig. 5C) and the resulted RMSD of R3 heavy atoms was ~2.8 Å (2nd open bar from the left in Fig. 6). This novel conformation, denoted as S1' Lec domain, was characterized by the R3 re-orientation via its apex around R85 followed by the counter-clockwise rotation and the slight deviation of R2 apex around N57 from S1 (Fig. 5A–B). Presence of S1' conformation in the two systems with sufficient long simulation time (>10 ns) imparted the confidence that this novel conformation is a stable structure. By contrast, the equilibrated conformation from “B-S2” with bent EGF domain was similar to that of S2 Lec domain except of loosing of R3 loop (data not shown). Thus, re-orientating from B- to E-EGF domain induced the conformational changes of S1 Lec domain on R3 and N-terminal of R2 but the re-orientating from E- to B-EGF just relaxed S2 Lec domain at the R3 and the C-terminal of R2.
The emergence of S1' conformation from S1 induced by the deletion of EGF domain or re-orientation from B- to E-EGF domain proposed the possible allostery between S1 and S1'. To further test the universality of S1' conformation and to elucidate the allosteric pathway from S1 to S1' Lec domain, Set IV simulations were performed, combined with the interchanged EGF orientations, by tuning those key residues of the entrance residues W1, A28, I29, and E34 of Lec domain, the pivot residue W62 of R2, and the ending residue D89 of R3, which seemed to be crucial in the interconversion between B-S1 and E-S2 (7). One series of simulations were done for “E-S1” structure combined with key residues tuned to those of S2 conformation (Fig. 6). Averaged RMSD of R3 heavy atoms to crystallized S1 (solid bars) or S1' conformation from equilibrated S1 Lec domain (open bars) was calculated from last 2 ns equilibration and the threshold of 3.0 Å was used to determine if the conformational change takes place or not. It was found that, as compared to the equilibrated S1 conformation from B-S1 or reconstructed “E-S1” structure (1st or 2nd set of bars from the left), the additional adjustment of residue W1 or D89 orientation tuned to that of S2 did not induce conformational change of S1 Lec domain (RMSD <3.0 Å, 3rd and 6th solid bars). The combined tuning of W1/A28/E34 or W1/A28/E34/W62, however, resulted in the conformational change (RMSD >3.0 Å, 4th and 5th solid bars), resulting in the conformations similar with that of equilibrated S1 Lec domain (RMSD 3.0 Å; open bars) (Fig. 6). All the resulted structures were distinct from that of S2 Lec domain (RMSD >3.0 Å, open bars in Fig. S1 of File S1.). In addition to the R3 RMSD calculated from aligning rigid regions of Lec domain and used to determine both the conformational change and the re-orientation of R3, we defined an alternative RMSD by aligning R3 itself. The calculations supported the above observation of the transition from S1 to S1' (Fig. S2 in File S1.). Thus, our simulations suggested that no allostery takes place from S1 to S2 conformation regardless of interchanging EGF domain and/or tuning key residues. Another series of simulations for “B-S2” conformation were also conducted by oppositely tuning same residues of S2 conformation to those of S1 conformation. All the R3 RMSD was found to yield >3.0 Å when aligning the rigid regions to those of crystallized S1 (solid bars) or S2 (open bars) conformation (Fig. S3 in File S1.). Combined with the RMSD of R3 itself (Fig. S4 in File S1.), no allostery but only the loosing of R3 was observed in S2 Lec domain. Taken together, these results suggested that the interaction network attributed to W1 or D89 residue may not be prerequisite for allostery of Lec domain, as predicted previously . It was further confirmed that the interchanged allostery between S1 and S2 structure did not take place directly and that S1' conformation could be a universal allosteric conformation of Lec domain originated from S1.
Then, what's the underlying allosteric pathway from S1 to S1' conformation? Further analyses demonstrated that the allostery was initiated by the breakage of water bridge among Q30 of R1, K67 of R2, and E135 near H region and resulted in extending EGF domain (pink newcartoon and CPK residues in Fig. 7A'). Originally the water bridge networks between Q30 and K67 and between Q30 and E135 in B-S1 structure (blue newcartoon and licorice residues in Fig. 7A) formed a closed cavity with several water molecules tightly linking R1 and part of R2 (W62-K67) (Fig. 7B). The disruption of water bridge near H region induced the opening of cavity gate by separating side-chains of K67 and Q30 and swallowed additional water molecules (Fig. 7B'). It also weakened the interaction between R1 and R2 loop (W62-K67) and pushed rotationally R2 away from original location (Figs. 7C and 7C'). The motion of R2 was then transmitted to the loop (H108-K113) through the β7 and β8 sheets via strong Hbond interaction between R54 of R2 and D89 of R3 and the disulfated bond between C90 and C109, which, in turn, pushed C109 residue away (Figs. 7D and 7D') and disrupted the interaction of C109 backbone Hbond with N87 of R3. The release of N87 from C109 resulted in the R3 re-orientation and, finally, terminated the allostery of Lec domain (Figs. 7E and 7E'). In short, the destroy of hinge water bridge induced the opening of cavity gate governed by Q30-K67 interaction and weakened the interaction between R1 and part of R2 (W62-K67), followed by the re-orientation of R2 transferred to C109 via the interaction of R2 N-terminal and R3 C-terminal as well as the rigid C90-C109 disulfated bond, which finally activated the conformational change of R3 through the breakage of N87-C109 Hbond. Such an allosteric pathway was further confirmed by the following correlations: The re-orientation of R2 (defined by RMSD of part of R2 (K55-W60)) was reinforced with width of cavity gate (defined by distance between two atoms of Q30OE1 and K67NZ) (Fig. 7G); The breakage of N87-C109 Hbond (denoted as distance between two atoms of N87O and C109N) was enforced by R2 re-orientation (Fig. 7H); And the allostery of R3 (defined by RMSD of R3) correlated positively with N87-C109 Hbond disruption (Fig. 7I).
Simulations of S1 Lec domain by deleting EGF domain (Fig. 2A, B) and reconstructed “E-S1” structure by reorienting EGF domain from bent to extended conformation combined with tuning the key residues (Figs. 4A–B, ,6,6, and S1–S2 in File S1.) demonstrated the stability and university of S1' Lec conformation. It is still possible, however, that the novel conformation was imposed seemingly by manually deleting the EGF domain or modulating its orientation. To exclude the possibility, additional simulations were performed on crystallized B-S1 and E-S2 structures without any constraints (Set V in Table 1). Similar to those for S2 Lec domain (Figs. 2C–D) and reconstructed “B-S2” structure by reorienting EGF domain from extended to bent conformation combined with tuning key residues (Figs. S3–S4 in File S1.), 30 ns equilibration of E-S2 exhibited similar relaxation of R3 loop without intrinsic conformational change (1st sets of bars in Figs. S3–S4 in File S1.). 45 ns equilibration of B-S1, however, demonstrated the spontaneous change of Lec domain conformation as well as the reorientation of EGF domain after ~5 ns fluctuation around B-S1 structure (Figs. 8A–B). Specifically, the Lec domain started to deviate from S1 for ~15 ns transition (7.5–20.5 ns) and then evolved into S1' around 25 ns, with the RMSD of R3 heavy atoms to S1 (black) and S1' (red) running up to ~6.2 Å and down to ~2.8 Å, respectively (Fig. 8A). In contrast to the S1' conformations resulted from EGF deletion or reorientation, the S1' conformation spontaneously obtained from B-S1 equilibration was unstable sufficiently, which finally evolved away into other conformations with high RMSD of R3 fluctuating to ~5.8 Å around 31 ns and even higher value during last 3 ns (Fig. 8A, red). EGF orientation also exhibited the multiple phases with more than three times of transition between bent (~122°) and extended (~138°) conformations (Fig. 8B). Thus, the novel S1' conformation could exist spontaneously with less stability.
Crystallized Lec-EGF domains of E-selectin liganded with sLeX presented the similar characteristics with B-S1 of P-selectin in both Lec conformation and EGF orientation (6). Here a 45 ns simulation for E-LE domains was performed to further test the universality of S1' allostery from S1 in different selectin members. The RMSD of R3 heavy atoms relative to P-selectin S1 (black) and S1' (red) conformation indicated that E-selectin Lec domain also experienced the spontaneous allostery to S1' after ~15 ns fluctuation around S1, where the allostery happened sharply and the S1 and S1' conformations were ultimately distinct (Fig. 8D). E-selectin EGF domain re-oriented to the high angle from the initial 120° and presented more extended angles with the maximum of ~165° at ~30 ns than that of P-selectin (Fig. 8E).
We further correlated the allostery of Lec domain to S1' with the reorientation of EGF domain to extended states, as observed in the simulations of P- and E-LE domains. The distribution of EGF orientation of S1' conformation (red) overlapped partially to that of S1 conformation (black) in both P- (Fig. 8C) and E- LE domains (Fig. 8F), suggesting that not all the large (or small) Lec-EGF angles are necessarily corresponding to S1' (or S1) conformation. It was also found that the distribution shifted rightward from S1 to S1' conformation, implying that S1' conformation favors more extended EGF orientation and that the extension of EGF orientation promotes the possibility of allostery from S1 to S1'. Moreover, the slight (Fig. 8C) or dramatic (Fig. 8F) shifting in the distribution was well correlated with the transient (Fig. 8A) or sharp (Fig. 8D) difference in the RMSD for P- or E-LE domains. Taken together, our simulations indicated that allostery from S1 to S1' could happen spontaneously, which is universal for both P- and E-selectin.
Allostery is essential for a protein to present different functional states. For example, catch bond behavior (bond lifetime increases with applied force) of bacterial FimH was attributed to the conformational change of Lec domain induced by pilin domain , . Allostery of interdomain region of vWF A1 domain possibly regulated catch bond behavior with platelet glycoprotein Ib (GPIbα) , . Activation of headpiece of integrin α5β1 contributed to the presence of catch bond with fibronectin ligand . Specifically, catch bond nature of a selectin molecule, first visualized in the forced dissociation of P-  or L-selectin  from PSGL-1, was correlated biologically with a shear-threshold feature for leukocyte tethering and rolling adhesion mediated by selectin-ligand interactions , . An allosteric model for catch bond of P-selectin-PSGL-1 interaction was proposed from analyzing the two crystallized P-LE structures, which assumed that the high-affinity conformation of P-LE (S2) induced by forced allostery from its low-affinity conformation (S1) is favored to its ligation . While a few measurements confirmed the occurrence of P-selectin allostery , , it is still unknown if the interchanged transition between S1 and S2 takes place directly along the predicted pathway and what the dynamic pictures are. Thus, MD simulations of P- and E-LE were performed in the current study, attempting to visualize the dynamic pathways of structural allostery for Lec domain. By elucidating the impacts of the stability of Lec domain alone, the ligation of SGP-3 ligand, the orientation of EGF domain, and the stability of LE domains, our results indicated that the allostery of Lec domain existed and the orientation of EGF domain induced the conformational change of Lec domain. Three novel observations were found in the simulations: the first is that the interchanged transition between S1 and S2 was unable to take place directly but a stable, novel conformation (S1') was presented originating from S1, the next is that the conformation of equilibrated S1 Lec domain alone was similar with that of equilibrated S1' conformation when EGF domain for S1 was tuned from B to E orientation (Figs. 4 and and5),5), and the final is that the allostery from S1 to S1' and the reorientation of EGF domain could happen spontaneously where the S1' conformation favors the extended EGF orientation (Fig. 8).
At least three lines of evidence were proposed from experimental measurements for the co-existence of multiple conformations of P-Lec EGF domains. First, both the slow and fast phases concurrently exist in the dissociation kinetics for a wild-type P-selectin construct (in the order-of-magnitude of ~10−1 and ~10−2 s−1) or a wedge mutant that opens the interface between Lec and EGF domains (in the magnitude of ~10−2 and ~10−3 s−1) , suggesting that at least three kinetic phases exist to correspond to three stable conformations of the protein containing Lec, EGF and SCR1 domains. Similar measurements were also obtained for a cleft P-selectin mutant with Lec, EGF, and SCR1-2 domains that opens within Lec domain  where the kinetic model used to probe the two equilibrated conformations (bent and extended) for binding to and dissociating from the chip surface is no longer applicable to estimate the reliable parameters, indicating that more than two P-selecitn conformations may exist. Second, not only the existence of EGF domain is pre-requisite for sufficient binding of P-selectin to its liagnd , , but the substitution of EGF domain of L-selectin with the homologous domain from P-selectin also enhances the binding to L-selectin ligand under shear flow while its equilibrium features toward soluble ligands remains the same , implying that the cooperativity of Lec and EGF domains promotes different kinetic phases. Third, a recent report on a triphasic force dependence of lifetime of E-selectin-ligand bond (personal communications) further supported that multiple stable conformations of P-selectin be expected from the experiments. Several testable predictions such as crystallizing the S1' structure or characterizing the high-affinity phase for P-LE wedge cleft can be proposed to elucidate the biological significance of the novel S1' conformation in future studies.
Dynamic allosteric pathways found in the current study were also different from those proposed by comparing the conformational differences of crystallized B-S1 and E-S2 structures . In the previous prediction, the extending from B- to E-EGF orientation induced directly the allostery of Lec domain from S1 to S2 via sequential interactions along the following pathway: First, the extending orientation broke up the interaction network near H region between EGF and Lec domains, followed by the disruption of interaction among residues W1, A28, and E34 via a water molecule. Next, the resulted re-orientation of R1 loop drove the tuning of W62 side-chain and then pushed the anti-parallel β-sheets of R2 away to re-distribute of hydrogen network of R54-D89 and K55-N83 between R2 and R3. Finally, the re-orientation of R3 induced the conformational transition from S1 to S2. Our MD simulations, however, proposed a distinct allosteric pathway. Here the side-chain re-orientation of residues W1 and R89 (the two key residues assumed to be responsible for the allostery ) did not induce the conformational change of Lec domain from S1 to S2 or S1', indicating that the breakage of either the interaction network of W1 around H region or the linkage between R2 and R3 via R89 was not required for allostery. By contrast, re-orientation of residue W62 (the pivot residue involved in the allostery ) promoted the conformational change from S1 to S1', implying that this residue was engaged in the allosteric pathways via pushing R2 away sufficiently. This alternative pathway was consistent with the previous observations that both the A28H mutation  and the introduction of a glycan wedge between EGF and Lec domains  resulted in the allostery of P-selectin by destroying the closed cavity between R1 and R2 loops (W62-K67). Conclusively, the five sets of simulations (~275 ns totally) defined well the specificity and stability of S1' conformation and the allosteric analyses indicated that the novel S1' conformation takes place along a different pathway.
While the allostery of Lec domain from S1 to S1' could happen universally, is it also possible to further induce the conformational change from S1' to S2? PSGL-1 ligation is one of potential pathways since the ligation of Lec domain makes it possible to alter the conformation of Lec domain. For example, the fluorescent intensity of labeled P-selectin was enhanced up to 10–13% when binding to PSGL-1 ligand , implying that the conformational change might occur within P-selectin molecule. Regardless of that S2 conformation was stabilized by the presence of SGP-3 ligand (Fig. 3), however, our results seemed not to support the conformational transition from S1 or S1' to S2 in the equilibration simulations of S1, B-S1, “E-S1”, or “E-S1'” when ligated to SGP-3 (Figs. 3, and S5 in File S1.). Another possibility lie in the exertion of applied force that may be required to induce the allostery to S2 in the forced dissociation of Lec-SGP-3 complexes . Again, the expected transition to S2 did not take place in our SMD simulations of B-S1-SGP-3 or E-S1'-SGP-3 complex (Fig. S6 in File S1.). Instead, the S1 conformation could evolve into S2-like via S1' conformation in the forced unfolding of EGF domain probably due to the resulted large Lec-EGF angle (cf. supplementary text and Fig. S7 in File S1.). In fact, the prolonged equilibration simulation of B-S1 exhibited the similar transition at the end phase of simulation (~43–45 ns) (Fig. 8A–B). Noting that the allostery from S1' to S2-like conformation followed the extension of EGF orientation, it is reasonable to assume that the allostery from S1 to S1' and further to S2-like conformation could happen spontaneously and the three conformations favor different EGF orientations where the S2-like conformation corresponding to the most extended EGF domain.
Our findings of an alternative allostery of P-selectin Lec domain also supported the prediction that the presence of EGF domain regulated the P-selectin-PSGL-1 binding by changing the conformation of Lec domain upon EGF orientation , where the S1' conformation favored the extended EGF orientation than that of S1 conformation (Fig. 8). This allosteric model can be used to interpret the catch bond nature since high force promotes strong and long-lived P-selectin-PSGL-1 interaction by turning EGF domain to extended orientation and then inducing the corresponding allostery of Lec domain. Besides, our simulations were not inconsistent with another sliding-rebinding model of catch bond where the forced opening of interdomain H hinge promoted the formation of new interactions to slow down unbinding and to prolong bond lifetime . While the sliding-rebinding model and the corresponding measurements  supported that the flexibility of interdomain hinge and the extension of EGF-Lec angle tilted the binding interface to the direction of external force and allowed the two contact sides sliding against each other, our simulations proposed the possibility for conformational change of Lec domain upon EGF re-orientation to alter the interaction. Future integration of the three structural or allosteric models is required to elucidate the intrinsic mechanisms of conformational changes in dominating P-selectin-PSGL-1 interaction.
Finally, the current study provided the dynamic pictures of P-selectin EGF and Lec structures to visualize the conformational change of Lec domain from the viewpoints of EGF orientation, interaction network of EGF-Lec interdomain, and interaction between P-LE and SGP-3 ligand. It was indicated that S1 and S2 structures are unable to interchange directly but able to go through a novel conformation of S1' Lec domain with an alternative allosteric pathway. Our results furthered our understanding in the structure-function relationship of P-selectin.
Supplementary Text and Supplementary Figures of S1-S7.
MD simulations were performed using Computational Facility, Institute of Mechanics, and Supercomputing Center, Chinese Academy of Sciences (CAS).
Competing Interests: The authors have declared that no competing interests exist.
Funding: This work was supported by National Natural Science Foundation of China grants 10702075, 30670172, 30730032, and 10332060 (http://www.nsfc.gov.cn/Portal0/default106.htm), National Key Basic Research Foundation of China grant 2006CB910303 (http://www.most.gov.cn/eng/index.htm), Knowledge Innovation Program of CAS grants KSCX2-YW-N-011 and KJCX2-YW-L08 (http://english.cas.cn/), and National High Technology Research and Development Program of China grant 2007AA02Z306 (http://www.most.gov.cn/eng/programmes1/200610/t20061009_36225.htm). The funders had no role in study design, data collection and analysis, decision to publish or preparation of the manuscript.