Allosterism in human complement component 5a (hC5a): a damper of C5a receptor (C5aR) signaling

The phenomena of allosterism continues to advance the field of drug discovery, by illuminating gainful insights for many key processes, related to the structure–function relationships in proteins and enzymes, including the transmembrane G-protein coupled receptors (GPCRs), both in normal as well as in the disease states. However, allosterism is completely unexplored in the native protein ligands, especially when a small covalent change significantly modulates the pharmacology of the protein ligands toward the signaling axes of the GPCRs. One such example is the human C5a (hC5a), the potent cationic anaphylatoxin that engages C5aR and C5L2 to elicit numerous immunological and non-immunological responses in humans. From the recently available structure–function data, it is clear that unlike the mouse C5a (mC5a), the hC5a displays conformational heterogeneity. However, the molecular basis of such conformational heterogeneity, otherwise allosterism in hC5a and its precise contribution toward the overall C5aR signaling is not known. This study attempts to decipher the functional role of allosterism in hC5a, by exploring the inherent conformational dynamics in mC5a, hC5a and in its point mutants, including the proteolytic mutant des-Arg74-hC5a. Prima facie, the comparative molecular dynamics study, over total 500 ns, identifies Arg74-Tyr23 and Arg37-Phe51 “cation-π” pairs as the molecular “allosteric switches” on hC5a that potentially functions as a damper of C5aR signaling.

In cellular network, protein-protein interactions are generally allosteric in nature (Nussinov, Tsai, & Ma, 2013), as the interacting proteins demonstrate a substantial change in their energy landscape (Swain & Gierasch, 2006;Uversky, 2013). In particular, allosterism has been associated with numerous ID proteins, including globular proteins with an ID region (Tompa, 2012), that are often targeted for their promiscuous role in modulating the cell cycle and signaling through the protein-protein interaction network (Ferreon, Ferreon, Wright, & Deniz, 2013;Hilser & Thompson, 2007;Xie et al., 2007). What is noteworthy that for information delivery, most of these signaling proteins undergo a series of order-disorder and vice versa transition at different time scales, which are subtly linked and controlled by specific long-range allosteric switches (Habchi et al., 2014;Uversky et al., 2008).
Under this background, it is intriguing that while the phenomenon of allosterism is well discussed for the membrane-embedded G-protein coupled receptors (GPCRs) (Christopoulos & Kenakin, 2002;Kenakin & Miller, 2010;Wang & Lewis, 2013), its role in the interacting protein ligands targeting the GPCRs has never been discussed in the literature. Human C5a ( h C5a) is one such cationic glycoprotein ligand (Klos, Wende, Wareham, & Monk, 2013), which is known to interact with two GPCRs, such as C5aR and C5L2 (Guo & Ward, 2005). While allosterism in C5aR has been discussed in the literature (Moriconi et al., 2014), it is yet to be discussed in h C5a. The h C5a-C5aR interaction is of therapeutic importance, as it elicits a plethora of physiological responses ranging from chemoattraction, multiple organ failure, ischemia, and reperfusion injury to apoptosis (Guo & Ward, 2005). The complete molecular interaction is not structurally established yet, though a highly refined model structure of C5aR in complex with a linear peptide agonist based on the C-terminus (CT) of h C5a has recently been described in the literature (Rana & Sahoo, 2015). It is well established that the mutation in the CT region of h C5a Toth et al., 1994), including the des-Arg 74 -h C5amimic of the proteolytic mutant of h C5a, produced in plasma - (Cain, Coughlan, & Monk, 2001;Higginbottom et al., 2005) down regulates both C5aR binding and signaling. In addition, mutational studies have also identified several pharmacophores on the core domain of h C5a that appreciably down regulates both C5aR binding and signaling Toth et al., 1994). However, unlike the CT of h C5a, the intermolecular interaction between the pharmacophores on the core domain of h C5a with the complementary residues on C5aR is yet to be established clearly. More importantly, the overall effect of such point mutations on the structure and dynamics of h C5a have not been explored, especially when it is suggested that chemical or metabolic changes (Wenthur, Gentry, Mathews, & Lindsley, 2014) can activate the "allosteric switches" on the protein ligands that can recruit distinct conformational microstates of GPCRs for modulating their pharmacological actions in tissues (Christopoulos & Kenakin, 2002;Wang & Lewis, 2013).
Nevertheless, while purification to homogeneity remains challenging (Renfer et al., 1986) for the plasmaderived native h C5a, the availability of the recombinant h C5a has advanced the structural studies (Cook, Galakatos, Boyar, Walter, & Ealick, 2010;Schatz-Jakobsen et al., 2014;Zhang, Boyar, Toth, Wennogle, & Gonnella, 1997). The recent structural studies on des-Arg 74 -h C5a indicate that the deletion mutation of Arg 74 induces a major conformational heterogeneity in h C5a, leading to the dimerization of des-Arg 74 -h C5a in crystal (Cook et al., 2010), where Chain A displays a four-helix bundle (RMSD~1 .48 Å) and the Chain B displays a three-helix bundle (RMSD~1.17 Å) topology (Figure 1), distinctly different from the conformation observed in the wild-type h C5a (Figure 2(a)). Interestingly, the Chain B displays a conformation (RMSD~.73 Å) that is incredibly similar to that of h C5a-A8 (Figure 2(b)), the C5aR antagonist (Schatz-Jakobsen et al., 2014) engineered from h C5a with 90% sequence identity. In contrast, no such conformational heterogeneity is observed between the crystal structures of m C5a (Figure 2(c)) and des-Arg 79 -m C5a (RMSD .36 Å, Supplementary Figure S1(a)) (Schatz-Jakobsen et al., 2014), including the h C3a and des-Arg-h C3a (RMSD~.29 Å, Supplementary Figure S1(b)) (Bajic, Yatime, Klos, & Andersen, 2013). Interestingly, h C5a is also not known to dimerize in solution. As evidenced in solution NMR studies (Zhang et al., 1997), the 74 amino acids in h C5a are coded to fold into a monomeric four-helix bundle structure intertwined by three flexible linker regions (L 1 : Tyr 13 -His 15 , connects helix 1 and 2; L 2 : V 28 -C 34 , connects helix 2 and 3; and L 3 : Arg 40 -Gly 44 , connects helix 3 and 4). Among the four helices, the helix 3 (Glu 35 -Ala 39 ) in h C5a is flanked between the flexible L 2 and L 3 regions and is the shortest one in length. In addition, the last 10 residues (Ala 63 -Arg 74 ) on the CT of h C5a display a partially ordered structure with a 4-residue (Gln 71 -Arg 74 ) helical turn. The h C5a harbors an unpaired cystiene (Cys 27 ) on the helix 2, beside the three disulfide linkages that hold the helices 2, 3, and 4 as a bundle, against the helix 1 through moderate hydrophobic interactions. Similarly, the m C5a that share~63% sequence identity with h C5a also display a monomeric four-helix bundle topology (backbone RMSD~1.12 Å) in crystals (Schatz-Jakobsen et al., 2014), held together by three disulfide linkages, including strong hydrophobic interactions between helix 1 and others (Figure 2(c)). It is noteworthy that the helix 3 in m C5a is of equivalent length to that of helix 2 on h C5a-A8. Interestingly, h C5a-A8 do not harbor an unpaired Cys 27 on the elongated helix 1, as seen on the helix 2 of h C5a, instead carries Arg 27 , similar to the Arg 32 found on the helix 2 of m C5a. However, unlike the unresolved CT in the m C5a, the h C5a-A8 (Figure 2(b)) displays an extended CT structure that is in contrast to the ordered CT of h C5a (Figure 2(a)).
In light of these structural studies, it is quite possible that other pharmacophores previously identified in the core region of h C5a may also confer similar subtle conformational changes, which might have a role in negatively regulating the function of h C5a. From structural perspective, it can be a challenge to capture the functionally regulatory, subtle conformational changes involved in the dynamics of information delivery over a distance in h C5a. Even sometimes, the preliminary CD spectroscopy studies on the mutant h C5a can be largely inconclusive (Bubeck et al., 1994). Nonetheless, barring des-Arg 74 -h C5a, no such detailed structural studies on the other known pharmacophores of h C5a have been undertaken so far. In fact, the structural studies (Cook et al., 2010) also do not provide sufficient insight to decipher the possible allosteric phenomena in h C5a in solution and its role in C5aR signaling (Moriconi et al., 2014). This provides the necessary momentum and the opportunity to study the structure-function relationship in the pharmacophores of h C5a by implementing the other available alternative techniques.
In this study, we have subjected the m C5a, h C5a, and its mutants to the conformational dynamics studies, by recruiting the all-atom molecular dynamics (MD) simulation techniques in explicit water, over a total 500 ns. The study attempts to address the following two concerns (1) what is the molecular basis of dimerization in des-Arg 74 -h C5a and whether allosterism has a role to play in the dimerization, and (2) whether the loss of binding or signaling between the mutant h C5a and C5aR − in lieu of point mutations in h C5a − is absolutely due to the absence of interacting residues or there is also an unknown allosteric component that collectively works as a damper of C5aR binding and signaling.

Materials and methods
PDB coordinates of the h C5a and others were downloaded from www.rcsb.org, as follows. 1KJS: h C5a; 3HQA: des-Arg-h C5a; 4P3A: m C5a; 4P3B: des-Arg-m C5a; 4P39: h C5a-A8; 4HW5: h C3a; and 4HWJ: des-Arg-h C3a. PyMOL (The PyMOL Molecular Graphics System, Version 1.2r1, Schrödinger, LLC) and Discovery studio (Accelrys) softwares were utilized for visualization, analysis, and presentation of the C5a structures. Sequence alignments of C5a were achieved using ClustalX (Larkin et al., 2007). Data were plotted in GraphPad Prism (version 6 for Windows, GraphPad Software, La Jolla California USA, www.graphpad.com). MD simulation of the h C5a and its variants were carried out at 300 K for over 50 ns each, following the procedures described previously (Joshi, Rana, Wangikar, & Durani, 2006;Rana & Sahoo, 2015). Briefly, the polar amino acids such as Lys and Arg were positively charged, whereas amino acids such as Asp and Glu were negatively charged, in all the systems, which were appropriately neutralized, by randomly placing requisite number of ions to the box prior to the energy minimization. For instance, under our simulating conditions, h C5a displayed + 5 charge, which was neutralized by placing 5 chloride ions in 7284 water molecules. Similarly, the m C5a also displayed + 5 charge, which was neutralized by placing 5 chloride ions in 6567 water molecules. The h C5a and its other variants were subjected to careful energy minimization in a cubic box with periodic boundary, recruiting the gromos-96 43a1 force field in GROMACS package (Hess, Kutzner, van der Spoel, & Lindahl, 2008). Placed in the center of a periodic box, the h C5a and others were energy minimized to 100 kJ mol −1 nm −1 tolerances with steepest descent, first in vacuum and then appropriately surrounded with simple point charge water model, as provided in GROMACS. Numerical integrations were performed in step size of 2 fs and coordinates were updated every 5 ps. Solvent density was set to the value corresponding to 1 atm at 300 K. Protein and solvent were coupled independently to Berendsen bath at 300 K, to the coupling time constant .1 ps. Protein was position restrained, while solvent was relaxed over 200 ps. Production MD was initiated thereafter, and the trajectory was sampled at 5 ps intervals. "Cation-π" interactions were calculated implementing the following cutoff (d ≤ 6 Å; θ < 90°) as described elsewhere (Marshall, Steele, Thanthiriwatte, & Sherrill, 2009;Rana & Sahoo, 2015). Briefly, the "cation-π" angles were obtained by recruiting the in-house program, which takes PDB files as input and then calculates the angle between the two normal, respectively, projected on the centroids of the aromatic ring and the guanidinium group of the arginine. The centroid-to-centroid distance, representing the "cation-π" interaction, was calculated using the inbuilt programs in GROMACS. Conformational clustering was performed (every 5 ps) as described, to the RMSD cutoff ≤1. 2 Å for the trajectories in water unless otherwise mentioned. Clusters having ≥100 conformers were considered as major microstates/clusters. The representative conformer of the major microstate, evolved over 50 ns of MD, was used for generating the Arg 37 and Arg 74 mutants of h C5a. Helix-to-coil transition in helix 3 was monitored every 100 ps interval over the duration of MD, under the guidance of Ramachandran diagram. The helix 3 is assigned a helical conformation if ≥50% residues accessed the α L conformations. Analyses of the trajectories were achieved by implementing the modules built into GROMACS.

Results
Probing the stability and dimerization switch in des-Arg 74 -h C5a As evidenced (Figure 1), the Chain A and Chain B are held together by five specific intermolecular interactions in the dimer of des-Arg 74 -h C5a (Cook et al., 2010). It is noteworthy that des-Arg 79 -m C5a does not display dimeric structures. More importantly, the dimers in des-Arg 74 -h C5a are also not covalently linked via the unpaired Cys 27 on h C5a, which suggests there may be an unidentified switch that triggers the dimerization of des-Arg 74 -h C5a in crystals. To probe the role of those interactions in the overall stability, the dimer was subjected to MD simulation in explicit water for over 50 ns at 300 K. As summarized in Figure 3, the hydrogen bonds, respectively, between the Lys 19 of Chain A and Ala 39 , Ile 41 of Chain B were maintained over 50 ns. However, among the two, the hydrogen bonding between Lys 19 of Chain A and Ala 39 of Chain B displays modest stability. On the other hand, out of the three salt bridge interactions (Donald, Kulp, & DeGrado, 2011), only one salt bridge between Glu 53 of Chain A and Lys 49 of Chain B was maintained over the duration of MD. Further, out of the two hydrophobic contacts (Bursulaya & Brooks, 1999) plotted between Chain A and B, the interaction between the Leu 61 of Chain A and Val 57 of Chain B appears modestly stable. Overall, though the dimer did not dissociate into monomers over 50 ns of MD, considering the moderate stability of the intermolecular interactions in water, it is quite possible that the dimers might dissociate into monomers over a longer time scale. Further, it is interesting to note that the Lys 14 and His 15 , respectively, in Chains A and B access the opposite quadrants (Supplementary Figure S2) of the Ramachandran space (Ramachandran & Sasisekharan, 1968). In Chain A, the Lys 14 access the β-sheet region, whereas it is mostly localized in the α R -helix region in Chain B. Similarly, the His 15 is, respectively, populated in the α L -helix region in Chain A, and the α R -helix region in Chain B. Thus, it appears that perhaps the Lys 14 and His 15 in L 1 region play a significant role in conformational heterogeneity or structural allosterism in h C5a, which might be influencing the dimerization of des-Arg 74 -h C5a in crystals. To further probe the role of L 1 region residues toward allosterism, the monomers of the des-Arg 74 -h C5a were subjected to comparative MD studies with monomers of h C5a and m C5a.

Conformational dynamics of h C5a compared to m C5a and des-Arg 74 -h C5a monomers
To probe whether the inherent conformational dynamics has any role in the dimerization process, both h C5a and m C5a, including the Chain A and Chain B of des-Arg 74 -h C5a were subjected to comparative MD studies, over 50 ns each in explicit water. Over the duration of MD, neither h C5a nor m C5a were able to access the conformational states demonstrated by the Chain A or Chain B in des-Arg 74 -h C5a dimer (Figure 1). In agreement, the results summarized in Figure 4 indicate that both h C5a and m C5a harbor a stable topology (Supplementary Figure S3). However, interestingly, the Lys and His residues in L 1 region of h C5a and m C5a demonstrate a significant difference in their overall conformational accessibility (Supplementary Figure S4). Grossly, while Lys 14 on h C5a accessed all the four quadrants (β-sheet, α L -helix, flat ribbon, and α R -helix) with a modest presence in the β-sheet region, the Lys 19 on m C5a predominantly accessed the α R -helix and flat ribbon quadrants of the Ramachandran plot. On the other hand, the His 15 on h C5a accessed both the β-sheet and α L -helix quadrants, whereas the His 20 on m C5a preferred to be localized, only on the β-sheet quadrant of the Ramachandran plot. In addition, as evidenced in Supplementary Figure S2 Thus, the Chains A and B of des-Arg 74 -h C5a were independently subjected to MD studies over 50 ns and further compared with the h C5a. No significant conformational change was noticed for the isolated Chain B of des-Arg 74 -h C5a dimer (Supplementary Figure S6). However, the isolated Chain A of des-Arg 74 -h C5a dimer displayed further folding (Figure 5), demonstrating the movement of helix 1 to form a similar four-helix bundle topology, matching appreciably with the h C5a (Figure 2(a)). In support, the Lys 14 and His 15 on Chain A also displayed similar conformational accessibility, as observed in h C5a (Supplementary Figure S7(a) and (b)). Interestingly, deletion of the Arg 74 from the NMR structure of h C5a also abolished the conformational access of Lys 14 from the α Rhelix region (Supplementary Figure S7

Identifying the allosteric switch region on h C5a
In agreement with the crystal structures (Cook et al., 2010), but in contrast to the NMR structure (Zhang et al., 1997), the central conformer of the major microstate of h C5a (Figure 4(a)) demonstrates an unordered CT (Ile 65 -Arg 74 ). The Arg 74 side chain appears flexible (Supplementary Figure S3(c)) in h C5a, as it did not sustain the hydrogen bonding over 50 ns of MD, observed with the Leu 72 and Leu 2 (helix 1) in the energy minimized NMR structure (Supplementary Figure S8). However, further analysis evidenced that the Arg 74 is involved in a fluctuating, but strong "cation-π" interaction (Ma & Dougherty, 1997;Marshall et al., 2009) (d ≤ 6 Å; θ < 90°) with the Tyr 23 on helix 2 (Figure 6(a)) over the duration of the MD simulation. Interestingly, it is also evidenced that the Tyr 23 is involved in a network of hydrogen bonding with both Gln 3 and Glu 7 on helix 1 (Figures 2(a), 6(b) and (c)). Such hydrogen bonding of Tyr 23 could be crucial in modulating the conformational space of the Lys 14 and His 15 residues in L 1 region, which helps in maintaining the four-helix bundle structure of h C5a, by forming a helical hairpin between the helix 1 and helix 2. It is noteworthy that the "cation-π" interaction between the Arg 74 -Tyr 23 can strongly modulate the rotameric fluctuations of Tyr 23 , which ultimately help in maintaining the hydrogen bonds between the helix 2 and helix 1 of h C5a. It is clear that deletion of Arg 74 results in the loss of "cation-π" interaction, enhancing the rotameric fluctuation in Tyr 23 (Supplementary Figure S9  . Role of Tyr 23 in topological stability of h C5a. (a) Strong, but fluctuating "cation-π" interaction observed between the Arg 74 and Tyr 23 in h C5a over 50 ns of MD. The dotted line represents the distance cutoff for the "cation-π" interactions. The "cationπ" angle (mean ± SD) is also provided as insert. (b) Strong hydrogen bonding noted between the side chains of Tyr 23 (helix 2) and Gln 3 (helix 1) of the h C5a. (c) Strong hydrogen bonding observed between the side chains of Tyr 23 (helix 2) and Glu 7 (helix 1) of the h C5a. and Glu 7 . Thus, it is not surprising that the deletion of Arg 74 leads to a three-helix bundle structure (Figure 1) in des-Arg 74 -h C5a (Cook et al., 2010), which undergoes dimerization in crystal. On the other hand, similar "cationπ" interactions between the Tyr 28 −Arg 79 cannot be established in m C5a, due to the unresolved CT. However, over 50 ns of MD, the Tyr 28 (Figure 2(b)) also demonstrated a consistent hydrogen bonding with the Glu 12 on helix 1 (Supplementary Figure S9(b)), similar to the one observed between Tyr 23 and Glu 7 in h C5a (Figure 6(c)), suggesting a structural role of Tyr 28 in maintaining the four-helix bundle structure of m C5a. It is also noteworthy that in m C5a, the helix 1 is much tightly packed to helix 4 than in the h C5a (Figure 2 and Supplementary Figure S10) via hydrophobic interactions (Bursulaya & Brooks, 1999). This further explains the minimal structural change observed in the des-Arg 79 -m C5a, in contrast to the des-Arg 74 -h C5a by the deletion of the C-terminal Arg residue.
Moreover, conformational clustering indicates that over 50 ns, h C5a is largely a three-helix bundle (Figure 4(a)), as the short helix 3 demonstrates conformational flexibility by undergoing a helix-coil transition (Supplementary Figure S11(a)) (Lifson & Roig, 1961), predominantly merging into the L 2 and L 3 regions. In contrast, no such helix-coil transition is observed for the helix 3 of both the Chain A and Chain B of des-Arg 74 -h C5a (Supplementary Figure S11(b) and (c)). Similarly, the helix 3 of m C5a also appears conformationally rigid and does not display any helix-coil transition over 50 ns (Supplementary Figure S11(d)). In agreement, further backbone RMSD analysis indicates that the "L 2 -helix 3-L 3 " region (Val 28 -Gly 44 ) on h C5a is highly conformationally flexible, in comparison with the entire protein backbone (Figure 7(a)). This is unique, as neither the monomers of the des-Arg 74 -h C5a nor the m C5a demonstrate such conformational flexibility at "L 2 -helix 3-L 3 " region ( Figure 7(a)). In addition, the root-mean-square fluctuation (RMSF) analysis per residue evidences that among all the residues, Asn 29 , Arg 37 , and Arg 40 are highly flexible in the "L 2 -helix 3-L 3 " region of h C5a (Figure 7(b)). Surprisingly, the corresponding residues on the m C5a (Asn 34 , Arg 42 , and Arg 45 ), including the residues on both the Chain A and Chain B of des-Arg 74 -h C5a appear rigid (Figure 7(b)). It is also noteworthy that while Arg 37 display the highest fluctuation in h C5a, the Tyr 36 display the highest fluctuation in m C5a. Interestingly, mutation of Tyr 36 in m C5a (Figure 2(c)) to NPA (p-nitrophenyl alanine) has recently been shown to produce an anti-m C5a vaccine (Kessel et al., 2014), effective against the collagen-induced arthritis (Kessel et al., 2014). Similarly, previous mutational studies have also identified Arg 37 as a potential pharmacophore that affect the potency of h C5a toward C5aR (Kola, Baensch, Bautsch, Klos, & Kohl, 1999;Toth et al., 1994). However, unlike the Arg 74 , the role of Arg 37 in conformational heterogeneity of h C5a is not known.
To further probe the structural link between Arg 37 and Arg 74 , the Arg 37 /Ala, Arg 37 /Lys, and Arg 37 /Asp mutants were subjected to MD studies, respectively, over 50 ns each. The results summarized in Figure 8 provide unusual dynamics of information delivery, involving the Arg 74 -Tyr 23 and Arg 37 -Phe 51 "cation-π" pairs as the "allosteric switches" in h C5a. As illustrated in Supplementary Figure S11, the helix 3 is mostly in a coil conformation in h C5a, predominantly helical in des-Arg 74 -h C5a, whereas fluctuates between helix and coil conformations in Arg 37 mutants (Supplementary Figure S12). Interestingly, in respect to the wild-type h C5a, mutation of Arg 37 not only affects the conformation of helix 3, but also affects the distal Arg 74 -Tyr 23 "cation-π" interaction (Marshall et al., 2009) weakly in Arg 37 /Ala, moderately in Arg 37 /Lys and strongly in Arg 37 /Asp mutants (Figure 8 and Supplementary Figure S13). In contrast to the native h C5a, it is further evidenced that proteolytic removal of Arg 74 results in the flipping of the Arg 37 side chain (Figure 9(a)), which forms a strong "cation-π" interaction with the Phe 51 on helix 4, in both Chain A and Chain B of the des-Arg 74 -h C5a dimer (Figure 9(b)). On the other hand, similar structural coordination between the Arg 42 and Arg 79 cannot be established between the structures of the m C5a and des-Arg 79 -m C5a, primarily due to the unresolved CT in native m C5a (Schatz-Jakobsen et al., 2014). However, a consistent "cation-π" interaction between Arg 42 -Phe 56 is evidenced in both m C5a and des-Arg 79 -m C5a (Supplementary Figure S14). Overall, the data suggest that the Arg 37 -Arg 74 pair can modulate the functionality of h C5a through an allosteric mechanism.

Discussion
The phenomenon of allosterism is completely unknown in h C5a and so as its effect on C5aR signaling. Nevertheless, many pharmacophores have been identified on h C5a that potentially modulates the C5aR signaling (Bielecka et al., 2014;Johnson & Chenoweth, 1985;Kola et al., 1999;Mollison et al., 1989;Toth et al., 1994). Surprisingly, the Figure 8. Effect of the Arg 37 mutation on the distally operative Arg 74 -Tyr 23 "cation-π" interaction. The representative structure of the (a) Arg 37 /Ala, (b) Arg 37 /Lys, and (c) Arg 37 /Asp mutants of h C5a are derived from the major microstates evolved for each mutant, over 50 ns of MD. Notes: The helices are, respectively, numbered. The corresponding allosteric effect on the distal Arg 74 -Tyr 23 "cation-π" distance and angle (mean ± SD) monitored over 50 ns of MD is also presented. The dotted line indicates the cutoff distance for the "cation-π" interaction. Figure 9. Proteolysis or mutation of the Arg 74 allosterically triggers the Arg 37 -Phe 51 "cation-π" interaction in des-Arg 74 -h C5a. (a) Comparison of the Arg 37 -Phe 51 "cation-π" distance and angle (mean ± SD) between the native h C5a and the monomers of the des-Arg 74 -h C5a, over 50 ns of MD. (b) Cartoon representation of the Arg 37 -Phe 51 "cation-π" interactions, respectively, in the NMR structure of h C5a and crystal structure of the des-Arg 74 -h C5a.
potential effect of such mutations on structural allosterism in h C5a, affecting the C5aR signaling, is not discussed in the literature so far. It is not even discussed whether h C5a undergoes a conformational change post binding to C5aR. The absence of h C5a-C5aR structural complex makes the problem little more complicated. In fact, the human CXCR4 receptor (with 22 unresolved residues on N-terminus (NT)) is the lone example that has been co-crystalized so far with a viral chemokine (vMIP-II) (Qin et al., 2015). The structural complex with 1:1 stoichiometry strongly supports a two-site binding interaction, demonstrating a single contiguous binding site on CXCR4 that docks the NT of the vMIP-II to the receptor. It is noteworthy that the NT of CXCR4 also interacts with the core of vMIP-II that demonstrates an appreciable conformational change in its loop and NT region (Liwang, Wang, Sun, Peiper, & Liwang, 1999;Qin et al., 2015), post binding to CXCR4. On the other hand, the linear peptide agonist [YSFKPM-PLaR; a = D-Ala; based on the CT peptide of h C5a]bound meta-active model structure of C5aR (Rana & Sahoo, 2015) also indicates a single contiguous binding site on the extra cellular surface of C5aR (lacks 26 residues on NT) similar to CXCR4, further justifying a similar binding interaction between the CT of h C5a and C5aR (Rana & Sahoo, 2015). Thus, it is quite likely that h C5a-C5aR interaction will also demonstrate a similar two-site 1:1 stoichiometry, as observed for CXCR4 receptor (Qin et al., 2015), by involving both the CT and the core of h C5a. In this context, the conformational dynamics in h C5a and its role both in dimerization of des-Arg 74 -h C5a and subsequent C5aR signaling warrants a discussion.
Numerous studies have previously established the C-terminal Arg 74 as a potent pharmacophore on h C5a (Bielecka et al., 2014;Cain et al., 2001;Higginbottom et al., 2005;Schatz-Jakobsen et al., 2014;Toth et al., 1994). It is no denying that absence of Arg 74 attenuates the normal function of h C5a, but whether the associated structural changes (Figure 1) also play a role in signal attenuation is not clear and thus need to be probed. Interestingly, the current MD study on des-Arg 74 -h C5a (Figure 3) does not favorably support the stability of the dimer in solution. In fact, the previous mass spectroscopy data on the plasma-derived des-Arg 74 -h C5a also evidences the presence of a monomer in solution (Reis et al., 2012). Thus, it is apparent that dimerization has a minimal role in the loss of function, displayed by des-Arg 74 -h C5a, which could be most likely due to the conformational allosterism triggered by the absence of Arg 74 . In agreement, a recent study evidences that citrullination of Arg 74 in h C5a ( h C5a-Cit) by PPAD (peptidyl arginine deiminase), an enzyme unique to P. gingivalis bacteria completely abrogates the function of h C5a (Bielecka et al., 2014). This suggests the structural and functional importance of Arg 74 in h C5a for C5aR signaling, though it is not clear from the study (Bielecka et al., 2014) whether h C5a-Cit undergoes dimerization in solution, as observed in the crystals of des-Arg 74 -h C5a (Cook et al., 2010). Interestingly, the Chain A of the des-Arg 74 -h C5a dimer undergoes further folding during MD demonstrating a four-helix bundle conformation ( Figure 5), similar to the native h C5a. However, the Chain B maintains its three-helix bundle structure (Supplementary Figure S6(b)) over the duration of MD. It is worth mentioning that under less optimal folding conditions, it is not unusual for recombinant proteins to demonstrate conformational heterogeneity (Baruah, Bhattacherjee, & Biswas, 2012;Jordan et al., 2009;Yao, Young, Norton, & Murphy, 2011), as observed in Chain A and Chain B of des-Arg 74 -h C5a. Thus, it is quite possible that during refolding, the recombinant h C5a may also fold into more than one closely related conformer, which can be really difficult to purify to homogeneity under certain buffer conditions. Interestingly, in a recent study (Bielecka et al., 2014), such type of conformational heterogeneity is also evidenced in the HPLC traces of commercially available recombinant h C5a. Thus, it is quite likely that under favorable buffer conditions and above the physiological concentration, increased conformational flexibility in the L 1 region (Lys 14 and His 15 ), induced by the disruption of Arg 74 -Tyr 23 "cation-π" interaction (Figure 6(a) and Supplementary Figure S13), may force the dimerization of des-Arg 74 -h C5a monomers in crystals, which may have a minimal biological significance.
Further, the MD studies (Figure 7) also suggest that the "L 2 -helix 3-L 3 " region (Val 28 -Gly 44 ) is highly conformationally flexible compared to the other regions on h C5a and could be functionally essential. This is in agreement with the previous studies, where it has been shown that an antisense peptide (Fujita et al., 2004) targeting the Arg 37 -Glu 53 region abrogates the function of h C5a. Among the pharmacophores identified on the "L 2 -helix 3-L 3 " region of h C5a, Arg 37 on helix 3 deserves most of the attention, as some study suggests the minimal influence of Arg 40 on C5aR binding (Bubeck et al., 1994;Kola et al., 1999). It is suggested that Arg 37 interacts with some unidentified negatively charged residues on C5aR and thus has functional importance (Zuiderweg, Nettesheim, Mollison, & Carter, 1989). In agreement, prior studies also demonstrate that mutations of Arg 37 to Ala/Lys/Asp in h C5a abrogate both the binding and signaling in C5aR (Toth et al., 1994). Interestingly, compared to Arg 37 /Lys, the Arg 37 /Ala mutant of h C5a displays apparently similar loss in C5aR binding, despite the absence of a positively charged side chain, whereas Arg 37 /Asp mutant of h C5a demonstrates a maximum loss in C5aR binding. In agreement with the binding affinities, the mutants also demonstrate a clear differential effect [Arg 37 /Ala < Arg 37 /Lys < Arg 37 /Asp] toward C5aR signaling, which surprisingly correlates perfectly with the strength of the Arg 74 -Tyr 23 "cation-π" interactions, respectively, presented in Figure 8 and Supplementary Figure S13. As evidenced, the Arg 74 -Tyr 23 "cation-π" interaction appears weak in Arg 37 /Ala, moderate in Arg 37 /Lys, and strong in Arg 37 /Asp mutant, compared to wild-type h C5a (Figure 8). This suggests that increase in the helical conformation of helix 3 (Supplementary Figure S12) allosterically strengthens the Arg 74 -Tyr 23 "cation-π" interaction ( Figure 8), which alternatively restricts the availability of the Arg 74 for properly docking the h C5a to C5aR (Rana & Sahoo, 2015), potentially contributing toward the loss of C5aR signaling. On the other hand, the Arg 74 on h C5a appears flexible (Supplementary Figure S3(c)) over the duration of MD and this inherent flexibility could be very important for the functional regulation of h C5a. It is suggested that the h C5a gets rapidly converted to des-Arg 74 -h C5a by the serum carboxypeptidase N (CPN) (Plummer & Hurwitz, 1978), as soon as it is released by the complement system. However, it is not clear whether CPN completely desarginates h C5a before binding to C5aR, which is unlikely given the importance of Arg 74 on h C5a for C5aR binding and signaling. In fact, some studies have also evidenced that internalization of C5aR involves the h C5a with intact antigenicity (Hetland et al., 1997). One possible explanation could be that the docking of the h C5a to C5aR confers conformational rigidity to Arg 74 , which helps the membrane-bound carboxypeptidase M (CPM) (Marquez-Curtis et al., 2008) in converting the h C5a to des-Arg 74 -h C5a, thereby preparing C5aR for signal attenuation and subsequent internalization. Interestingly, CPMs are also known to play a role in modulating the inflammatory response in tissues (Bauvois, 2001;Marquez-Curtis et al., 2008;Skidgel, Stanisavljevic, & Erdos, 2006). Further, as evidenced in Figure 9, the loss of Arg 74 in recombinant h C5a also triggers the Arg 37 -Phe 51 "cation-π" interaction, partly responsible for the conformational rigidity demonstrated by the helix 3 on des-Arg 74 -h C5a (Supplementary Figure S11).
Moreover, the Arg 74 -Tyr 23 and Arg 37 -Phe 51 residue pairs are also conserved across the species (Supplementary Figure S15). It is also evidenced that functionally both Arg 74 and Arg 37 have a very low tolerance to mutation (Toth et al., 1994). It is also apparent from the MD studies that Arg 74 interacts with Tyr 23 (Figure 6), which plays a structural role in h C5a by forming a network of hydrogen bonds with the residues on helix 1. Thus, increased rotameric fluctuation in Tyr 23 (Supplementary Figure S9(a)) can potentially induce conformational heterogeneity in h C5a, by influencing the conformational accessibility of the Lys 14 and His 15 (Supplementary Figures S2 and S4) on L 1 region, alternatively affecting the function of h C5a. In light of this, it is not surprising that mutation of His 15 /Ala abrogates both binding and signaling of h C5a (Toth et al., 1994). Further, mutation of Tyr 23 /Gly or nitration of Tyr 23 by tetra nitromethane, under nondenaturing condition has also been shown to affect the functionality of h C5a (Johnson & Chenoweth, 1985;Mollison et al., 1989). Similarly, it is evidenced that mutation of Phe 51 to any other aromatic residues displays a dramatic alteration in function of h C5a, further suggesting a structural role of Phe 51 in h C5a (Toth et al., 1994). It is noteworthy that Phe 51 , which is structurally constricted between a set of disulfide bridges, can also induce allosterism upon mutation, negatively influencing the function of h C5a. Nevertheless, it supports the absence of Arg 37 -Phe 51 "cation-π" interaction in native h C5a, as evidenced in the NMR structure (Zhang et al., 1997). In contrast, des-Arg 74 -h C5a evidences the presence of Arg 37 -Phe 51 "cation-π" interactions (Cook et al., 2010). In summary, it appears that the Arg 74 -Tyr 23 and Arg 37 -Phe 51 pairs are the nature designed allosteric switches ( Figure 10) that play a significant role in attenuating the h C5a-C5aR binding and signaling.

Conclusion
The functionally essential L 2 -helix 3-L 3 region on h C5a is highly conformationally flexible and dynamic in nature, a feature absent in m C5a. The des-Arg 74 -h C5a dimer in crystals could be a result of conditional allosterism in h C5a. Moreover, the des-Arg 74 -h C5a dimer appears to be dynamic and marginally stable in solution and thus may have a minimal functional significance. The flexibility of Arg 37 on h C5a favors the helix-coil transition in helix 3, which is completely abrogated in des-Arg 74 -h C5a, by the loss of Arg 74 -Tyr 23 "cation-π" interaction, triggering the flipping of Arg 37 to form a strong "cation-π" interaction with the Phe 51 . Further, mutation of Arg 37 also reduces the inherent flexibility of Arg 74 , by strengthening the Arg 74 -Tyr 23 "cation-π" interaction in h C5a. Overall, the Arg 74 -Tyr 23 and Arg 37 -Phe 51 "cation-π" pairs act as "allosteric switches" in modulating the structure and function of h C5a, by controlling the conformational freedom of the Lys 14 and His 15 residues in the L 1 region. The study provides the first glimpse of the allosteric component in h C5a, which will surely help in discovering future potential neutraligands of h C5a.

Supplementary material
The supplementary material for this paper is available online at http://dx.doi.10.1080/07391102.2015.1073634.