International Union of Basic and Clinical Pharmacology. LXXXVII. Complement Peptide C5a, C4a, and C3a Receptors

a. Agonists

Although the pentapeptide Met-Gln-Leu-Gly-Arg that is analogous to the C terminus of C5a is inactive at the C5a₁ receptor (Chenoweth and Hugli, 1980), the discovery that the C-terminal octapeptide of human C5a, His-Lys-Asp-Met-Gln-Leu-Gly-Arg, has weak agonist activity (Kawai et al., 1991, 1992) has provoked intense research into the mechanism of receptor activation and the development of antagonists.

A series of decapeptide analogs first reported by Ember et al. (1992), substituting Phe for His67, had increased potency over the native sequence. These were later developed into a large series of constrained peptides. In some of these, the insertion of Pro into the sequence has resulted in greater agonist activity (Sanderson et al., 1994) but also in a loss of selectivity (Scully et al., 2010). EP-54 (Tyr-Ser-Phe-Lys-Pro-Met-Pro-Leu-dAla-Arg) is an agonist at both the C5a₁ receptor and C3a receptor but has little or no activity at the second C5a receptor C5a₂ receptor (also known as C5R2, C5L2, GPR77) (Scola et al., 2007). In contrast, EP-67 (Tyr-Ser-Phe-Lys-Asp-Met-Pro-(Me)Leu-dAla-Arg) is a weak agonist at both the C5a₁ receptor and the C3a receptor but may have greater activity at the C5a₂ receptor (Kawatsu et al., 1996; Short et al., 1999; Taylor et al., 2001; Vogen et al., 1999a, 1999b).

Peptides such as EP-54 and EP-67 have been described as “response selective,” producing different responses at the C5a₁ receptor depending on the cell type under study (reviewed in Taylor et al., 2001). However, a more likely explanation is the nonselectivity of these peptides, with varying affinities for two or more of the known complement peptide receptors. Peptide ligands have been found to have conserved turn structures (Tyndall et al., 2005), and C5a and its analogs have a β/γ turn motif that has also been observed in bradykinin, enkephalin, and vasopressin. The conservation of the turn structure may be linked to the mechanism of receptor activation because a two-step binding mechanism is a common feature of secretin family GPCR (Tyndall et al., 2005).

b. Antagonists

Further development of C-terminal peptides resulted in a family of hexapeptide analogs of the form N(Me)-Phe-Lys-Pro-dCha-Xxx-dArg, which were potent agonists and antagonists depending on the nature of the fifth residue (Konteatis et al., 1994). Increasing aromaticity at this position was found to enhance antagonist activity, with Trp providing complete antagonism. From these studies, a cyclic peptide antagonist (PMX53; 3D53, (Ac)Phe-[Orn-Pro-dCha-Trp-Arg]) has been developed that shows improved affinity (Wong et al., 1999) (Fig. 5A) and has a β-turn structure similar to the C5a C terminus, as determined by NMR (Zhang et al., 2008).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 5.

C5a1 receptor peptide antagonists. (A) PMX53 (AcPhe-[Orn-Pro-dCha-Trp-Arg]). (B) PMX205 hydrocinnamic acid substituted derivative of PMX53. (C) JPE-1375. (D) L156,602.

PMX53 (pIC₅₀ = 7.05 in human neutrophil membranes) is an insurmountable antagonist for C5a₁ receptor (Strachan et al., 2000, 2001) with no discernible activity at the C5a₂ receptor (Scola et al., 2007). Like many peptides, PMX53 has low oral bioavailability, but its long-lasting effect means that even once-daily oral dosing is sufficient to maintain circulating levels in the rat (Morgan et al., 2008). The development of PMX53 has been discontinued (http://www.evaluatepharma.com/Universal/View.aspx?type=Story&id=178099), although PMX205, a more stable derivative (Delisle Milton et al., 2011) with a hydrocinnamic acid moiety at the N terminus (Fig. 5B), is now being used in animal models of disease, including rat and mouse models of neurologic diseases such as Huntington’s disease, Alzheimer’s disease, and amyotrophic lateral sclerosis (ALS) (Fonseca et al., 2009; Woodruff et al., 2006, 2008). However, PMX205 has been reported as unlikely to be of benefit in ALS patients (ALS-TDI http://www.als.net/ALS-Research/PMX205/ALS-Topics/).

PMX53 also has limited selectivity, in common with the related compound JPE-1375 (Hoo-Phe-Orn-Pro-(d-HLeu)-Phe4F-Phe) (Fig. 5C), and it also binds to other receptors such as NK2 and Mas-related gene 2 receptor (Schnatbaum et al., 2006; Subramanian et al., 2011). Despite this, PMX53 (and, to a lesser extent, JPE-1375) has been of immense value in defining the role of C5a in a wide range of animal models of disease. Described in more detail in reviews (Monk et al., 2007; Klos et al., 2009; Qu et al., 2009; Woodruff et al., 2011), these include inflammatory bowel disease, ischemia-reperfusion injuries, sepsis, and arthritis. Unfortunately, PMX53 and JPE-1375 have limited activity at the rodent C5a₁ receptor in transfected cell lines, particularly when used with recombinant rodent C5a (Waters et al., 2005; P. N. Monk, unpublished observations), making the reported in vivo activities more difficult to analyze.

JPE-1375 was developed as a linear analog of PMX53 by replacing the Arg with Phe and adding hydroorotic acid (Hoo) in place of the N-acetyl group (Proctor et al., 2006; Woodruff et al., 2006). The potency of JPE-1375 is comparable with PMX53, but the former has increased stability in liver microsome preparations. JPE-1375 has been discontinued (Qu et al., 2009) despite showing some promise in AMD (Ricklin and Lambris, 2007), renal allograft survival (Gueler et al., 2008), experimental tubulointerstitial fibrosis (Boor et al., 2007), and atherosclerotic plaque stabilization (Shagdarsuren et al., 2010). The hexadepsipeptide L156,602 (Fig. 5D) is a nonselective antibiotic antagonist for the C5a₁ receptor from Streptomyces with a pIC₅₀ = 5.7 at the C5a₁ receptor, although extensive modification failed to improve selectivity (Hensens et al., 1991; Tsuji et al., 1992a,b, 1995). The C terminus of C5a has also been exploited by the use of phage display to allow selection of libraries of C5a mutants to produce a potent antagonist, A8^Δ71–73 (Heller et al., 1999). This polypeptide binds both the C5a₁ receptor and the C5a₂ receptor (Otto et al., 2004) and has some activity in vivo (e.g., Zhang et al., 2009), although its size and protein derivation make it unsuitable for drug development.

a. Glycosylation

The human C3a receptor is reported to be highly glycosylated (Mizuno et al., 2007) and has two N-linked glycosylation sites (Asn9, Asn194), in the N terminus and second extracellular domain, respectively. In contrast, the mouse C3a receptor has four potential sites in this domain (Tornetta et al., 1997), but it is not clear what functional role this glycosylation plays.

b. Tyrosine Sulfation

Five or six tyrosine residues in the 172-amino-acid second extracellular loop (174, 184, 188, 317/318) of the human C3a receptor are sulfated in vivo (Gao et al., 2003). Tyr174 sulfation is critical for high-affinity binding of C3a but not for receptor activation by C-terminal peptide analogs of C3a, suggesting that Tyr174 directly interacts with the core of C3a.

c. Phosphorylation

The C3a receptor is known to undergo serine/threonine phosphorylation after ligand binding, mediated by G protein-coupled receptor kinases (Langkabel et al., 1999), probably GRK2 and GRK3. The C terminus of the C3a receptor contains 10 serine/threonine residues that may undergo phosphorylation. Mutation of these to alanine indicates that disruption of phosphorylation of Ser465/470 and Thr463/466 inhibits ligand-induced internalization. Ser449 is not involved in internalization but is involved in signal transduction (Settmacher et al., 2003).

d. S-acylation

A potential S-palmitoylation site is present at the intracellular C terminus (Cys468) of the human C3a receptor and also in the rat and mouse but not guinea pig C3a receptor. It is not known if this Cys is actually modified in vivo. A potential eighth helix at the C terminus (Asn431–Gln451) is suggested by sequence similarities to bradykinin R2 and other receptors (Feierler et al., 2011). Interestingly, the C5a₁ and C5a₂ receptors do not have this basic sequence and also contain proline, a potential helix blocker.

a. G protein mediated

By comparison with C5a, C3a is generally a much weaker chemotactic stimulus in leukocytes (Fernandez et al., 1978), despite sharing many of the same signaling mechanisms. Following the binding of C3a to C3a receptor, the primary signaling mechanism activated is through the pertussis toxin (PT)-sensitive G protein Gα_i in human and mouse immune cells such as neutrophils (Norgauer et al., 1993), eosinophils (Elsner et al., 1994), and microglia (Moller et al., 1997). C3a has also been reported to decrease intracellular cAMP levels in murine dendritic cells (Li et al., 2008), which might be dependent on Gα_i. C5a has been shown to have this effect as well in cell lines (Vanek et al., 1994) and in DC (Peng et al., 2009b). However, in endothelial cells, C3a signaling caused a PT-insensitive cytoskeletal response and ERK1/2 activation, attributed to activation of Gα12 and/or Gα13 despite the availability of Gα_i for coupling to C5a₁ receptor in the same cells (Schraufstatter et al., 2002). The promiscuous PT-insensitive G protein Gα16 can also couple to C3a receptor in cotransfected cell lines (Crass et al., 1996). Differences have also been observed in the subsequent changes in intracellular free Ca²⁺ (Ca²⁺_i), with C3a typically stimulating a smaller and transient increase in intracellular Ca²⁺_i in neutrophils solely due to the influx of extracellular Ca²⁺ whereas C5a also causes the release of Ca²⁺ from intracellular stores and a more prolonged elevation due to the activation of phospholipase C (PLC) (Norgauer et al., 1993). However, in microglia, the Ca²⁺_i responses are equal (Moller et al., 1997), and in human MSC both C3a and C5a cause a prolonged activation of protein phosphorylation (Schraufstatter et al., 2009). Similarly, in a human astrocytic cell line, both C3a and C5a stimulated the release of Ca²⁺ from intracellular stores, although the response to C3a was smaller (Sayah et al., 2003).

b. Arrestin mediated

In mast cells, although C3a provokes only a transient increase in Ca²⁺_i, the activation of ERK1/2 and Akt phosphorylation is sustained (Venkatesha et al., 2005). This disparity in the temporal aspects of C3a signaling in mast cells has been explored in great detail, and it is known that degranulation is uncoupled from other responses, such as de novo production of the chemokine CCL2 (Ahamed et al., 2001). Here, the use of a C3a receptor mutant, truncated at the C terminus to remove potential serine/threonine phosphorylation sites, has shown that production is absolutely dependent on receptor phosphorylation whereas degranulation was not affected. The mutant receptor failed to bind β-arrestin 2, suggesting that arrestins can act as mediators of additional signaling mechanisms for the C3a receptor. However, both degranulation and the production of CCL2 were equally sensitive to PT, suggesting that Gα_i activation is still critical, most likely through ERK1/2 activation. Arrestins were originally described as adaptors that link ligand-activated GPCR to the cellular internalization machinery (Wilden et al., 1986; Lohse et al., 1990), an important part of the complex desensitization process (Jalink and Moolenaar, 2010). The G_q-coupled protease-activated receptor PAR2 couples to ERK1/2 through two pathways, one G protein dependent and one dependent on β-arrestin/Src kinase that results in ERK1/2 activation by endocytosed receptors in a particular subcellular location (DeFea et al., 2000). However, the early phase of ERK1/2 activation by C3a receptor is not dependent on Src (Ahamed et al., 2001), so the mechanism by which β-arrestin mediates C3a receptor signaling is still unclear. For the C3a receptor, at least, different arrestins appear to play different roles. The silencing of β-arrestin 2 in human mast cells prevents the desensitization and internalization of C3a receptor, leading to a prolonged increase in Ca²⁺_i (Vibhuti et al., 2011) but has no effect on degranulation whereas silencing of β-arrestin 1 could actually inhibit degranulation. Interestingly, both arrestins were observed to suppress ERK1/2 activation by the C3a receptor, in contrast to results observed with other GPCR (Defea, 2008). Phosphorylation of the C3a receptor by GRK2, 3, 5, and 6 has been implicated in the control of the association with arrestins (Guo et al., 2011), with different GRK involved in different aspects of the response to C3a. Thus, signaling by C3a receptor, at least in mast cells, is complex and is critically dependent on arrestins as well as G proteins. This sensitivity to arrestins may explain why C3a receptor signaling is often more transient and weaker in nature relative to that of C5a₁ receptor.

a. Glycosylation

There is one potential site for N-linked glycosylation on human C5a₁ receptor, at Asn2. Removal of N-linked glycosylation by endoglycosidase F or by mutation of Asn2 had little effect on either affinity for the ligand or expression of the receptor at the cell surface (Pease and Barker, 1993).

b. Tyrosine sulfation

There are three tyrosine residues at the N terminus of C5a₁ receptor, two of which (Tyr11, Tyr14) are proximal to Asp residues, making them potential sulfation sites (Rosenquist and Nicholas, 1993). Both Tyr11 and Tyr14 are sulfated, and this modification is essential for the binding of C5a (Farzan et al., 2001), C5a des-Arg (Scola et al., 2007), and CHIPS (Ippel et al., 2009; Liu et al., 2011c).

c. Phosphorylation

Phosphorylation of the C5a₁ receptor after ligand binding or phorbol ester treatment has been linked to desensitization (Ali et al., 1993). The major phosphorylation sites are at the C terminus (Ser314, Ser317, Ser327, Ser332, Ser334, and Ser338) (Giannini et al., 1995) although the third intracellular loop also contains a potential PK-C site (Lys239-Thr-Leu-Lys) (Bock et al., 1997). Mutation of Ser332, Ser334, and Ser338 to Ala reduces phosphorylation by 80% (Giannini et al., 1995) and also inhibits receptor internalization (Naik et al., 1997). This inhibition was thought to be due to a loss of association with molecules important for internalization such as β-arrestin, dynamin, and clathrin. Interestingly, β-arrestin 1 and 2 were still able to bind to phosphorylation-deficient C5a₁ receptor but much more weakly than to a wild-type receptor (Braun et al., 2003). However, internalization of the receptor, one mechanism for desensitization, was found to be dependent only on amino acids 335–350 in a series of truncated C5a₁ receptor mutants, beyond the majority of the phosphorylation sites (Bock et al., 1997). Similarly, another study found that although Ser334, Ser327, Ser332, and Ser338 could be phosphorylated by PK-Cβ, these sites were functionally redundant for internalization and desensitization and even that, at high concentrations of C5a, β-arrestin binding to the C terminus could actually inhibit phosphorylation (Pollok-Kopp et al., 2007). Thus, the role of phosphorylation in the control of receptor function remains unclear.

a. C5a₁ receptor binds both C5a and C5a des-Arg

Determining the locations and mechanisms of interaction between C5a₁ receptor and its ligands is vital to the production of effective therapeutics. As for most members of the GPCR family, there is no available crystal structure for the C5a₁ receptor, so structural knowledge of the receptor has been based on extensive ligand binding, receptor chimera, and mutagenesis studies. Molecular modeling of the receptor now plays a major role in our understanding, with a recent model of a ligand-bound C5a₁ receptor providing novel insights into the mechanisms of ligand and receptor interaction (Nikiforovich et al., 2008). The most extensively studied interaction to date is between the receptor and the full-length natural ligand C5a. In vivo, C5a is rapidly converted into C5a des-Arg by carboxypeptidase enzymes, which remove the terminal arginine residue (Bokisch and Muller-Eberhard, 1970). Experimentally, C5a binds the C5a₁ receptor with a K_d of 1 nM, whereas the truncated form has a binding affinity which is 10- to 100-fold lower than that of C5a and is a partial agonist at the C5a₁ receptor (Higginbottom et al., 2005). Interactions with this truncated form will also be considered here, where data are available.

a. Binding site one: C5a₁ receptor N terminus

Yeast random saturation mutagenesis (RSM) and NMR studies on the N terminus of the C5a₁ receptor have indicated the importance of the many acidic residues in C5a binding (Chen et al., 1998; Hagemann et al., 2006). In general, single/double mutations of these have had little effect on binding whereas various multiple mutations within the N terminus of the C5a₁ receptor have significantly reduced the affinity of the C5a₁ receptor for C5a. For example, the C5a₁ receptor(D15,16,18,21N) reduced C5a affinity 40-fold, and the C5a₁ receptor(D10,15,16,18,21N) reduced affinity by 133-fold (DeMartino et al., 1994), whereas the C5a₁ receptor(D10N), C5a₁ receptor(D27N), and C5a₁ receptor(D21,27N) had no effect on C5a binding (Mery and Boulay, 1994). The lack of importance of individual residues in forming the binding site was supported by the yeast RSM study, which concluded that no single residues were essential for ligand activation of the C5a₁ receptor (Hagemann et al., 2006). More recently, site-specific disulfide-trapping experiments performed in yeast have identified several potential specific points of contact between C5a and the N terminus of the C5a₁ receptor (Hagemann et al., 2008).

The N terminus of the C5a₁ receptor is highly flexible, resulting in many possible low-energy conformations for the interaction between the C5a and C5a₁ receptors (Nikiforovich et al., 2008). Some interactions were consistent across many of the possible conformations and can be rationalized using available mutagenesis data, especially as these data were not used in the building of the model. The aspartate residues Asp15, Asp16, and Asp21 were predicted in most conformations to form a salt bridge with Lys17 within the N terminus of C5a₁ receptor, indicating that the reductions in C5a binding affinity induced by multiple aspartate replacements likely result from changes to the overall conformation of the receptor N terminus. Asp27 was predicted to form a less important salt bridge than neighboring Asp28, in accordance with the lack of disruption in binding upon single mutation of this residue to asparagine (Mery and Boulay, 1994). Asp18 was predicted not to be involved in salt bridge formation within the C5a₁ receptor, but in a small number of conformations was predicted to interact with Lys19 or Lys20 of C5a, supported by mutagenesis data showing a reduction in binding when these lysine residues within C5a are mutated (Bubeck et al., 1994; Toth et al., 1994). Other important potential contact points include Arg46 (the side chain of which in certain conformations could interact electrostatically with Asp10 or Asp16) and Cys27, which has also been predicted to contact fragment 24–30 of the C5a₁ receptor by RSM (Hagemann et al., 2006) and the modeling study (Nikiforovich et al., 2008).

The importance of tyrosine sulfation in forming the ligand-binding site in other GPCRs (Hsu et al., 2005) led to the investigation of Tyr11 and Tyr14 in C5a₁ receptor, which have been shown to be sulfated. The C5a₁ receptor mutations Y14F and Y11F induced a 50% reduction or complete loss of binding affinity for C5a, respectively, indicating the importance of these sulfations in the formation of the N-terminal binding site (Farzan et al., 2001). The binding of CHIPS, which acts as an antagonist for the C5a₁ receptor, has also been shown to be dependent on the sulfation of tyrosine residues 11 and 14 (Ippel et al., 2009; Liu et al., 2011c) (Section IV.C.2).

b. Binding site two: transmembrane/extracellular loop regions of C5a₁ receptor

One approach used to ascertain the general regions of C5a₁ receptor involved in the second binding site was the construction of chimeric C5a₁ receptor/FPR1. Replacing the first extracellular loop of the C5a₁ receptor with the corresponding region of the FPR1 had no effect on binding affinity for C5a whereas replacement of EC2 or EC3 abolished binding (Pease et al., 1994). A lack of direct interactions between EC1 and C5a was also supported in the 2008 model (Nikiforovich et al., 2008), but the effects of several point mutations within EC1 that did alter ligand interactions indicate that this region is important for overall structural conformation (Cain et al., 2001b); the presence of the Trp-Phe-Xxx-Gly motif in EC1, which is highly conserved among GPCR, supports this (Klco et al., 2006).

Mutagenesis studies identified Ile116, Arg175, Arg206, Glu199, Asp282, and Val286 as potential interaction sites in ligand binding (Fig. 8). A recent model of C5a bound to the C5a₁ receptor has suggested a more extensive list of potential interaction sites with residues 59–74 of C5a being predicted to interact with the side chains of Leu117, Met120, Tyr121 (TM3); Leu167, Phe172 (TM4); Leu187, Cys188, Asp191, His194 (EC2); Glu199, Arg200, Ala203, Arg206, Leu207, Leu209, Pro214 (TM5); Met265 (TM6); and Leu277, Asn279 (EC3) (Nikiforovich et al., 2008). Some of these residues had already been identified as potential ligand-binding sites in mutagenesis studies.

Analysis of mutations of residue Glu199 at the top of TM5 has produced some conflicting biologic results regarding receptor activation. An interaction between Glu199 of the C5a₁ receptor and Lys68 of C5a has been suggested in several studies and has been shown to be important for activation of the C5a₁ receptor by C5a des-Arg, but not C5a (Crass et al., 1999c; Monk et al., 1995). Specifically, an E199K mutation of the C5a₁ receptor produced a lower binding affinity for wild-type C5a and a higher binding affinity for the C5a mutant K68E, indicating the presence of a salt bridge between these two residues of ligand and receptor (Crass et al., 1999c). The recent model of C5a bound to the C5a₁ receptor does indeed predict salt bridge formation between Lys68 of C5a and the side chain of residue Glu199 of the C5a₁ receptor, with the E199K mutation also disrupting hydrogen bonding between residues Glu199 and His194/Gln71 of the C5a₁ receptor (Nikiforovich et al., 2008). In addition to the interaction with Lys68, Glu199 has also been predicted to interact with Arg74 after a lack of response of an E199K mutant to agonists lacking a C-terminal arginine (Crass et al., 1999b; Higginbottom et al., 2005).

Another residue at the extracellular face of TM5 that has been proposed to interact with the C-terminal arginine of C5a (Arg74) is Arg206 (DeMartino et al., 1995; Gerber et al., 2001); however, subsequent studies found only a small effect on receptor activation of the R206A mutation (Cain et al., 2001a). This, along with the fact that the truncated ligand C5a des-Arg binds to the R206A mutant but does not activate it, led to the hypothesis that this mutation simply alters the global structure of the receptor (Monk et al., 2007). R206A mutants have been reported to have varying effects on C5a binding affinities from no change (DeMartino et al., 1995) to significant reductions in binding (Raffetseder et al., 1996). As stated previously, Arg206 has been predicted from modeling to interact with C5a. Within this model, the R206A mutation disrupts the predicted interaction between Arg206 and Arg74 of C5a but induces no major changes in residue-residue interactions and does not support experimental data indicating a complete absence of C5a binding (Nikiforovich et al., 2008).

Buried deep in TM5, Trp214 in the human C5a₁ receptor is conserved in the gerbil C5a₁ receptor but not in rodent receptors or in human C5a₂ and C3a receptors. If substituted by its murine equivalent, Leu, the resulting C5a₁ receptor mutant (W213L) is no longer antagonized by W54011 or NDT9520492. Conversely, the murine C5a₁ receptor could be sensitized to these antagonists by the reciprocal substitution L214W (Waters et al., 2005). Interestingly, PMX53 showed no activity at either the wild-type murine receptor or the mutant L214W, suggesting that this antagonist binds at a different site.

Other residues that have been predicted to interact with the C terminus of C5a include Arg175 and Asp282. Arg175 is located in EC2, and it too has been predicted to interact with Arg74 of C5a with R175A and R175D mutations, both causing a reduction in C5a binding affinities and activation of C5a₁ receptor (Cain et al., 2003; Higginbottom et al., 2005). In the C5a bound model of C5a₁ receptor, no major disruptions in binding were predicted with an R175D mutant, indicating no specific interaction between Arg175 and C5a (Nikiforovich et al., 2008). However, the mutagenesis data can be explained by the loss of a salt bridge between Arg175 and Glu179 in the mutant receptors, an interaction that may be crucial in stabilizing the “open” conformation of EC2. Asp282 is at the extracellular face of TM7 and is predicted to form an important interaction with Arg74 of C5a. In support of this, the D282R mutant is relatively unresponsive to C5a but sensitive to C5a des-Arg and analogs (Cain et al., 2001a, 2003). However, again minimal changes were detected in the C5a bound model of the C5a₁ receptor for the D282R mutant. Although both Arg175 and Asp282 were not predicted to interact directly with C5a in this model, they have previously been predicted to interact with the cyclic hexapeptide antagonist PMX53, along with Glu199, Arg206, and Ile116 among others (Higginbottom et al., 2005).

Ile116 in TM3 of the C5a₁ receptor was predicted to be part of an activation switch after an I116A mutation altered the activity of the C5a peptide C-terminal analog (Phe-Lys-Pro-dCha-Trp-dArg) from antagonist to agonist (Gerber et al., 2001). Further studies mutating this residue found an increased affinity for agonists but a decreased activation of the C5a₁ receptor, indicating that the mutation reduced the activation efficiency of the C5a₁ receptor (Higginbottom et al., 2005). In more recent modeling studies, the side chains of C5a fragment 59–74 have been predicted to contact the nearby Leu117 residue (Nikiforovich et al., 2008) but not Ile116 itself, whereas modeling based on the binding of hexapeptide antagonist has predicted interactions with Ile116 (Gerber et al., 2001; Higginbottom et al., 2005). Val286, in TM7, has also been proposed to be involved in the activation switch mechanism, with its side chain being proposed to point toward that of Ile116, allowing the two residues to form a binding cleft for ligands. In recent models, Val286 too has been predicted to interact with peptide ligands but not intact C5a (Higginbottom et al., 2005; Nikiforovich et al., 2008).

Much of the information now available indicates that different ligands bind the C5a₁ receptor in different ways; even just the cleaving of the terminal arginine residue in C5a des-Arg changes the predicted interactions between receptor and ligand. Through modeling of the ligand-receptor interaction, two important differences in the way in which C5a des-Arg binds compared with C5a were the lack of an interaction between Arg206 and the missing Arg74 of the ligand, and the lack of an interaction between Arg200 and Asp69. In addition to these specific missing interactions, the C terminus of C5a des-Arg was predicted to bind in an entirely different orientation to C5a (Nikiforovich et al., 2008). Available antagonists and analogs of the native ligand are again predicted to interact with different residues of C5a₁ receptor; identifying the mechanisms important in eliciting desired effects, such as antagonism, will be crucial in directing the future development of therapeutic agents.

In addition to identifying ligand-binding positions, many mutations in the second extracellular loop of the C5a₁ receptor have also been found to result in constitutive receptor activity, identifying a negative regulatory role of EC2 (Klco et al., 2005). As constitutively active GPCR are thought to most likely represent the activated GPCR state, the study of such mutants could provide vital insight into the activation switch of GPCR in general. In modeling studies, the EC2 loops of constitutively active C5a₁ receptor mutants were found to contact the EC3 loops, leading to interaction of EC2 domains with TM3 (Nikiforovich and Baranski, 2012). This in turn triggers the movement of TM7 toward TM2 and TM3 with a resultant change in hydrogen bonding between these regions known to be crucial for C5a₁ receptor activation (Nikiforovich et al., 2011).

c. Comparison with C5a₂ receptor

The second identified receptor for C5a, the C5a₂ receptor, shares ∼35% identity with the C5a₁ receptor; in contrast with the C5a₁ receptor, the C5a₂ receptor binds C5a des-Arg and C5a with similar affinities (Scola et al., 2007), implying different binding mechanisms. Interfering with the N terminus of the C5a₂ receptor using an antibody did not alter the binding of C5a to the C5a₂ receptor, but it did significantly inhibit the binding of C5a des-Arg to the C5a₂ receptor (Scola et al., 2007). Replacing the N terminus of the C5a₂ receptor with that of the C5a₁ receptor did not affect the affinity for C5a des-Arg. Many of the residues at the extracellular face of the transmembrane domain of C5a₁ that have been shown to influence ligand binding are also present in the C5a₂ receptor (Fig. 8).

a. G protein mediated

Often the most potent of the complement peptides, C5a has also been the most intensively studied with regard to signal transduction (reviewed in Monk et al., 2007; Klos et al., 2009). The C5a₁ receptor couples primarily to the PT-sensitive G protein Gα_i2 (Sheth et al., 1991; Skokowa et al., 2005) in cells such as neutrophils and mast cells or, less frequently, to PT-insensitive G proteins such as Gα16/Gα15 (Amatruda et al., 1993; Monk and Partridge, 1993; Davignon et al., 2000) in cells of the monocytic lineage. The C5a₁ receptor is unusual among GPCR in being precoupled to G proteins (Siciliano et al., 1990); mutants that are unable to precouple in this way have reduced affinity for C5a (Raffetseder et al., 1996). Precoupling is also seen in a mutant of the α_2B adrenoreceptor, where it increases agonist potency (Ge et al., 2003); it is tempting to speculate that this may also be the case for the C5a₁ receptor, allowing the response to C5a to predominate over responses to other chemoattractants such as C3a.

As with the C3a receptor, the downstream response to receptor ligation is cell type dependent. For example, the Ca²⁺ response in C5a-activated neutrophils is predominantly due to release from intracellular compartments, whereas in monocyte-like cells a much greater contribution from extracellular influx is observed (Monk and Partridge, 1993). Similarly, mast cells respond to C5a with a rapid and transient increase in Ca²⁺_i (Hartmann et al., 1997) whereas microglial cells respond with a more prolonged response (Moller et al., 1997). In neutrophils and macrophages, sphingosine-1-phosphate (S-1-P) production by sphingosine kinase 1 is required for the C5a-stimulated release of Ca²⁺ from intracellular stores (Ibrahim et al., 2004; Melendez and Ibrahim, 2004). S-1-P also up-regulates the expression of the C5a₂ receptor (the second C5a receptor) in mouse neutrophils, which is thought to be a protective response against endotoxemia (Bachmaier et al., 2012).

b. Arrestin signaling

Phosphorylation of the C5a₁ receptor leads to association with arrestins and subsequent targeting to clathrin-coated pits for internalization (Braun et al., 2003). PK-CβII and GRK2 interact with intracellular loop 3 and the C-terminal domain of the C5a₁ receptor and may be responsible for the serine/threonine phosphorylation of the receptor (Suvorova et al., 2008). A dileucine motif in the C terminus of the C5a₁ receptor is thought to stabilize the “eighth helix” (Gln305–Arg320) proximal to the membrane (Suvorova et al., 2008). This structure, sometimes stabilized by a palmitoylated cysteine residue, often acts as a protein interaction site in GPCR (Huynh et al., 2009) and may be involved in receptor internalization (Thomas et al., 1995). Interestingly, the other receptors—the C3a receptor and C5a₂ receptor—both have cysteine residues in the C-terminal domains whereas the C5a₁ receptor does not (Fig. 8). In some GPCR with cysteine residues in the C-terminal domain, such as bradykinin B2, helix 8 is critical for the association with β-arrestins, so the lack of a cysteine residue in this position in the C5a₁ receptor may be responsible for some of the differences in signaling observed for the complement peptide receptors. Unlike the C3a receptor, there is no evidence that the C5a₁ receptor signals through arrestins (Gripentrog and Miettinen, 2008), although an association occurs (Braun et al., 2003; Kalant et al., 2005; van Lith et al., 2009). Kinase activation by the C5a₁ receptor is almost entirely Gα_i dependent (Buhl et al., 1994; Gripentrog and Miettinen, 2008), with little evidence to date that arrestins are directly involved. It is also interesting that the C terminus of the C5a₁ receptor can be substantially deleted (from residue 311) without a negative effect on signaling (Matsumoto et al., 2007a,b; Monk et al., 1994), although it is required for trafficking to the cell surface and for internalization (Bock et al., 1997).

The ability of the C5a₁ receptor to form homodimers and heterodimers has been demonstrated (Geva et al., 2000; Huttenrauch et al., 2005; Klco et al., 2003). In both cases, ligation of a dimerized C5a₁ receptor can cause the phosphorylation and/or internalization of the partner receptor. The functional consequences of this ability to form oligomers are unknown but may be linked to the cross-modulation of the response to chemoattractants or the rapid down-regulation of the C5a₁ receptor in severe conditions such as sepsis.

a. Glycosylation

A potential N-linked glycosylation site exists at Asp3 of the human C5a₂ receptor, and, although the glycosylation status of this residue has not been formally investigated, the apparent molecular weight observed by Western blot analysis suggests that the receptor is glycosylated in vivo (Okinaga et al., 2003).

b. Tyrosine sulfation

Like the C5a₁ receptor, there are three tyrosine residues at the N terminus of the C5a₂ receptor, all of which (Tyr8, Tyr10, and Tyr13) are proximal to acidic residues, making them potential sulfation sites (Rosenquist and Nicholas, 1993). Mutation of Tyr10 or Tyr13 but not Tyr8 can affect ligand binding to the C5a₂ receptor, but only the binding of C5a des-Arg is significantly inhibited (Scola et al., 2007).

c. Phosphorylation

The C5a₂ receptor has only seven serine/threonine residues at the C terminus, in comparison with the C5a₁ receptor which has 11. A low level of phosphorylation of the C5a₂ receptor has been observed after C5a treatment of transfected cells (Okinaga et al., 2003), suggesting that few of these residues are phosphorylated, perhaps because of the lack of the PK-CβII binding site found on the third intracellular loop of the C5a₁ receptor (Section IV.C.2).

d. S-acylation

Similar to the C3a receptor, the C5a₂ receptor has a cysteine in the C-terminal domain that is a potential S-acylation site. It is not known whether any modification of this residue occurs in vivo.

a. Intracellular or extracellular localization for C5a₂ receptor

Although it is generally accepted that mouse, rat, and human PMN express the C5a₂ receptor, controversy exists about C5a₂ receptor localization in this cell type. Some investigators did not detect significant amounts of the receptor on the cell surface (Otto et al., 2004; Bamberg et al., 2010), but others could verify surface expression in binding assays or by flow cytometry (Okinaga et al., 2003; Gao et al., 2005; Rittirsch et al., 2008). Finally, Scola et al. (2009) suggested a wide natural variation in C5a₂ receptor surface expression between human individuals. In some cases, opposing findings were subsequently published by the same group (Okinaga et al., 2003; Bamberg et al., 2010). These discrepancies may in part be due to differential regulation of C5a₂ receptor expression depending on the activation status of the cells. Also, not all of the polyclonal antisera or monoclonal antibodies used may have been suitable to detect the C5a₂ receptor on primary cells. Some antibodies raised against antigenic peptides or against transfected cells fail to recognize their antigen on primary cells, despite being able to detect it on transfected cells. This should be taken into account whenever newly developed antibodies are used for detection of this receptor on primary cells (Jorg Köhl, personal communication).

b. Regulation of expression

The regulation of C5a₂ receptor expression is influenced by proinflammatory and anti-inflammatory signals. Several cell lines have been shown to express the C5a₂ receptor either constitutively or after induction with certain substances. On HeLa cells, the C5a₂ receptor but not the C5a₁ receptor is detected at low levels, and it decreases upon treatment with IFN-γ or tumor-necrosis factor α (TNF-α). Both the C5a₁ and C5a₂ receptors are absent from the surface of native HL-60 or U937 cells, but their expression can be induced with Bt₂-cAMP or IFN-γ, whereas TNF-α has no effect (Johswich et al., 2006). In rat astrocytes, C5a₂ receptor mRNA and protein were found to be up-regulated by noradrenaline, associated with suppression of nitric oxide synthase 2 and NF-κB genes (Gavrilyuk et al., 2005). Huber-Lang et al. (2005) could show that prolonged exposure of rat and human PMN to C5a in vitro decreased C5a₂ receptor expression. Likewise, C5a₂ receptor is reduced on PMN isolated from septic patients or from rats subjected to sepsis induced by cecal ligation and puncture (CLP). TLR activation in human PBMCs was found to inhibit C5a₂ receptor expression, thereby increasing cellular responsiveness to C5a₁ receptor signaling. Thus, the C5a₂ receptor appears to be differentially regulated in different cell types, and its expression level may influence cellular functions, especially in the context of inflammation.

a. C5a and C5a des-Arg

The C5a₂ receptor was initially speculated to be the long sought-after C4a receptor (Ohno et al., 2000; Lee et al., 2001), but in fact turned out to be a second human receptor for C5a, binding both C5a and C5a des-Arg with high affinity (Cain and Monk, 2002; Okinaga et al., 2003; Johswich et al., 2006). The K_d values of C5a₂ receptor binding were in the range of ∼3–10 nM for C5a, and ∼12–36 nM for C5a des-Arg, respectively, with C5a₂ receptor having 10- to 100-fold higher affinity for C5a des-Arg in comparison with the C5a₁ receptor. Although the affinity of the human C5a₂ receptor is about equal for C5a and C5a des-Arg, both mouse and rat C5a₂ receptor homologs clearly prefer species-matched C5a des-Arg over C5a, with mouse C5a₂ receptor displaying an over 4000-fold higher affinity for C5a des-Arg than for C5a. As in humans, the rodent C5a₂ receptor outmatches the rodent C5a₁ receptor in its affinity for C5a des-Arg (Scola et al., 2007). Moreover, in one study, the human C5a₂ receptor exhibited distinctly (∼100-fold) slower on-rates and slightly (∼3-fold) slower off-rates for C5a at 0°C and in the presence of azide (Okinaga et al., 2003).

b. Other complement peptides

Whether the C5a₂ receptor can bind C3a/C3a des-Arg or C4a/C4a des-Arg is a matter of controversy. Moderate to low binding of C3a and C4a to a site distinct from the C5a binding site was first detected in 2002 (Cain and Monk, 2002). Shortly thereafter, Kalant et al. (2003) reported binding of the des-arginated ligands C3a des-Arg and C4a des-Arg to the C5a₂ receptor, and observed increased triglyceride synthesis in C5a₂ receptor-transfected cells after stimulation with C3a des-Arg, also referred to as acylation-stimulating protein (ASP), or with C3a. In contrast, others were unable to detect binding of any ligands apart from C5a and C5a des-Arg to C5a₂ receptor (Okinaga et al., 2003; Johswich et al., 2006), and provided evidence that C3a and C3a des-Arg interact with plastic surfaces in a pattern highly suggestive of specific receptor-ligand binding (Johswich et al., 2006). The complex issue of C5a₂ receptor as a functional receptor for C3a des-Arg is reviewed in more detail in Section IV.E.

c. Comparison with C5a₁ Receptor ligand binding

C5a and C5a des-Arg binding by the C5a₂ receptor differs from that of the C5a₁ receptor, despite the two receptors sharing patterns of tyrosine and acidic N-terminal residues as well as charged and hydrophobic residues in the extracellular loops and transmembrane regions (Scola et al., 2007). In the C5a₁ receptor, these residues are known to interact with the core and C terminus of C5a, respectively (Oppermann et al., 1993; Pease et al., 1994). However, the well-characterized antagonist PMX53 (Wong et al., 1999; Taylor et al., 2001) and other effective peptidic and nonpeptidic ligands for the transmembrane hydrophobic pocket of the human C5a₁ receptor are moderate-to-poor inhibitors of ligand binding to the C5a₂ receptor, suggesting that the C5a core segment is the most relevant for C5a binding to the C5a₂ receptor (Okinaga et al., 2003; Otto et al., 2004; Scola et al., 2007). An antibody raised against the N terminus of human C5a₂ receptor inhibited the binding of C5a des-Arg to the C5a₂ receptor but left C5a binding unaffected, while substituting the N terminus of the C5a₂ receptor with that of the C5a₁ receptor had little effect on the high affinity of the C5a₂ receptor for C5a des-Arg (Scola et al., 2007). Finally, mutational analyses revealed three N-terminal residues of the C5a₂ receptor that were essential for C5a des-Arg binding but had little involvement in C5a binding (Scola et al., 2007). Taken together, these findings indicate that the N terminus is an important but not the only determinant of the high affinity of the human C5a₂ receptor for C5a des-Arg, and that, unlike the C5a₁ receptor, critical N-terminal residues of the C5a₂ receptor interact with C5a des-Arg only, not with C5a.

d. Modulation of ligand binding

While a number of blocking antibodies specific for the mouse, rat, or human C5a₂ receptor have been developed and employed by different investigators in vitro and in vivo (Gao et al., 2005; Cui et al., 2007; Lee et al., 2008; Bamberg et al., 2010), to our knowledge the only antagonists for the human C5a₂ receptor are the C5a mutant jun/fos-A8, and its derivative jun/fos-A8^Δ71–73, which inhibit ligand binding to both the C5a₁ receptor and C5a₂ receptor (Heller et al., 1999; Otto et al., 2004).

a. G protein mediated

Despite sharing the 7TM structure of typical GPCR, the C5a₂ receptor is widely accepted to be uncoupled from G proteins (Cain and Monk, 2002; Kalant et al., 2003; Okinaga et al., 2003; Johswich et al., 2006; Scola et al., 2009). This is believed to be due to changes in conserved sequences shared by most functional chemoattractant or chemokine receptors, including the C5a₁ receptor. Primarily, both mouse and human C5a₂ receptors lack the highly conserved “DRX” motif usually found at the end of TM3 (Oliveira et al., 1994; Ballesteros et al., 1998), actually Asp-Arg-Phe in the human C5a₁ receptor. In the C5a₂ receptor, it reads Asp-Leu-Cys, thus lacking the arginine residue that is most important for the interaction with G proteins (Savarese and Fraser, 1992; Scheer et al., 1996; Cain and Monk, 2002; Okinaga et al., 2003). Mutating the respective arginine in FPR, CCR5, histamine H₂ receptor, and other receptors greatly decreases or altogether abolishes their ability to couple to G proteins (Prossnitz et al., 1999; Alewijnse et al., 2000; Lagane et al., 2005; Rovati et al., 2007). Conversely, C5a binding to a C5a₂ receptor in which Asp-Leu-Cys is mutated to Asp-Arg-Cys induces a low level of intracellular Ca²⁺ mobilization in HEK293 cells cotransfected with Gα16, indicative of a partial gain-of-function in coupling to this rather promiscuous G protein subtype (Okinaga et al., 2003). However, RBL-2H3 cells transfected with the C5a₂ receptor mutated to resemble the C5a₁ receptor at this motif did not respond to C5a or C5a des-Arg in an intracellular free Ca²⁺ assay (Scola et al., 2009). Given that rat basophilic leukemia (RBL) cells do not endogenously express Gα16, variations in the amount and selectivity of the two G proteins may explain these differing results.

Further residues in the mouse and human C5a₂ receptors that differ from conserved sequences in the C5a₁ receptor and other GPCR are in the “NPXXY” motif (Asn-Pro-Leu-Met-Phe in C5a₂ receptor) in TM7, and deletions of serine/threonine residues and a basic region (²³⁹Lys-Thr-Leu-Lys in C5a₁ receptor) that truncate the third intracellular domain (Cain and Monk, 2002; Scola et al., 2009). Substitution of the analogous serine/threonine residues with alanine in the C5a₁ receptor does not interfere with ligand binding but abolishes downstream signaling, suggesting that this region is also involved in G protein coupling (Bock et al., 1997). Different mutations of the NPXXY motif have been shown to inhibit the activation of G_q by the cholecystokinin B receptor and to disturb internalization, ligand binding, or coupling to selected downstream pathways in the case of FPR (Gales et al., 2000).

In an effort to further elucidate the regions responsible for uncoupling the C5a₂ receptor from G proteins, a recent study compared the C5a₁ receptor and the C5a₂ receptor with four substitution mutants of the human C5a₂ receptor: DRX (Asp-Arg-Phe), NPXXY (Asn-Pro-Met-Leu-Tyr), DRX/NPXXY double mutant, a chimeric receptor carrying the third intracellular loop of C5a₂ receptor, and a triple mutant containing all of the mentioned modifications. Unlike the C5a₁ receptor, none of the mutants exhibited ligand-induced internalization, Ca²⁺ mobilization, or translocation of endogenous β-arrestin 1 in RBL-2H3 cells, and neither did the wild-type C5a₂ receptor (Scola et al., 2009). Taken together, multiple structural features contribute to effectively restrict G protein-coupled signaling by the C5a₂ receptor.

In accordance, several groups have consistently found that binding of C5a, C3a, C4a, or their des-arginated derivatives to the wild-type C5a₂ receptor in vitro does not support G protein-mediated cellular responses such as Ca²⁺ mobilization, degranulation, chemotaxis, or activation of the mitogen-activated protein kinase (MAPK) pathway, in striking contrast to the diverse effects induced by the C5a₁ receptor (Cain and Monk, 2002; Kalant et al., 2003; Okinaga et al., 2003; Johswich et al., 2006; Scola et al., 2009). Pretreatment of human C5a₂ receptor-expressing RBL-2H3 cells with C5a, C5a des-Arg, C3a, or C4a could facilitate FcεRI-mediated secretory responses (Cain and Monk, 2002), implying that, at the most, the C5a₂ receptor exhibits very limited capacities for signal transduction under particular conditions.

b. Internalization of C5a₂ receptor

The C5a₂ receptor has been suggested to act as a decoy receptor, limiting the availability of C5a or C5a des-Arg to competing receptors, especially the C5a₁ receptor (Okinaga et al., 2003; Scola et al., 2009). This concept was first described in 1993 for IL-1RII (Colotta et al., 1993), and has since then been extended to GPCR including D6, DARC, US28, CCX CKR (CCRL-1), CRAM-B (CCRL-2), and CXCR7 (Nibbs et al., 1997; Gosling et al., 2000; Randolph-Habecker et al., 2002; Lee et al., 2003; Burns et al., 2006; Hartmann et al., 2008; Leick et al., 2010). By sequestering ligands and targeting them for degradation, these receptors negatively regulate cellular responses to inflammatory mediators. Decoy receptors are commonly uncoupled from G proteins, most of them due to lack of the DRX motif. Unlike signaling receptors, they do not undergo ligand-induced internalization, but some of them recycle constitutively between the cell surface and intracellular compartments, thus efficiently removing ligands from the extracellular space. This may lead to a relatively high proportion of intracellular receptors, as demonstrated for D6 (Weber et al., 2004).

It has been demonstrated by the use of human PMN or transfected cell lines, including RBL-2H3, that significant amounts of C5a₂ receptor are located intracellularly, and that human C5a₂ receptor is neither phosphorylated nor internalized upon binding of C5a or C5a des-Arg (Cain and Monk, 2002; Okinaga et al., 2003; Scola et al., 2009; Bamberg et al., 2010). Instead, clathrin-dependent constitutive internalization occurs, resulting in net transport of ligand into the cell and in ligand degradation (Scola et al., 2009). This appeared to be more efficient for C5a des-Arg than for C5a, suggesting that the C5a₂ receptor might be responsible for removing the more abundant des-arginated form from the circulation (Scola et al., 2009). Interestingly, a predominantly intracellular location and constitutive internalization were recently reported for FPR3, which, despite being capable of eliciting cellular responses, does not seem to activate G proteins (Rabiet et al., 2011). In the respective study, the N-terminal extracellular region and TM1 were identified to contain the structural determinants of this unusual recycling behavior, which appears to be clathrin- and β-arrestin-independent. Based on these results, the investigators hypothesized that FPR3 may serve a ligand scavenging or intracellular receptor function, in analogy to what has been described for the C5a₂ receptor (Scola et al., 2009; Bamberg et al., 2010). Again, contrasting data have come from studies investigating the C5a₂ receptor as a receptor for C3a des-Arg /ASP: here, ligand-induced clathrin-dependent receptor endocytosis and recycling were observed in HEK293 cells, and were impaired by an S323I mutation, indicating that this residue is vital for C3a des-Arg /ASP-induced C5a₂ receptor activation (Cui et al., 2009b).

c. Arrestin-mediated signaling

Because there is no evidence for the involvement of G proteins in the observed C5a₂ receptor-mediated effects, the idea emerged that the C5a₂ receptor uses G protein-independent signaling, such as pathways targeted by β-arrestins (reviewed in Defea, 2008). A β-arrestin 1-GFP fusion protein was translocated to the human C5a₂ receptor in transfected HEK293 cells after stimulation with C5a, C5a des-Arg, C3a, or C3a des-Arg (Kalant et al., 2005). The presence of large amounts of exogenous protein or varying preferences of the C5a₂ receptor for β-arrestin subtypes depending on the cell type may account for the divergence between these and other results (Scola et al., 2009). Later, two independent studies of C5a₂ receptor-transfected cells concordantly reported C5a-induced translocation of β-arrestin 2 (Cui et al., 2009b; van Lith et al., 2009). In addition, the role attributed to the C5a₂ receptor as a functional receptor for C3a des-Arg (ASP) in lipid metabolism seems to be associated with recruitment of β-arrestin 2-GFP (Cui et al., 2009a). Furthermore, it was recently shown in the human mast cell line HMC-1 that G protein-mediated ERK1/2 activation downstream of the C3a receptor may be negatively regulated by β-arrestins (Vibhuti et al., 2011). Colocalization of β-arrestin 1 with the C5a₁ and C5a₂ receptors occurs in endosomal vesicles of C5a-stimulated human PMNs and was found to be dependent on C5a₁ receptor activation (Bamberg et al., 2010). Assuming a regulatory interplay of the C5a₂ receptor with the C5a₁ receptor, as they had already hypothesized earlier (Gerard et al., 2005), the investigators showed that two putative signaling events downstream of β-arrestin 1, ERK1/2 activation and chemotaxis, are enhanced after antibody blockade of the C5a₂ receptor. However, it remains to be confirmed that β-arrestin 1 mediates C5a₁ receptor-induced functions in human primary cells or transfected cells, as they may as well result from G protein activation (Buhl et al., 1994; Nilsson et al., 1996; Haribabu et al., 1999; Schraufstatter et al., 2009; Nishiura et al., 2010b;). In summary, there are credible indications of an active signaling function of the C5a₂ receptor involving β-arrestin, but further confirmation should be aimed at elucidating how this modulates the cellular response to C5a.

d. Dimerization

For the C5a₁ receptor, homodimerization as well as heterodimerization with CCR5 have been observed, and have been shown to lead to homologous or heterologous co-internalization of phosphorylation-deficient C5a₁ receptor or of CCR5, respectively, upon C5a binding to the wild-type C5a₁ receptor (Huttenrauch et al., 2005; Rabiet et al., 2008). Interestingly, the chemokine receptor CXCR7, which is known to associate with Gα_i but does not elicit any intracellular signaling (Levoye et al., 2009), has been found to heterodimerize with CXCR4. The dimers were constitutively associated with β-arrestin and concomitantly decoupled from Gα_i, resulting in potentiated CXCL12-induced signaling via MAPK pathways to promote cell migration (Decaillot et al., 2011). It is tempting to speculate that a similar interaction could take place between the C5a₁ receptor and the C5a₂ receptor, shifting or switching downstream signaling from G protein-dependent to β-arrestin-mediated events, and thus providing at least some explanation for the unresolved controversies about the functions of the C5a₂ receptor. This possibility has been considered, but no evidence was seen in mouse cells (Chen et al., 2007), and it was not reported in human PMN, even when modulation of C5a₁ receptor signaling was observed (Bamberg et al., 2010). The investigators considered differential regulation of downstream pathways rather than dimerization as the underlying mechanism (Chen et al., 2007; Bamberg et al., 2010). However, more recent evidence points to the formation of hetero- and homodimers by C5a1 and C5a2 (Croker et al., 2012; Poursharifi et al., 2013), although the functional significance is still unclear.

e. Cross-talk with Toll-like receptors

Apparently, nonsignaling 7TM receptors may also act as chemokine chaperones or transporters (Middleton et al., 1997; Nibbs et al., 2003), but to date, none of these functions have been attributed to the C5a₂ receptor. Lately, a cross-talk between TLR and the C5a₁ receptor/C5a₂ receptor has been described, in which TLR activation reduces C5a₂ receptor activity, thereby releasing the C5a₁ receptor from negative modulation and enhancing the inflammatory response to C5a (Raby et al., 2011). This study suggests that in the course of every inflammatory reaction, a balance between TLR-induced cellular hypersensitivity to C5a and the C5a₂ receptor counteracting this effect has to be established, and that the net outcome of the process varies considerably among individuals.

a. Lack of an acylation-stimulating protein knockout model

C3a des-Arg /ASP as a cleavage product of C3 is generated in all situations where the complement cascade is activated by the main activation routes. Unfortunately, it is impossible to generate a knockout mouse with an isolated deletion of C3a des-Arg /ASP without interfering with other functions of the complement system.

b. Problems with C3 knockout animals

Although mice deficient in the precursor molecule C3 have erroneously been described as “ASP knock-out mice” or “ASP-deficient mice” (Cianflone et al., 1999), beyond lacking C3a des-Arg /ASP (Murray et al., 1999c,d, 2000; Xia et al., 2002, 2004) these animals are deficient in main complement effector functions downstream of C3—that is, opsonization due to C3b, inflammation and immune modulation mediated by C3a and C5a, sublytic cell activation, and cell lysis of bystander cells caused by MAC. Thus, C3 knockout mice cannot be considered an adequate model to specifically investigate the role of C3a des-Arg (ASP).

c. Receptor for acylation-stimulating protein

As described previously, the C5a₂ receptor is widely accepted as a receptor for C5a and C5a des-Arg. Its role as a signaling receptor for C3a des-Arg /ASP, however, is controversial (Kalant et al., 2005; Johswich et al., 2006; Johswich and Klos, 2007; Klos et al., 2009). Thus, mice with a C5ar2 gene knockout might be more appropriate for the analysis of ASP-related issues than C3 knockout mice, the caveat being that it remains to be proven that the C5a₂ receptor is indeed a functional receptor for ASP.

d. C3a des-Arg regarded as inert

Traditionally, a large number of researchers working in the complement field consider C3a des-Arg as a relatively stable metabolite without any biologic potential, making it a useful marker for complement activation (Burger et al., 1988; Klos et al., 1988; Hartmann et al., 1993; Stove et al., 1995). Although C5a des-Arg retains some affinity for its receptors, most complement researchers assume that C3a des-Arg does not interact with the C3a receptor (Klos et al., 1992; Ames et al., 1996; Crass et al., 1996; Johswich et al., 2006). It is thus neither internalized nor degraded, remaining detectable in the bloodstream. In line with this point of view, correlations of an increase in C3a des-Arg /ASP in body fluids of humans or animals with a certain genotype or a disease suggest a role of complement activation and inflammation in general, and not necessarily a role of this particular complement cleavage product and its assumed function. This is further stressed by the fact that unbalanced obesity and insulin resistance are thought to be linked to chronic inflammation in adipose tissue (Ouchi et al., 2011), and that complement activation is a typical early component of an inflammatory response (Manabe, 2011).

e. Difficulties working with C3a/C3a des-Arg

Partially due to their charge, C3a and C3a des-Arg are difficult to handle, more difficult than C5a and C5a des-Arg (Hugli, 1975). C3a and C3a des-Arg /ASP show a significantly higher nonspecific binding to cells and surfaces (Burg et al., 1996). For this reason, plastic materials have to be coated with silicone, albumin, or other inert proteins when handling C3a or C3a des-Arg /ASP. Synthetic C-terminal peptides of C3a with high purity and more favorable physiochemical properties have been used to study binding and activation of C3a receptor by C3a (Ames et al., 1996; Ember et al., 1991; Hugli and Erickson, 1977; Kretzschmar et al., 1992). However, such peptides have rarely been used to study the interaction between the C5a₂ receptor and C3a des-Arg, and they mimic the effects of ASP only partially (Murray et al., 1999a).

f. Simulation of specific binding

In addition, an easily misleading, rare phenomenon can simulate specific binding in the absence of any cellular receptor. This occurs when binding studies with adherent cells are performed in cell culture dishes or 96-well plates, or when membranes instead of a sucrose gradient are used to separate free from cell-bound labeled ligands. This binding of ¹²⁵I-C3a or ¹²⁵I-C3a des-Arg (ASP) to the surface of materials occurs despite the presence of albumin or dry milk powder (but because of albumin: Cui et al., 2009a), and can be effectively and dose-dependently competed with nonlabeled C3a or C3a des-Arg, thereby imitating specific binding with a K_d of the “plastic C3a/C3a des-Arg receptor” in the nanomolar range (EC₅₀ of ∼20 nM). This phenomenon is not observed using ¹²⁵I-C5a or C5a des-Arg. To prevent this pseudo-specific nonreceptor binding of ¹²⁵I-C3a and ¹²⁵I-C3a des-Arg /ASP, all plates and materials have to be preincubated with protamine sulfate (or histones) in addition to albumin (Johswich et al., 2006). To address this easily misleading feature of C3a or C3a des-Arg /ASP in binding studies, the use of control cells that do not express the ASP receptor in question are essential. Likewise, when investigating primary cells, the experiment must also be performed in the absence of any cells.

g. Purification of acylation-stimulating protein

The purity of plasma-derived C3a des-Arg (pASP) is another critical issue to be considered, especially in the context of the high concentrations used to show ASP-mediated functions. Human C5a and C5a des-Arg bind to and activate the C5a₁ receptor with an EC₅₀ of ∼1 nM (Burg et al., 1995); the EC₅₀ of C3a needed for C3a receptor activation is only slightly higher (Klos et al., 1992; Settmacher et al., 2003). The first biologic effects of these fragments can even be observed at ∼0.1 nM. Of note, in experiments on metabolic responses, pASP has been used at concentrations ranging from 30 nM to 10 μM, with an EC₅₀ of one to several hundred nanomolars. As C3a/C3a des-Arg and C5a/C5a des-Arg are related cationic peptides sharing common features like molecular weight, isoelectric point, or acid stability (Hugli et al., 1975b, 1981a; Paques et al., 1980; Daumy et al., 1988), it is rather difficult to obtain highly pure preparations of C3a des-Arg from plasma without low contaminations of other fragments still being present. Thus, one must exclude even minor contaminations of C5a/C5a des-Arg or C3a to rule out the possibility of a (still intriguing) effect of these polypeptides on lipid and glucose metabolism. Additionally, when observed more than 20 years ago, mast-cell activation by C3a and C3a des-Arg at concentrations exceeding 10 μM was proposed to be based on “a nonspecific mechanism similar to that of other polybasic compounds” (Fukuoka and Hugli, 1990).

h. Recombinant acylation-stimulating protein

The use of recombinant C3a des-Arg (rASP) could solve this problem. However, recombinant purified protein has been consequently applied only in some experiments of more recent publications (Cui et al., 2009b; Paglialunga et al., 2010). Formerly, it seemed to have been difficult to produce in large amounts (K. Cianflone, personal communication) and was therefore reserved for a few key experiments that revealed rASP-induced effects in the micromolar range (Murray et al., 1997). In any case, adequate controls are needed to exclude contaminations such as LPS, which might interfere with biologic responses in cell culture or in animal models, when rASP is prepared from Escherichia coli (Roy et al., 2011).

a. Introduction

Based on these considerations, and to identify the studies providing the best evidence for the proposed role of C3a des-Arg /ASP and the C5a₂ receptor, this review will address the following questions:

1. How credible is the evidence that ASP is identical with C3a des-Arg, and that it influences energy metabolism?

2. How indisputably has it been demonstrated that the C5a₂ receptor plays a role in energy metabolism?

3. How convincing is the evidence that C3a des-Arg /ASP binds to and signals via the C5a₂ receptor?

4. Can the C3a/C3a receptor or the C5a/C5a des-Arg /C5a₁ receptor be ruled out as confounding mediators and signaling components?

5. Do the effects observed and the concentrations of ASP used in cell culture represent a physiologic situation in vivo?

6. Is the effect on lipid metabolism a main physiologic function of the complement system and ASP? Or should it be considered as one component of “inflammatory” diseases including obesity?

b. Changes after application of an oral fat load in diets

Using a competitive enzyme-linked immunosorbent assay (ELISA), it was shown in 1989 that the serum levels of C3a/C3a des-Arg increased significantly within hours after an oral fat load, but not after challenge with glucose (Cianflone et al., 1989a, 1992). However, this was not affirmed by others (Charlesworth et al., 1998; van Oostrom et al., 2004). In moderately hypercholesterolemic women, ASP levels were altered after hydrogenated fat consumption (Matthan et al., 2001). Plasma ASP concentration and adipose tissue C3 mRNA expression were higher in obese than in lean men, and fasting plasma ASP concentration and C3 mRNA expression negatively correlated with insulin sensitivity (Koistinen et al., 2001). Moderate weight loss of obese women or a prolonged fast led to decreased ASP plasma levels (Sniderman et al., 1991; Cianflone et al., 1995). After gastric bypass surgery in morbidly obese subjects, body weight, ASP, and leptin decreased and adiponectin increased, while plasma lipids and insulin resistance improved (Faraj et al., 2003). Thus, diet and obesity can influence ASP concentrations in plasma.

c. Modified acylation-stimulating protein levels in various diseases and during a modified hormonal status

Plasma ASP was significantly higher in patients with coronary artery disease, and there was an association with plasma triglyceride, very low-density lipoprotein (VLDL) cholesterol, VLDL, apolipoprotein B (apoB), and certain apolipoprotein E (apoE) phenotypes (Cianflone et al., 1997). Children with Prader-Willi syndrome frequently suffer from high ASP levels and dyslipidemia (de Lind van Wijngaarden et al., 2010). Increased levels of plasma ASP, CRP, insulin, cholesterol, and nonesterified fatty acids are present in women with polycystic ovary syndrome compared with controls (Wu et al., 2009). In pregnant women, ASP levels were markedly elevated. They correlated with elevated levels of triglycerides, apoB, low-density lipoprotein cholesterol, and body mass index (BMI) (Saleh et al., 2007). In neonates, cord blood ASP showed a positive correlation with fetal birth weight above average, and with maternal plasma TG (Saleh et al., 2008).

For a complementologist, an increase of C3a des-Arg /ASP in body fluids indicates that an activation of the complement system has taken place. One can assume that besides this (surrogate) marker, all effector molecules that are released during activation of the complement cascade, such as C3a, C5a, C5a des-Arg, C3b, or the MAC, were at least transiently present as well. A direct conclusion regarding a biologic function of C3a des-Arg should thus not be drawn. In any case, it is remarkable that a correlation of complement activation and lipid metabolism has been observed in various settings. This supports the concept that unbalanced obesity is linked to chronic low-grade inflammation, and suggests that drastic fat uptake can rapidly trigger a low inflammatory response.

d. Animal models, complement factor C3, acylation-stimulating protein, and energy metabolism

Serum from obese rats was found to contain almost twice as much C3 as serum from lean rats (Boggs et al., 1998). In C3^−/− mice, which lack significant parts of the complement system, including ASP, a marked delay in postprandial TG clearance, reduced body fat, and lower leptin levels were observed in comparison with wild-type animals. Also, postprandial nonesterified fatty acids were increased in serum/plasma independently of low or high fat diets, and modest changes in insulin/glucose metabolism implied increased insulin sensitivity (Murray et al., 1999d). In contrast and of note, the Wetsel group found no significant differences in the plasma TG, cholesterol, or free fatty acid concentrations of fasted C3^−/− mice compared with their wild-type litter mates. Plasma lipoprotein analyses indicated no significant differences in various lipoproteins and triglycerol in the absence of ASP. The C3^−/− animals were not impaired in their ability to clear TG and free fatty acids from their circulation after an oral fat load (Wetsel et al., 1999). These contradicting results might be caused by differences in the C3^−/− and wild-type mouse strains bred and used by the two research groups in different facilities. It suggests that the genetic background of these animals should be checked thoroughly (e.g., by SNP analysis) to prove sufficient and effective backcrossing.

A third research group cross-bred C3^−/− mice to mice deficient in both apoE and the low-density lipoprotein receptor (Apoe^−/− Ldlr^−/− mice). Compared with controls, mice additionally lacking C3 had higher serum TG levels and a more proatherogenic lipoprotein profile, and lower body weight and fat content. In contrast, there were no differences between factor B-deficient mice, being additionally analyzed where complement activation of the alternative pathway is hampered, and wild-type animals. Moreover, the size of lesions in the aorta was significantly higher in C3^−/− mice than in controls. The investigators concluded that complement activation by the classic or lectin pathway exerts atheroprotective effects, possibly through the regulation of lipid metabolism (Persson et al., 2004).

e. C5ar2^−/− mice and blockade of C5a₂ receptor by antibodies

In studies conducted by or in cooperation with the Cianflone group, C5ar2^−/− mice also showed changes in their metabolism as compared with wild-type controls: these mice are hyperphagic on a low-fat diet. Nevertheless, their body weight and adipose tissue mass did not differ from that of wild-type litter mates. On a high-fat diet, average adipocyte size and adipose tissue triglyceride/DNA content were significantly reduced. Postprandial clearance of TG was delayed in C5ar2^−/−animals, and the synthesis of adipose tissue triglycerides, lipolysis, and fatty acid re-esterification were significantly reduced. Indirect calorimetry measurements suggested preferential fatty acid utilization over carbohydrate (Paglialunga et al., 2007).

In vivo, application of anti-ASP and anti-C5a₂ receptor antibodies did not alter body weight, adipose tissue mass, food intake, or levels of insulin, leptin, or adiponectin. However, the blockade particularly of the anti-C5a₂ receptor induced a significant delay in TG and nonesterified fatty acids clearance and decreased perirenal, hepatic, and muscle TG mass. Adenosine monophosphate-activated protein kinase (AMPK) and lipoprotein lipase (LPL) activity were increased by this treatment. Thus, the ASP/C5a₂ receptor neutralizing antibodies that effectively block the ASP-C5a₂ receptor interaction altered lipid distribution and energy utilization. Moreover, continuous administration of anti-ASP antibodies as compared with an IgG negative control had effects on energy expenditure and glucose storage, increasing in vivo whole-body energy utilization (Cui et al., 2007; Paglialunga et al., 2010).

f. Systemic application of purified acylation-stimulating protein or recombinant acylation-stimulating protein

Injection of 500 μg i.p. of ASP accelerated TG clearance after a fat load in vivo in C57BL/6 mice, and reduced plasma glucose within a few hours (Murray et al., 1999b). Untreated male and female C3^−/− mice had elevated oxygen consumption, increased fatty acid oxidation in liver and muscle, and increased inguinal fat and muscle mRNA expression. Fatty acid incorporation into lipids was decreased but could be normalized by administration of ASP together with oral fat intake (Xia et al., 2004). However, assuming complete resorption of 500 μg of ASP from the peritoneum, its redistribution into an estimated 2 ml of total blood volume should lead to ∼25 μM ASP in the circulation. For pASP, a purity of 99% was ascertained by mass spectrophotometry, potentially permitting up to 250 nM of contaminating proteins with a molecular weight similar to ASP in the bloodstream. Of note, this impedes interpretation of data from the corresponding experiments because ∼10–50 nM C3a, C5a, or C5a des-Arg would already result in maximal cellular responses mediated by the C3a receptor or C5a₁ receptor. Thus, when applying 500 μg of pASP, it cannot be excluded that contaminating factors could be responsible for the observed effects.

Yet there are more convincing studies with a continuous administration of recombinant instead of purified ASP over up to 4 weeks that clearly demonstrate decreased fasting levels of nonesterified fatty acids and decreased energy expenditure in C3^−/− mice, but only small (if any) changes in body weight or food uptake (Paglialunga et al., 2010). As a negative control, saline is applied instead of an inert recombinant protein purified from E. coli in parallel with the rASP. In differentiated adipocytes, 100 nM rASP resulted in a drastic increase in fatty acid uptake in vitro. This was blocked in vitro by incubation with an anti-ASP antibody, excluding the possibility that contaminations, such as LPS, are responsible for the observed effect on adipocytes. The interpretation of the data obtained in the mouse is more difficult due to the absence of a similar specificity control (i.e., inhibition by antibody directed against ASP) and also due to the fact that, surprisingly, the achieved and effective serum steady-state levels of rASP delivered by an osmotic pump were very small (i.e., only in the range of 0.25 nM). This large difference in concentrations needed between the in vitro assay to see cellular responses and the in vivo experiment makes it more difficult to draw conclusions about a common underlying mechanism.

On the other hand, continuous application of antibodies against endogenous ASP with corresponding IgG controls in the animal model resulted in clear effects on the lipid and glucose metabolism (Paglialunga et al., 2010). Moreover, injection of rASP into the third ventricle of the brain inhibits food intake and locomotor activity in rats (Roy et al., 2011). Again, data interpretation is limited by the absence of a negative control which could exclude that impurities in the rASP might account for effects in this system. Further, it is difficult to judge what concentrations of rASP are reached in the cerebrospinal fluid, and whether they are in the physiologic range or rather correspond to a situation of severe neurologic disease such as meningitis. Nevertheless, these are interesting data as they indicate additional functions for the complement system and its cleavage products.

g. In vitro data using purified acylation-stimulating protein, synthetic C3a-related peptides, or recombinant acylation-stimulating protein

There are a variety of older publications using human pASP. They have been presented and reviewed in detail elsewhere (Johswich and Klos, 2007; Klos et al., 2009). Therefore, this review focuses on a selection of key data from more recent publications, preferentially based on the use of recombinant ASP or synthetic peptides. For in vitro studies on human skin fibroblasts (HSF) and murine 3T3 cells, ASP was chemically and enzymatically modified; additionally, synthetic C3a-analogous C-terminal peptides, known ligands of the human C3a receptor, were analyzed (Murray et al., 1999a). The results show that the N-terminal region of ASP plays a minor role in receptor binding (determined by inhibition of ¹²⁵I-ASP binding) and triacylglycerol synthesis. An intact disulfide-linked core region is essential for lipid synthesis but not for receptor interaction. The C3a-analogous C-terminal peptide P117 (containing a C-terminal Arg) can inhibit ASP binding and can partially activate lipid synthesis, suggesting “that this region participates but is not sufficient for complete receptor binding” (Murray et al., 1999a), whereas smaller C-terminal peptides were ineffective.

On the human C3a receptor, P117 has an EC₅₀ in the range of 50 nM (while ASP is completely ineffective) (Martin et al., 1997; Johswich and Klos, 2007; Klos et al., 2009). Triacylglycerol synthesis was stimulated by P117 at 1000-fold higher concentrations of 4–40 μM, similar to the molar concentrations of ASP needed. Surprisingly, in competition studies using 1 nM ¹²⁵I-ASP as a tracer on HSF, unlabeled P117 and ASP were equally potent, with an IC₅₀ in the range of only 100 nM (pIC₅₀ = 7). This behavior is rather surprising as the ligand concentrations needed for binding and dose-response curves are usually quite similar.

As mentioned previously, C3a and C3a des-Arg show high nonspecific binding as well as “pseudo-specific nonreceptor binding” to various surfaces in the absence of any cell, with a K_d close to that observed in binding studies (Johswich et al., 2006). This suggests that the binding of ¹²⁵I-ASP and the surprisingly low K_d observed in the Cianflone group, at least in earlier studies on adipocytes of fibroblasts, are best explained by this effect. Of note, in experiments of the Klos group, neither 100 nM fluorescence-labeled (FL) pC3a nor the des-arginated peptide (FL-ASP) bound to C5a₂ receptor expressed on human embryonic kidney (HEK) cells as compared with HEK control cells, whereas FL-C3a was still able to bind to the C3a receptor (Johswich et al., 2006). In contrast, when similar studies were performed by the Cianflone group, using FL-rASP, HEK cells stably expressing C5a₂ receptor bound and internalized this ligand more strongly than nontransfected HEK control cells (Cui et al., 2009a). Both studies using FL-ligands seem to be thoroughly performed, so there is presently no explanation for the differing results.

With an EC₅₀ of ∼1.5 μM, the rASP used here was much more active in stimulating lipid synthesis and fatty acid uptake than the ASP purified from blood and former preparations. In the same study, C5a₂ receptor-expressing HEK cells bound 1 nM of ¹²⁵I-ASP much better than nontransfected cells or the no-cell control with a K_d of 34 nM. HEK and Chinese hamster ovary (CHO) cells expressing C5a₂ receptor bound FL-rASP in a dose-dependent manner at similar concentrations (Cui et al., 2009a).

h. Summary and assessment of the role of C3a des-Arg/ASP and C5a₂ receptor

The pathogenesis of obesity-related metabolic dysfunction is considered to include chronic low-grade inflammation (Ouchi et al., 2011). Various cytokines and chemokines act as adipokines, simultaneously influencing inflammation and metabolism (Kershaw and Flier, 2004). The complement system appears to be linked with adipokine signaling in multiple ways: cells in the adipose tissue can produce various complement factors (Choy et al., 1992; White et al., 1992), and adiponectin itself can trigger the classic pathway of complement activation (Peake et al., 2008). Moreover, complement factor D has been identified as an adipokine. Considering this connection between obesity and inflammation, it does not seem unlikely that other components of the activated complement system might play a similar dual role as proinflammatory mediators and metabolic modulators. In accordance, activation of the complement cascade as evidenced by the increase of the “activation marker” C3a des-Arg /ASP is associated with changes in lipid and glucose metabolism. In this context the publications of the Cianflone group are very informative.

The majority of investigations using ASP purified from blood (or synthetic C3a-related peptides) and most data from animal models are in accordance with a potential role of ASP in metabolism, but they do not provide final proof that C3a des-Arg is indeed the responsible and active mediator. In particular, they do not exclude that contaminations with C3a or C5a des-Arg, and their interactions with the C3a receptor or C5a₁ receptor, respectively, might be causing some of the observed effects of pASP. A few recent key publications using recombinant C3a des-Arg (ASP) provide much stronger evidence for an active role of ASP and for the C5a₂ receptor as its receptor. However, there are some contradictory results from both animal studies (Wetsel et al., 1999) and in vitro experiments (Johswich et al., 2006) that should be borne in mind when assessing the role of C3a des-Arg as a metabolic regulator.