Complement activation by drug carriers and particulate pharmaceuticals: principles, challenges and opportunities

Abstract

Considering the multifaceted protective and homeostatic roles of the complement system, many consequences arise when drug carriers, and particulate pharmaceutical formulations clash with complement proteins, and trigger complement cascade. Complement activation may induce formulation destabilization, promote opsonization, and affect biological and therapeutic performance of pharmaceutical nano- and micro-particles. In some cases, complement activation is beneficial, where complement may play a role in prophylactic protection, whereas uncontrolled complement activation is deleterious, and contributes to disease progression. Accordingly, design initiatives with particulate medicines should consider complement activation properties of the end formulation within the context of administration route, dosing, systems biology, and therapeutic perspective. Here we examine current progress in mechanistic processes underlying complement activation by pre-clinical and clinical particles, identify opportunities and challenges ahead, and suggest future directions in nanomedicine-complement interface research.

1. Introduction

The complement system, comprising >30 soluble and membrane-bound proteins, is an integral constituent of the innate immune system [1]. The lectin, alternative, and classical pathways are generally responsible for activation of the complement cascade (Figure 1). Complement orchestrates various protective, regenerative and inflammatory responses by bridging innate and adaptive arms of immune system, as well as cross talk with biological cascades such as the coagulation (contact) system (Figure 2) [1,2]. The two important defense functions of complement activation are opsonization, and the assembly of pore forming membrane-attack complex [1,3]. However, uncontrolled complement activation is deleterious to the host, causing cellular injury (e.g., through the assembly and insertion of the membrane attack complex into cell membranes), and contributing to inflammation, development and progression of disease (e.g., through liberation of excessive anaphylactic peptides) [2–6].

Figure 1.

Complement activation pathways. Steps A–E shows C1q-mediated classical pathway (CP) activation and formation of the C3 convertase (C4b2a) formation. Steps F & G depict merging with amplification route of the alternative pathway (AP) for more C3b deposition (opsonization) through C3bBb assembly (AP C3 convertase), and eventual formation of AP C5 convertase formation (C3bBb3b). Classical pathway activation can directly form a C5 convertase (C4b2a3b) as well. Steps H–J represent the terminal pathway of the complement system, regardless of initiation pathway, showing formation of the lytic complex C5b-9. Lectin pathway (LP) is triggered through binding of collectins/MASPs (K) to the particle surface, leading to the formation of C3 convertase (steps L–N). The scheme also shows AP tick-over, initiated by meta-stable C3(H₂O) in the fluid phase (step O) eventually leading to the formation of AP C3 convertase (C3bBb) (steps P–R), and its stabilization on binding to properdin (triangle P) (step V). Steps S–U represent a proposed artificial situation, where properdin may act as a pattern-recognition molecule, and on binding may recruit C3b, and form AP C3 convertase (step V). Image credit: Peter Popp Wibroe.

Figure 2.

Complement activation through contact (coagulation) system. The cascade is initiated on deposition of Hageman FXII of the contact pathway of coagulation on a negatively charged surface [Scheme (1)]. On binding, FXII is auto-activated to αFX11a. On adsorption, some misfolded proteins and protein aggregates may also auto-activate FXII. Then, αFX11a activates prekallikrein (PKAL) to form kallikrein (KAL), which cleaves C3 and C5 to generate their respective anaphylactic peptides C3a and C5a without involvement of canonical complement pathways. KAL could further generate βFX11a, which subsequently cleaves C1, resulting in activation of the classical pathway of the complement. In addition to these, αFX11a could also induce coagulation cascade with downstream generation of thrombin, and thrombin in turn is capable of cleaving C3 and C5, liberating C3a and C5a, respectively. Complement activation can also proceed on deposition and activation of FVII-activating protease on polyanionic domains [Scheme (2)] through thrombin activity. Modified and redrawn from [146].

The role of complement in drug delivery systems has roots in early liposome research where liposomes were used as model biological membranes to study how bilayer composition, fluidity, and curvature trigger complement activation, and induce lysis [7]. These studies not only advanced our mechanistic understanding of pore-forming membrane-attack complex (C5b-9 complex) insertion into biomembranes, but also led to the development of liposome-based (including haptenized vesicles) immunoassays for biomarker detection (e.g., antigens and antibodies), and evaluation of serum/plasma complement hemolytic levels [8–12]. Soon, it was realized that liposome-induced complement activation has many consequences in liposome application for targeted drug delivery. For instance, the insertion of the lytic C5b-9 complex into the liposomal bilayer could induce substantial leakage of encapsulated therapeutic cargo [9,13]. Moreover, complement activation was enriching liposomes with opsonic complement fragments C3b and iC3b, thereby priming vesicles for rapid interception by complement receptor-bearing macrophages of the reticuloendothelial system, as well as circulating monocytes, and neutrophils [14–16]. This defensive mechanism was limiting liposome targeting to sites outside the reticuloendothelial system. Other studies identified significant interspecies and interindividual variation in liposome-mediated complement activation pathways [16,17].

Since realization of the role of complement in liposome stability and biological performance, many studies began to examine the multifaceted roles of complement system in broader nanomedicine engineering, performance, and safety [18–20]. As such, the past two decades have witnessed many developments in mechanistic understanding of complement activation by preclinical, and clinical particulate formulations as well as serendipitous discovery of some complement-evading nanoparticles and drug carriers [18–20]. Collectively, these studies have identified physicochemical characteristics (e.g., particle curvature, size, shape and surface properties such as patterns generated by polymers and polymer configuration, and spatial distribution of functional groups) and integrated bio-interfacial parameters that modulate complement activation mechanisms and pathways, and a number of important findings have emerged. Here, we briefly outline lessons learned, and suggest future direction in particle-complement interface research in relation to complement activation mechanisms, formulation design, adjuvanticity, and combination approaches with complement inhibitors.

2. Current progress

2.1. Multi-parametric events in complement activation

Collective evidence from the past two decades has shown that nanoparticle-mediated complement activation is a multi-parametric event, where size, shape, and surface properties seem to cooperatively modulate enzymatic complement cascade [18,20]. For instance, curvature and surface stabilizers (e.g., dextran coating) affect IgM binding, and conformation. In the case of small (100 nm or below) dextran-coated nanoparticles, adsorbed anti-dextran IgM molecules are strained, and assume a staple conformation that triggers C1 binding activating the classical pathway [21]. On the other hand, while IgM binds to larger nanoparticles of similar surface properties equally well, it assumes a planar conformation [21]. This attracts C1 weakly, and hence low levels of complement activation [21]. Contrary to such observations, macrophage clearance tends to increase with particle size when normalizing surface area [21,22]. This may be a consequence of reduced local surface area available for the build up of the alternative pathway amplification loop (Figure 1) (where the size of the alternative pathway C3 convertase is 20–30 nm) on high curvature particles. On low curvature surfaces, the affinity of factor H (a regulatory complement protein) for the opsonic C3b (which is already deposited on the surface) decreases to a level, which is similar to that of factor B for C3b. This shifts the balance from regulation to amplification, thus ensuring covalent deposition of more C3b molecules, which is a signal for rapid phagocytic clearance through complement receptors [1]. Eventually, through many sequential steps, and with predominant involvement of factors H and I, surface-bound C3b molecules undergo inactivation, and cleavage, resulting in the release of many smaller fragments (e.g., C3f and C3c) leaving C3dg bound to the nanoparticles [1]. In addition to topological issues surrounding the surface assembly of C3 convertases, C3b binding is also rate limiting depending on reactivity of the activating surface for covalent binding as well as the available surface area [18,23]. It is estimated that each surface-bound C3b molecule could occupy an area of 40 nm² [21,24]. Due to topological restraints, with high curvature particles the bulk of C3b molecules instead of being deposited on the surface are released into the fluid phase. Thus, many small particles (<40 nm) may escape efficient C3b opsonization; however, this does not mean that small particles are poor activators of the complement system, where anaphylactic peptides may still be liberated due to assembly of fluid phase C3 convertases. Accordingly, complement activation studies should monitor multiple activation products in the fluid phase in addition to C3b opsonization. It is also plausible that ultra small nanoparticles (e.g., particles < 10 nm) may trigger complement indirectly. For instance, this may occur when a few ultra small particles bind to large glycoproteins, where conformational changes in glycoprotein could expose complement-activating domains. Complement activation may even be related to nanoparticle-mediated protein aggregation in the fluid-phase, and their deposition on vascular endothelium.

Contrary to the belief that stealth (PEGylated) nanoparticles can deter protein adsorption and fend off body’s defenses, PEGylated liposomes, including clinical formulations such as Doxil® (and follow-on products), trigger the human complement system [25–28]. The anionic phosphate-oxygen moiety of methoxyPEG-phospholipid moiety was found to be a key culprit in triggering complement, since methylation of phosphate-oxygen prevented complement activation, and opsonization processes with PEGylated vesicles [28]. More recently, C3 opsonization of nanoparticles was not only shown to be continuous and variable in vivo, but also a vital role for non-specific blood protein adsorption in complement activation, including PEGylated vesicles, was identified [23,29]. Apparently, some of these proteins generate necessary antigenic epitopes for docking of a few natural antibodies of IgG class on some, but not all nanoparticles, which subsequently serve as target for nascent C3b attack, and formation of alternative pathway C3bBb-properdin convertases [29]. Dynamics of non-specific blood protein binding might to some extent be responsible for interindividual disparities seen in complement activation kinetics, and opsonization as discussed above [30]. It is also plausible that misfolded or denatured protein aggregates on nanoparticles may also contribute to complement activation through contact system (Figure 2).

The ability of PEGylated nanoparticles to trigger complement, initiated further studies in mapping the effect of polymer density, and mobility on complement activation. Indeed, conformation and/or architectural arrangement of surface protected PEG chains (and other related polymers), which in turn is also modulated by curvature, were shown to shift complement activation from one pathway (e.g., classical) to another (e.g., lectin pathway) (Figure 3) [31], which has now been extended to other coating materials such as polysaccharides [32–34]. Although the mechanistic details pertaining these dynamics are still to be worked out, these observations indicate that such intricate patterning might transiently generate ‘pathogen-mimicking’ clusters for differential docking of complement sensing molecules either directly or through non-specific protein deposition. Intriguingly, the terminal regions of surface projected polymers such as poloxamers and poloxamines, which are used to sterically stabilize nanospheres, resemble structural motifs of N-acetyl-D-glucosamine/D-mannose recognizable by lectin pathway initiators mannose-binding lectin (MBL) and L-ficolin [18]. Thus, in close proximity, polymer dynamics might transiently generate water-coordinated end-terminal clusters of appropriate architecture for MBL and/or L-ficolin docking, thereby triggering complement activation through the lectin pathway [18,31].

Figure 3.

Complement activation pattern of poloxamine 908-coated polystyrene nanospheres. The projected conformation of polyethylene oxide (PEO) chains of poloxamine 908 can shift complement activation from one pathway to another. With Type A nanospheres complement activation is exclusively due to C1q-mediated classical pathway (CP) and alternative pathway (AP). With Type B, complement activation still proceeds through AP, but instead of CP, lectin pathway (LP) is triggered. With Type C nanospheres, complement activation arises from LP, but contribution of AP only arises from the amplification loop. However, with Type C nanospheres, the extent of complement activation is considerably lower than Types B and A.

2.2. Species disparities in complement activation and dangers of extrapolation

There are reported disparities in nanomaterial-mediated complement activation processes among different species [35–39]. Indeed, this goes back to earlier work in liposome pharmacokinetic studies, where ganglioside GM1 liposomes showed long circulating properties in mice, but not in rats [35,40]. The short circulation half lives in rats was apparently due to presence of natural anti-GM1 antibodies and vesicular complement activation, which facilitated liposome recognition by the hepatic Kupffer cells [35]. More recent studies, however, have indicated that assessment of complement activation by various drug carriers and diagnostic nanoparticles in animals or in their sera/plasma may not necessarily represent human events. For instance, the so-called stealth poly(oxazoline)-grafted nanoparticles, which circulate for prolonged periods of time in rodents and resist complement activation, are potent activators of the human complement system, where complement fixation (C3b/iC3b opsonization) on nanoparticles promotes their uptake by primary human phagocytes (Figure 4) [39]. Another example is dextran-coated iron oxide nanocrystals, which trigger complement activation differently in human and mice sera [36–38]. While dextran cross-linking with epichlorohydrin prevented complement activation in mice, neither this step nor modifications of alcohol functionalities in dextran with alkylating and acetylating agent had any effect on complement activation in human sera [38]. Collectively, these studies suggest the importance of testing immune compatibility of particulate diagnostics and drug carriers with human materials, rather than extrapolating results from animal sources.

Figure 4.

Complement activation by poly(2-methyl-2-oxazoline)-coated nanoparticles (PMOXA-NP). Panel (a) SDS-PAGE of human C1q and its binding to PMOXA-NP in saline. C1q did not bind to uncoated NP. (b) In human serum PMOXA trigger complement activation through C1q-mediated classical pathway and promotes NP uptake through C3 opsonization by human macrophages. PMOXA-NP do not trigger mouse complement and remain stealth on challenge with murine macrophages. Modified with permission [39].

2.3. Uncontrolled complement activation and immunosuppression: pitfalls and benefits

Uncontrolled complement activation is deleterious to the host, causing cellular injury and contributing to broader development of inflammatory diseases [6,41]. For instance, intratumoral complement activation by nanoparticles, and particularly nanoparticles that trigger the alternative pathway, can promote tumor growth as shown in a syngeneic tumor model in immunocompetent mice [42,43]. The role of complement in tumor progression is complex and could involve many components and players. For instance, liberated C5a molecules promote influx of immunosuppressive T_reg cells, and neutrophils [44–46]. Among its other roles, C5a can inhibit the production of IL-12 family cytokines in macrophages, and induce macrophage polarization by activation of the nuclear factor-κB [47,48]. Likewise, C3a—C3aR signaling also promote tumor proliferation by inhibiting neutrophil, and CD4⁺ T cell responses [49].

The expression of many complement proteins is also increased in many tumors [41], where some promote tumorigenesis independently of complement activation. One example is C1q, which is expressed in the stroma, and vascular endothelium of several human malignant tumors [50]. The mechanisms by which C1q promotes tumor growth is poorly understood, but C1q is capable of inducing macrophage polarization, and suppressing macrophage nucleotide-binding domain leucine-rich repeat and PYRIN domain containing receptor 3 (NLRP3) inflammasome activation [51]. Furthermore, C1q polarized macrophages express elevated levels of programmed death-ligand 1 and −2, and promote T_reg proliferation [52]. It is therefore tempting to speculate that some extravasated polyanionic nanoparticles could bind intratumoral C1q, and modulate intratumoral macrophage (and other cells) inflammatory responses differently via its collagen-like domain. However, such responses may depend on C1q conformation, and the avidity of the bound particles for C1q receptors gC1q and cC1q (calreticulin).

It should be emphasized that the bulk of experimental anti-cancer nanomedicine research to date has been conducted in immunocompromised/inbred mouse strains, which often have complement deficiencies, and this may have contributed to over-interpretation of therapeutic efficacy [43]. Accordingly, by considering the integrated roles of complement proteins in tumor immunology, the extent of anti-cancer nanomedicine accumulation in tumors and their mode of local complement triggering could play an important role in therapeutic outcome in human subjects. Translational oncology research utilizing nanomedicine-based approaches should therefore consider, and re-evaluate these eventualities. Finally, with increasing application of nanoparticles as theranostics [22], the role of local nanoparticle-mediated complement activation in inflammatory conditions such as atherosclerosis, arthritis, and many skin diseases also requires re-evaluation. It is also tempting to speculate that some pathological conditions, and depending on the disease stage, may benefit through compartmental accumulation of complement activating nanoparticles, which through controlled complement activation could recruit immunosuppressive cells.

2.4. Complement activation and infusion-related adverse reactions: random acts of intervention?

Acute adverse reactions to nanomedicine injection/infusion in human subjects are idiosyncratic, non-IgE-dependent, and often result in cardiopulmonary distress with varying severity as well as other symptoms [53,54]. These responses are complex, but share similarities to immune responses seen with viral gene therapy vectors [55], and perhaps related to immune cell activation (e.g., macrophages, natural Killer cells, and dendritic cells) through crosstalk and cooperative effects between different receptors, such as Fcγ-receptors (FcγRs), and Toll-like receptors [54]. Some reports, however, have suggested a causal role for complement activation, and anaphylactic peptides C3a and C5a in adverse reactions to nanomedicines [reviewed in 56], but clear evidence linking complement to infusion reactions in human subjects (as well as preclinical models) is still missing [53,54]. At least, complement activation was not observed in nonhuman primates, and phase I patients on exposure to a PEGylated RNA aptamer (pegnivacogin) [57]. However, allergic-like reactions to pegnivacogin in a number of human subjects were antibody-mediated (due to high levels of pre-existing anti-PEG antibodies), where relatively large amounts of PEG (64 mg PEG for an 80 kg patient) were delivered in a single bolus dose of pegnivacogin [57]. This is presumably linked to rapid formation, and binding of pegnivacogin-immunoglobulin complexes to FcγRs, and signaling through them. The immunoglobulin-dependent responses in this trial [57] are in line with a study indicating that human sera contain low levels of anti-PEG antibodies, with nearly 7% and 1% of all specimens containing anti-PEG IgG and IgM in excess of 500 ng/mL [58]. However, not all patients with high levels of anti-PEG antibodies experienced allergic reactions to pegnivacogin [57]. This either suggests differences in susceptibility to antibody-triggered reactions among individuals (e.g., signaling threshold and/or desensitization through FcγRs) or differences in properties of anti-PEG antibodies in individual patients.

Since pigs instantly respond to nanomedicine injection with symptoms similar to responsive human subjects [59], a number of studies have tried to establish correlation between nanoparticle-mediated complement activation in human sera, and cardiopulmonary responses in pigs [25,26,56,60–62]. These empirical approaches, however, could not satisfactorily establish and mechanistically validate a role for the extent of complement activation in infusion reactions. For instance, a clinical study involving 29 cancer patients reported acute allergic reactions in 13 individuals on Doxil® infusion; however, Doxil®-induced complement activation was seen in 21 patients [26]. Among the sensitive patients, 12 had elevated plasma sC5b-9 levels (a marker of complement terminal pathway), but among the 16 non-responding individuals, 9 also had significant rises in plasma sC5b-9 level [26]. On the other hand, injection of minute quantities of Doxil® (0.1 mg lipid/kg body weight) induces reproducible cardiopulmonary distress in pigs, while doubling the dose causes cardiac arrest [63]. Doxil®, and many other clinical and preclinical nanomedicines, however, are poor activators of the porcine complement, and especially at such low doses [54,63–65]. The argument that the porcine model is sensitive to low levels of anaphylactic peptides does not hold, since injection of low doses of C5a shows no dramatic hemodynamic disturbances in pigs [63]. In another study, lack of complement activation by intravenous lipid emulsions in human sera was confirmed, whereas intravenous injection of lipid emulsions into pigs induced significant hemodynamic disturbances, which correlated with rapid rises in plasma thromboxane levels [62]. The study concluded this as an exceptional case, suggesting pigs are not ideal models for studying infusion responses to intravenous lipids [62]. Intriguingly, adverse responses to intravenous lipid emulsions despite its wide use are rarely reported. Our own earlier studies with liposomes have also questioned the ‘human-porcine’ analogy [66]. While liposome pre-incubation with HDL substantially reduced complement activation in human serum, injection of liposome-HDL into pigs was associated with a slow, but extended decline in mean arterial pressure with no recovery for the duration of the experiment (30 min) [66]. On the other hand, liposomes alone induced a sharp drop in mean arterial pressure (with a magnitude comparable to the lowest level reached with liposome-HDL) otherwise with full recovery within 5 min, whereas HDL injection alone was harmless [66].

In contrast to the above, a large body of evidence has indicated a direct role for some macrophage sub-populations such as the resident pulmonary intravascular macrophages (PIMs), in the aforementioned reactions in the porcine model irrespective of complement activation [20,54,64,67–69]. PIMs are an important component of innate immune system in pigs, and other species such as sheep, cat and horse (but not humans), contributing to robust clearance of particulate matters from the blood circulation, which is associated with immediate release of proinflammatory cytokines, metabolites (e.g., thromboxane), and other mediators (e.g., procoagulant factors) on ingestion [64,69–72]. Indeed, robust particle clearance by PIMs in pigs and sheep correlates with peak periods of cardiopulmonary distress [20,54,64,69,71], and on PIM destruction adverse responses to particulate matters, and even endotoxin and zymosan (two potent complement activating agents) injection disappears [64,73,74]. Slowing down particle infusion/injection rate [20,75], and thus reducing the rate of particle presentation to responsive immune cells, overcome infusion reactions in pigs. This phenomenon has also been observed in many sensitive human subjects [53,54]. Other approaches that slow the rate of particle recognition and ingestion by PIMs, such as particle ‘hitchhiking’ on erythrocytes, and particle shape modification (e.g., rods and discs), also slow down or overcome cardiopulmonary distress in the pigs [64]. Pre-dosing with minute quantities of drug-free PEG-liposomes has also overcome anaphylaxis on Doxil® injection in pigs [76]. The molecular basis of these desensitization phenomena is currently unknown, but may involve pre-existing antibodies, signaling through FcγRII/RIII, and macrophage polarization [20,54,77,78]. With non-PEGylated and PEGylated liposomes, binding of circulating anti-phospholipid antibodies, antibodies that recognize epitopes presented by proteins that are intercalated into the PEG cloud, and pre-existing anti-PEG antibodies might contribute to FcγRII/RIII desensitization processes.

The exact role of complement, if any, in infusion reactions to nanopharmaceuticals still remain to be deciphered. At the same time, by considering the oversensitivity of the porcine model (due to PIM physiology), it is very clear that nanomedicine safety assessment in pigs is problematic for predicting idiosyncratic human reactions. The porcine model might represent extreme situations as in the human adult respiratory distress syndrome, where ‘induced’ PIMs are abundant, but the model cannot be justified for use in preclinical development phases pertaining nanomedicines, since its over-sensitivity will exclude many promising nanopharmaceuticals from clinical trials [20,54]. Variable responses to nanomedicine infusion seen in human subjects are most likely related to a pre-existing subset of responsive and/or ‘induced’ immune cells (e.g., macrophages, natural killer cells) as discussed elsewhere [54].

3. Challenges and looking towards future

3.1. Protein adsorption

Considering the role of non-specific blood protein adsorption in complement activation, a better understanding of dynamic protein adsorption/desorption processes on pharmaceutical particles as well as improved approaches that could limit protein deposition are required. Conversely, promoting deposition of complement modulators on nanomedicines may minimize or halt complement activation.

3.1.1. Water structure and the role of specific water-mediated interactions in protein adsorption

Many qualitative and quantitative time-resolved proteomic efforts attest to the complexity of protein deposits on many nanoparticle types [23,29,39,79,80], but protein fingerprinting may not necessary reflect real-time orchestrated molecular build-up and proteolytic blood cascade events in complement activation, and in some cases complement proteins may non-specifically deposit without triggering the cascade. Another limitation is inadvertent contamination with lipoproteins on nanoparticle separation from plasma/serum. Notably, some sub-fractions of human HDL (e.g., HDL₃) are enriched with many complement proteins, and regulators to include C3, C4a, C4b, C9, vitronectin, and clusterin (apolipoprotein J) [81], which may be picked on proteomic analysis, again leading to incorrect interpretation. Accordingly, controlled experiments with complement inhibitors seem necessary [39]. Similarly, proteomic approaches do not reveal conformational changes on protein adsorption and/or intercalation that trigger complement activation [29]. Consolidating functional studies are therefore necessary if protein fingerprinting on nanomedicines, and drug carriers is to be correlated with biological performance, and especially complement initiation, and cascade [29,39]. Notwithstanding, the majority of existing proteome interpretations are generic and/or inconclusive, and most likely a reflection of cumulative snapshots of non-specific surface events mediated by hydration- and structured-water at protein and nanoparticle domains. While specific water-mediated interactions in protein complexes have received considerable attention [82], such processes are yet to be translated to protein adsorption, and dynamics on pristine as well as polymer-projected surfaces. Not even two proteins of the same species may have identical hydration domains in a complex medium such as plasma. For instance, individual circulating albumin molecules may be different in terms of their bound fatty acids (or other hydrophobic companions), which ultimately generate a population of species with different hydration domains, and structured water. Therefore, protein binding/deposition may simply be ‘acts of random companionship’ controlled by hydration domains of the particle surface, and the nearby proteins, thus conforming to statistical, and non-specific binding. Interfacial waters may fill cavities, which in turn may mediate hydrogen bonding between a surface and a nearby protein, enhancing affinity without contributing to specificity. Furthermore, depending on surface free energy and nanoparticle surface characteristics such as physical defects, curvature, solvent-accessible surface area of functional groups, hydrophobic pockets and water retention times, biomolecule deposition may be inhomogeneous and patchy (Figure 5). On surface adsorption, solvation patterns of a typical protein may change (e.g., interior water molecules may escape) and this may promote non-specific interactions with the nearby proteins (resulting in protein build-up), or act as a nucleation site for fiber growth (which may generate patterns for antibody binding and complement sensing proteins), or alternatively causing protein destabilization and/or desorption. The ‘water-driven’ phenomenon might explain why so many different blood proteins simultaneously deposit on a typical particle surface, and why individual particles within a typical suspension may have different patterns of adsorbed proteins (or activate complement variably). From the complement point, the dynamics of hydration and structured water on a typical surface may play a modulatory role in spontaneous and ubiquitous formation of C3(H₂O), a rearranged form of C3, which initiates antibody-independent activation of the complement system through C3(H₂O)Bb convertase assembly [83]. Furthermore, pattern recognition by complement sensing proteins such as C1q, mannose binding lectin, collectins, and ficolins is, presumably, modulated by dynamics of hydration water. In support of this notion, structured water has already been indicated to regulate immune recognition [84]. For example, water molecules in the peptide-binding site of the major histocompatibility complex enhance binding energy, and provide mediating hydrogen bonds [84]. This non-specifically allows MHC to bind a large number of epitope sequences with high affinity [84]. Thus, studying the role of water-mediated interactions in surface protein association and folding, and particularly in relation to C3 and complement sensing molecules, may improve our understanding of complement activation kinetics on surfaces, thereby leading to better optimization of drug carrier engineering.

Figure 5.

Electron micrographs of model sulfated polystyrene nanospheres after dipping in fresh human plasma. (a–f) Transmission electron microscopy of polystyrene nanospheres (60 nm) dipped in neat human plasma for 2 min. The micrograph in (a) shows contrasting electron density on a nanosphere surface. This is a reflection of variable density of surface-bound proteins. The curved arrow depicts an electron dense region arising from excessive protein-protein interactions. Micrographs also show inhomogeneous and patchy protein deposition on many nanospheres as well as the presence of many electron dense spikes (small arrows). The spikes may represent nucleation sites arising from protein collapse into compact loops, strands, and turns leading to fibre formation. The arrows in (c) are example of protein deposits/fibers that have bridged adjacent nanospheres. (g) Scanning electron micrograph of untreated polystyrene nanospheres. Unpublished observations (S.M. Moghimi, Newcastle University).

3.1.2. Materials and surface engineering

Considering the role of blood protein binding in complement activation, design initiatives based on coating or grafting with super-hydrophobic low surface energy polymers and copolymers might overcome protein adsorption and resolve complement activation dilemma, but this approach is dependent on overall polymer biosafety, and fate in the body. Alternatively, a more conventional approach, and widely applicable to drug carriers such as polymeric nanospheres and liposomes, is surface fabrication through polymer pairing [85,86]. With conventional PEGylated surfaces, protein persistence at particle surfaces is modulated by polymer chain mobility and deformability [87]. Thus, surface modification with combinations of short and longer chain poly(ethylene glycols) and/or other polymers/copolymers might perturb energetic phases, and affect deformability of longer chain polymers through altered polymer configuration and structured-water arrangements, thereby negatively affecting protein adsorption and/or penetration into the polymer cloud. Indeed, a newly coarse-grained molecular dynamics simulations study has shown that surface pairing with appropriate combinations and proportions of long and short PEG chains (PEG₂₀₀₀ and PEG₅₅₀) stretches PEG₂₀₀₀ brushes from the surface, and minimizes binding and/or intercalation of a model protein into the PEG cloud [88]. When translated into PEG-paired nanoparticles, PEG-pairing dramatically reduced complement activation in human sera, which was not a reflection of increasing surface PEG density [88]. It would be interesting to adopt this strategy through synthesis of copolymers with super-hydrophobic-polyethylene glycol composition to further affect ion pairing [89], and modulate the extent of protein deposition/intercalation further. Another alternative surface strategy is hierarchical double-PEG layering, where the second PEG layer is covalently attached to underlying chemically reactive PEG molecules [90]. Whether this approach could overcome complement activation in human blood remains to be tested, but such hierarchical approaches may not only introduce more complexity to formulation reproducibility and process manufacturing steps, but also compromise the stability of pharmaceutical nanoparticles in the blood circulation.

3.1.3. From serendipity to complement regulatory protein recruitment

By serendipity a number of complement benign drug carriers have been identified. For instance, recently, a brain-specific phage mimetic particle was shown to bypass complement activation [91]. Other examples include non-lamellar liquid crystalline nanodispersions based on citrem [92,93], an anionic citric acid ester of monoglycerides, and liposomes bearing lipids with carboxylic acid and sulfonate functionalities in their bilayer [94]. Although the underlying mechanisms are still to be unraveled, and presumably related to the mode of functional group presentation both individually and in clusters [88], lack of complement activation at least by citrem nanodispersions and the aforementioned vesicles may be related to surface recruitment of the complement glycoprotein factor H, which possesses cofactor activity for factor I mediated C3b cleavage, and hence minimization of C3b opsonization events. Indeed, factor H is known for its binding to glycerol side chain and carboxyl moiety of sialic acid and many virulent pathogens escape complement attack as they express sialylated glycans and polysialic acid, which recruits factor H [95,96]. Many studies are examining interactions between factor H and native heparin sulfate for understanding the underlying mechanisms by which surface-bound factor H interacts with C3b bound surfaces [97]. Understanding of these events could also lead to development of better materials and drug carrier libraries, which could inherently recruit factor H. Taking these into account, there are proven initiatives that at least show heparin-coated nanoparticles are poor activators of the complement system [98]. Again, nanoparticle surface engineering with a phage peptide, which efficiently recruits factor H, has also overcome complement activation [99]. Our own efforts with dextran-coated superparamagnetic iron oxide nanoworms have shown a direct correlation between the amounts of surface deposited C3 and factor H, but elevation of factor H levels in sera/plasma had no significant effect on C3 opsonization [29].

As a protective measure on different biological surfaces, complement activation is further controlled by multiple membrane-embedded proteins as well as circulating effectors that have redundant activity [2]. Although engineering of complement evading nanoparticles could be extended with such inhibitors, some natural vesicles such as exosomes (conserved membranous vesicular of 30–140 nm released from late endocytic compartments) and cell-derived microvesicles carry GPI-anchored CD55 (a regulator of C3 and C5 convertases) and CD59 (an inhibitor of complement membrane-attack complex assembly) that escape complement surveillance [100]. However, some exosomes still trigger complement activation [100], and our own efforts have shown that exosomes derived from some metastatic and nonmetastatic malignant cells are potent activators of the human complement system, activating complement through the calcium-sensitive pathways [101]. To date, increasing evidence suggest exosomes and microvesicles as important players in a multitude of pro- and anti-inflammatory events [101–104]. For instance, exosome levels are elevated in conditions where complement activation is dysregulated (e.g., polytrauma, hemorrhage, septic shock) and the crosstalk between exosomes and the complement system (as well as the coagulation system) seems important in regulating the extent of complement activity, and hence by modulating innate and adaptive immune responses [101–104]. Considering exosome stability and biological barrier penetration properties, and hence their potential applications in drug delivery [105], it is therefore necessary to screen exosomes for complement activation. Additionally, a deeper insight of exosome-complement crosstalk mechanisms might lead to identification of better immunomodulatory strategies to combat inflammatory diseases through exosome bioengineering.

3.1.4. Transformation factor and diametrical outcomes

Dynamic delivery systems such as some forms of micelles may undergo rapid transformation in the blood, and the extent of complement activation by such systems may be related to these transformational events. Examples include pharmaceutical nonionic solubilizers such as Cremophor EL (C_EL) [106] and Pluronics [107,108]. C_EL is manufactured by reacting the triglyceride rich castor oil with ethylene oxide yielding a complex mixture of hydrophilic and hydrophobic components that forms star-shaped micelles in aqueous solution at concentration >60 μg/mL, which triggers human complement [106]. C_EL micelles affect the electrophoretic mobility of both HDL and LDL, and this is believed to occur from incorporation of some of the hydrophobic C_EL components into lipoproteins [109,110]. This partial hydrophobic component depletion transforms the remaining components into forming droplets of 100–300 nm in size, and these are thought to trigger complement [106]. Pluronic micelles such as those based on poloxamer 188 and poloxamer 407 also activate complement, but the extent of activation of terminal complement pathway by these micelles is dramatically reduced in sera with elevated HDL and LDL levels [107,108]. This protective role of lipoproteins is partly due to micellar interaction with apolipoprotein AI, AII, and B-100 [108]. Indeed, apolipoproteins AI and AII are inhibitors of C9 polymerization, which minimizes the generation of complement lytic complexes [111]. Furthermore, Pluronic micelles directly interact with human lipoproteins and chylomicrons, yielding complex species. For instance poloxamer 407 on interaction with human LDL generates two different types of ‘remodeled LDL’ particles; a population of diffused round-shaped electron-dense species of 80–200 nm in diameter, and a second population of relatively less denser but irregularly shaped species [108]. In addition to these, some populations of ‘remodeled LDL’ entrapped clusters of smaller particles [108]. It is plausible that these transformations may have regulatory roles in complement activation. On the other hand, interactions with chylomicrons and VLDL seem to promote complement activation (unpublished observations). Collectively, these observations may have implications in future design, selection and patient-centric application of micellar solubilizers, such as selective administration to cohorts with appropriate lipoprotein composition and circulating levels to minimize complement activation.

3.2. Complement inhibitors, complement-deficient humans, and preclinical models

Availability of human pathway-specific and complement-receptor inhibitors are opening doors for functionalization of future generations of nanomedicines (and perhaps in combination with the ‘don’t eat me signal’ recombinant CD47 [112]) or as a combination strategy for addressing safety challenges surrounding complement activation (e.g., as in complement-mediated tumor growth) [113]. These inhibitors comprise monoclonal antibodies, nanobodies, cyclic and linear peptides, aptamers, and small molecules targeting different complement proteins, proteases, and receptors (Figure 6) [2,113–115]. As for combination strategies, the use of complement inhibitors must be considered cautiously, and particularly in relation to C3 inactivity, since this could promote susceptibility to and reactivation of many bacterial infections, for example in the lung, on repeated dosing of C3 inhibitors. However, a recent study has shown no susceptibility to infections in nonhuman primates on prolonged treatment with C3 inhibitor compstatin [116], thereby strengthening the notion for combination therapies with nanomedicines in conditions where local complement activation is beneficial to disease progression.

Figure 6.

A selection of human-specific complement inhibitors under investigation.

Developments of species-specific complement inhibitors are equally important. This may allow for distinguishing and identifying similarities between human and animal responses (as well as variation between different preclinical models), and eventual validation of animal models for selected cases. An interesting example is Coversin, a 16.7 kDa recombinant protein originally discovered in the saliva of the Ornithodoros moubata tick, which prevents equally efficiently the cleavage of C5 in humans and pigs [117–119]. There are also indications that the human C1 esterase inhibitor is also effective in pigs and cats by acting on the classical pathway [120,121]. Nevertheless, such approaches may be linked with increasing availability of single/multiple complement gene knockout or spontaneous complement deficient animals [36,122–125] for mechanistic studies, and delineating suspect pathways. There are also spontaneous complement deficient large animal models such as hereditary C3 deficient dogs, where affected animals have less than 0.0003% of physiological amount of C3 in their plasma displaying reduced opsonic and chemotactic activities [124]. The CRISPER/Cas9-mediated gene targeting approach has also been used to generate C3-deficient pigs [125]. These C3 knockout pigs could be useful for reassessing the role of porcine C3 in PIM (and other immune cells)-mediated cardiopulmonary distress.

Finally, deficiencies of many complement proteins in human subjects are well known, and their role in inflammatory network is becoming clearer [53,126–129]. Perhaps, patient profiling for complement deficiencies should be considered in trials and treatments with nanomedicines.

3.3. Exploring complement adjuvanticity

Al(OH)₃ (along with potassium aluminum sulfate or alum) is a complement activator, and commonly used as an adjuvant in human vaccines [130]. Studies in complement receptor-deficient mice have shown impaired immunization responses with antigen-bound alum, thus implying a modulatory role for complement activation in immunization [131]. This is in agreement with the role of C3 and its cleavage product C3d (in fluid phase and surface bound) in T_H2 sensitization, B-cell activation and long-term memory [132–134]. Indeed, co-engagement of complement receptor 2 (CD21) through C3dg-tagged antigens/nanoparticles reduces the threshold for B cell receptor signaling, thus ensuring long-term B cell memory and optimal antibody production [132–135].

Polymeric nanoparticles, liposomes, hexosomes, cubosomes and many other particulate entities are gaining popularity in vaccination strategies both as an antigen depot systems and adjuvants [135,136], but show variable responses and effectiveness. The observed differences in the adjuvanticity of nanoparticles may in part be related to differences in the extent of complement activation. Thus, complement profiling per se might improve nanoparticle selection, and identify complement-activating platforms, and immunization routes (including sequential immunization through different routes) for favorable immune responses. Considering the role of complement in prophylactic protection, future studies should further consider and examine the role of complosome (the intracellular complement, which is present in all cells examined thus far), since complosome is apparently a critical regulator of basic cellular processes, where intracellular complement activation can trigger mammalian target of rapamycin (mTOR) activation, affect nutrient influx, augment glycolysis and oxidative phosphorylation, and regulate ROS production [137,138]. For example, resting CD4⁺ T cells have intracellular pools of C3 and C5, corresponding anaphylactic C3a and C5a receptors (C3aR and C5aR1), and the protease cathepsin L [139]. The latter continuously processes intracellular C3 as well C3(H₂O), internalized through a recycling pathway [140], where liberated C3a engages with lysosomal C3aR to induce low level mTOR activation for cell survival. On T cell receptor stimulation, C3 and its cleavage products and cathepsin L are translocated to the cell surface resulting in autocrine activation through C3aR and CD46, which cumulates in the altered expression of LAMTOR5 (hepatitis B X-interacting protein), the glucose transporter GLUT1, the amino acid transporter LAT1, NLRP3 and IL-1β [reviewed in 137,138]. CD46 can also activate intracellular C5 and liberate anaphylactic peptide C5a, which engages with intracellular C5aR1 to amplify ROS production, which in turn drives NLRP3 inflammasome assembly, and subsequent IL-1β production for optimal T_H1 induction [137,138].

4. Conclusions

The complement system plays multifaceted roles in nanomedicine performance; complement activation may not only improve therapeutic outcome, but might also compromise efficacy. During the past two decades we have learned much on complexities surrounding complement activation by different nano- and micro-particles, and unexpected processes have been discovered. In parallel, fundamental discoveries in this period have further improved our molecular understanding of complement activation processes and regulation, and identified crosstalk pathways by which complement regulates homeostasis, and disease pathogenesis. Collectively, these discoveries together with better understanding of systems immunology within the context of diseases will pave the way for future design of immunologically safer pharmaceutical colloids, viral vectors and sensors [55,141], and might introduce selective applicable combination of therapeutic strategies with complement inhibitors, including the newly described fusion protein inhibitors for dual targeting of lectin and alternative pathways [142]. Cancer serves as a good example. Intratumoral complement activation may depend on tumor type, location, developmental, and treatment stage as well as the host’s immune system responsiveness. Accordingly, experiments with large animals bearing spontaneous tumors (e.g., dogs) as well as immunocompetent animal models of human tumors, including ‘humanized’ species (mice) bearing elements of human immunity, should consider the role of complement functionality, variability, and crosstalk on testing the efficacy of anti-cancer nanomedicines [42,43,143]. Thus, in some cases complement activation may be beneficial. For instance, one study reported elevated tumor C3 levels between 1 and 24 h after photodynamic therapy, and increased alternative pathway activity in the serum between 1 and 3 days post therapy [144]. The efficacy of photodynamic therapy, however, was decreased on blocking anaphylactic peptide receptors, suggesting the importance of complement in photodynamic-mediated tumor destruction [144]. Finally, we foresee development of plasmonic, dynamic multi-component, and multi-parametric assay procedures, and platforms for simultaneous monitoring of nanomedicine-mediated complement activation, and crosstalk analysis with immune, and non-immune cells in the coming decade. Here, and by considering in vitro and in vivo differences seen in complement activation, C3 opsonization, and leukocyte responses [145], future efforts should also focus on developing species-specific in vitro complement assays that correlate with the in vivo dynamics of complement activation.