Variable Antibody-dependent Activation of Complement by Functionalized Phospholipid Nanoparticle Surfaces*

A wide variety of nanomaterials are currently being developed for use in the detection and treatment of human diseases. However, there is no systematic way to measure and predict the action of such materials in biological contexts. Lipid-encapsulated nanoparticles (NPs) are a class of nanomaterials that includes the liposomes, the most widely used and clinically proven type of NPs. Liposomes can, however, activate the complement system, an important branch of innate immunity, resulting in undesirable consequences. Here, we describe the complement response to lipid-encapsulated NPs that are functionalized on the surface with various lipid-anchored gadolinium chelates. We developed a quantitative approach to examine the interaction of NPs with the complement system using in vitro assays and correlating these results with those obtained in an in vivo mouse model. Our results indicate that surface functionalization of NPs with certain chemical structures elicits swift complement activation that is initiated by a natural IgM antibody and propagated via the classical pathway. The intensity of the response is dependent on the chemical structures of the lipid-anchored chelates and not zeta potential effects alone. Moreover, the extent of complement activation may be tempered by complement inhibiting regulatory proteins that bind to the surface of NPs. These findings represent a step forward in the understanding of the interactions between nanomaterials and the host innate immune response and provide the basis for a systematic structure-activity relationship study to establish guidelines that are critical to the future development of biocompatible nanotherapeutics.

Nanoparticles (NPs) 3 are emerging tools that could greatly impact medical diagnosis and treatment due to their potential to target cells and tissues with imaging agents and/or drug payloads (1). Their unique physical aspects, however, present challenges not encountered by small molecule therapeutics; nanoparticles must pass the scrutiny of a resident immune system that is poised to identify and destroy "foreign" intruders. Activation of immune effectors by nanoparticles could compromise their intended activities and/or cause serious off-target effects (2,3).
Lipid-encapsulated particles such as liposomes, micelles, and emulsions represent classes of NPs of varying sizes that have been introduced into the clinical setting. Although the simple phospholipid membrane exterior bears a striking resemblance to biological structures, complexation with PEG to produce a "stealthy" particle that avoids rapid reticulo-endothelial system clearance as well as alternative modifications with metals, unnatural lipids, drugs, and homing ligands result in unusual surface chemistries that may incite immune response (4,5). Such blood contact issues are well known clinically and present a fundamental challenge to the clinical translation of nanotechnologies for systemic applications. For example, Doxil TM , a PEG-coated liposomal form of doxorubicin, can elicit moderate to severe hypersensitivity reactions, often correlated to complement activation, in up to 45% of patients in some series (4). Similar reactions occurring in liposome-treated pigs can be diminished by pre-treatment with complement inhibitors (6). Thus, even PEGylated lipid-encapsulated NPs can activate the complement system, but the fundamental question will be to what degree. Unfortunately, the parameters controlling blood contact response to NPs are only anecdotally known from experience. Clearly, a systematic elucidation of nanoparticle-complement interactions is warranted to afford the rational design of safe nanomedicines.
The complement (C) system is a branch of immunity that can rapidly recognize and respond to bacteria, viruses, and infected cells with a powerful repertoire of opsonins, anaphylatoxins, and cytolytic agents (7)(8)(9). C activation can occur on virtually any unprotected surface and severe tissue damage can accompany intense C activity. C activation is mediated by three major pathways, the classical (CP), lectin, and alternative (AP) pathways. The CP is activated by antigen-antibody complexes and certain molecular patterns, the lectin pathway , and minority shareholders (less than 5%). 1  is activated by microbial polysaccharides, and the AP is activated by a variety of surfaces. In addition, the AP amplifies the activity of all the pathways. Each pathway results in the assembly of the C3 convertases, the central enzymes of the C cascade. C3 convertase cleaves the fluid phase protein C3 into the opsonin C3b and the anaphylatoxin C3a. Further convertase activity directs assembly of the membrane attack complex, and production of C5a, a second anaphylatoxin. The C activation products (C3a, C3b, C5a, and C5b) direct target clearance and membrane lysis (7,10) and promote local inflammatory reactions (11). Complement regulatory proteins defend host cells by inhibiting convertase activity (12), but their protective effects can be overcome by intense local or systemic C activity, whereupon severe tissue damage can occur. In mice, rapid C activation in the circulation induced by cobra venom factor can result in severe pulmonary microvascular endothelium injury leading to acute pulmonary failure. Lung injury is characterized histologically by edema, focal hemorrhage, fibrin deposition, and by formation of neutrophil and platelet aggregates in pulmonary capillaries (13)(14)(15). Local C activation can also produce severe detrimental effects as in the case of ischemia reperfusion injury (16). Perfluorocarbon (PFC) NPs are a class of lipid-encapsulated emulsions with particle sizes ranging between 200 and 300 nm. Typically, the particle exterior surfactant layer is predominantly composed of phosphatidylcholine, and the core is comprised of a perfluorochemical (PFC) that is chemically stable, nonmetabolizable, and intrinsically nontoxic (17). Homing ligands, e.g. monoclonal antibodies, peptides, or peptidomimetics, may be chemically cross-linked to the outer surface of the nanoparticles to afford active targeting to biomarkers. A wide variety of PFC NPs are currently being developed and have shown promise as therapeutic delivery systems and as imaging contrast agents when coupled with gadolinium (18). Here, we use several representative PFC NP formulations to dissect the interactions of lipid-encapsulated NPs with the C system.
We have developed a quantitative approach to examine complement-nanoparticle interactions, and we have used a mouse model to identify and evaluate the in vivo consequences of these interactions that are not evident from in vitro analysis alone. In overview, our results indicate that surface functionalization of NPs with certain chemical structures elicits swift C activation that is initiated by antibody, propagated via the classical pathway, and is independent of charge effects. In other instances, surface functionalization with alternative designs result in more limited C responses. These findings represent a step forward in the understanding of the interactions between nanomaterials and host immune functions and provide the basis for a systematic structure-activity relationship study to establish guidelines to the future rational design of immunocompatible nanotherapeutics.
CH50 Hemolysis Assay-To determine residual C activity (20), NPs (10% v/v) were incubated in 20% human serum in DGVB 2ϩ buffer (150 l total) for 5 min at 37°C. Reaction mixtures were then chilled to 4°C and subjected to centrifugation at 960 ϫ g for 15 min. Resulting supernatants were mixed with DGVB 2ϩ buffer to a total of 800 l, and titration curves were constructed from a series of reactions, each composed of diluted supernatant (150 l) plus 5 ϫ 10 7 (100 l) of IgM-sensitized sheep erythrocytes (EA, CompTech). Reactions were incubated at 37°C for 1 h with shaking, added to 667 l of DGVB 2ϩ buffer, and subjected to centrifugation (1000 ϫ g for 5 min). Degree of cell lysis was determined by measurement of A 414 . A value for complete cell lysis was provided by a control reaction consisting of EA mixed with water. Residual activity of NP-treated serum was compared with the residual activity of serum incubated with buffer alone. CH50 is equal to the dilution factor that results in 50% cell lysis.
A modified CH50 hemolysis assay was used to measure residual C activity of serum collected from NP-treated and control mice (21). In this assay, highly sensitized sheep erythrocytes were prepared by adsorbing additional IgG to commercially prepared EAs (CompTech). First, 1.25 ϫ 10 9 EA cells were resuspended in 1.25 ml DGVB 2ϩ buffer and rabbit anti-sheep erythrocyte polyclonal IgG diluted 1:30 in 1.25 ml of DGVB 2ϩ buffer was then added to EA cells, drop-wise with mixing. The mixture was incubated at 37°C for 15 min with shaking; cells were collected by centrifugation and resuspended in 8 ml of DGVB 2ϩ buffer.
All experiments with mice were performed in strict accordance to guidelines approved by the Division of Comparative Medicine at Washington University. Mice were injected intravenously with 10 l/g of body weight of each NP formulation (ϳ2-5Eϩ12 particles) or PBS. Animals were sacrificed 30 min after injection, and blood was collected from the inferior vena cava. Serum was separated and diluted 1:30 in DGVB 2ϩ buffer and mixed with highly sensitized EA freshly prepared as above, and residual C activity was measured by titration as described for human serum. Titration curves were constructed from a series of reactions, each composed of diluted serum (150 l) plus 1.67 ϫ 10 7 (33 l) of highly sensitized EA. Degree of cell lysis was determined as above.
In Vitro C Activation and Western Blot Analyses-NPs (10% v/v) were incubated in 20% human serum for up to 30 min at 37°C in GVB 2ϩ buffer (20 l total). In some cases, heat-inactivated serum or serum deficient in a particular C protein was used. Where indicated, C protein deficiencies were compensated with the relevant purified protein. Reactions were terminated by addition of 80 l of cold (4°C) EDTA buffer. Samples were centrifuged at 960 ϫ g for 15 min, and supernatants were reserved. NP pellets were resuspended in 100 l of EDTA buffer and washed three times in EDTA buffer. Sample supernatant (1 l in 24 l of SDS running buffer) and washed nanoparticles (entire pellet resuspended in SDS running buffer) were fractionated by SDS-PAGE under reducing conditions, transferred to PVDF, and probed with goat anti-human C3 (1:1200 dilution; CompTech), sheep anti-human C4BP (1:1000 dilution, Abcam, Cambridge, MA), or goat antifactor H (1:1000 dilution, CompTech) followed by the appropriate horseradish peroxidase-conjugated secondary antibody: (1:2000 to 1:10,000 dilution, Jackson ImmunoResearch Laboratories, West Grove, PA). The protein bands were then visualized with a SuperSignal West Pico Chemiluminescent Substrate (Pierce).
In Vivo C Activation and Western Blot Analysis-To examine for C activation in vivo, wild type C57BL/6 (The Jackson Louis) mice were injected intravenously with 10 l/g of body weight of each nanoparticle formulation (ϳ2-5Eϩ12 particles). In some cases, animals were pretreated with mouse IgM (1 mg intraperitoneally per mouse, Rockland Immunochemicals, Gilbertsville, PA) or mouse IgG (5 mg intraperitoneal per mouse, Jackson ImmunoResearch Laboratories). Thirty min after nanoparticle injection, the animals were killed, and blood was drawn from the inferior vena cava. To prepare plasma samples, blood was collected in EDTA-containing tubes and centrifuged for 5 min at 4°C to separate the plasma layer.
To examine for proteins bound to surface of NPs, blood from injected animals was collected in EDTA-containing tubes and centrifuged at 960 ϫ g for 15 min. The pellet was washed four times in EDTA buffer, resuspended in SDS running buffer, fractionated by SDS-PAGE under reducing conditions, and blotted for an IgM chain (1:1000 dilution, Jackson ImmunoResearch Laboratories) or IgG heavy and light chains (1:2500 dilution, Jackson ImmunoResearch Laboratories).

RESULTS
Nanoparticles Activate Human C System in Vitro-Several PFC NP formulations were selected for these studies (Table  1). Each NP featured an exterior shell composed of phosphatidylcholine or lipoid lecithin derived from egg and incorporating 300 -500 molecules (ϳ0.1 mol %) of the ␣ v ␤ 3 -integrin peptidomimetic, a homing ligand (19). Most of the experi-

Complement Activation by Lipid-encapsulated Nanoparticles
ments were conducted with two NP formulations, one incorporating 30 mol % Gd-diethylene-triamine-pentaacetic acid-bis-oleate (GdDTPA-BOA) and the other 30 mol % Gd-DOTA-NH 3 -caproyl-phosphatidylethanolamine (GdDOTA-PE). The phosphate group on GdDOTA-PE gives this NP formulation an overall more negative net charge potential (zeta potential at Ϫ60 mV) compared with the Gd-DTPA-BOA formulation (zeta potential at Ϫ39 mV). An NP conjugated only with the ␣ v ␤ 3 -integrin homing ligand (control NP) served as control.
Intense C activity results in the depletion of intact C3. This effect forms the basis for the CH50 assay, the classic way to quantitatively measure serum C activity by its capacity to lyse antibody-sensitized sheep erythrocytes (EA). We used this assay to measure the capacity of different NP formulations to activate C. Following a 5-min incubation in 20% human serum at 37°C, the NPs were separated by centrifugation, and the serum supernatants were assayed for residual C activity. As seen in Fig. 1, residual C activity was greatly diminished (to Ͻ25% of normal human serum; NHS) by incubation with the GdDOTA-PE NPs but was only partially diminished by GdDTPA-BOA NPs. In contrast, supernatants recovered from reactions with the control NPs were as active as NHS that was not exposed to NPs. Similar results were seen with control NPs treated for up to 30 min (data not shown). Together, these observations indicate that GdDOTA-PE incorporation into the lipid surfactant elicited a robust C response.
Several distinctive C3 fragments are generated during C activation (Fig. 2). Most prominent among them are ␣1 and ␣2, which can be found both covalently bound to target surfaces and in the fluid phase. These fragments were apparent on the surface of NPs (Fig. 3A) or in the reaction supernatants (Fig. 3B) within 1-2 min and reached near maximum by 5-15 min. Addition of EDTA to the reactions abolished C3 cleavage, as C activation pathways require Mg 2ϩ (Fig. 3A, lane 6). When NPs were washed extensively following the reactions and then examined for NP-bound proteins by gel electrophoresis followed by Western blot, we found that ␣1 and ␣2 fragments were prominent on the surface of the GdDOTA-PE NPs and much less pronounced on the control or GdDTPA-BOA NPs (Fig. 3C). We estimated that Ͻ5% of the C3b generated became affixed to a NP surface, whereas the rest remained in the fluid phase. This bias is consistent with other complement-target reactions.
To understand whether C activation is driven by the overall net negative charge on the NPs, we prepared a gadolinium chelate formulation with the phosphate group on GdDOTA-PE O-methylated to yield a more neutral NP (GdDOTA-PE CH3 , zeta potential at Ϫ20 mV, Table 1). Despite the neutralization of the phosphate group, we found that GdDOTA-PE CH3 NPs activated C to the same degree as GdDOTA-PE NPs (Fig. 1). An alternative NP formulation that included GdDOTA-PE without the phosphate group (GdDOTA-DAG, zeta potential at Ϫ20 mV, Table 1) appeared to elicit less robust C activation by qualitative Western blot analysis (Figs. 3D and 6C), but NP stability using this lipid chelate analog was poor, with significant degradation appreciated within days of synthesis, which prevented us from performing the more definitive quantitative CH50 assay. Taken together, these results suggest C activation is dependent on the chemical structure of the lipid-anchored chelates but independent of the overall charge on the NPs.
To determine which pathways were activated by PFC NPs, we incubated the GdDOTA-PE NPs with sera depleted of individual C components and examined the C3 fragments bound to the NPs. During C activation, C1 (along with C4) mediates assembly of the CP convertase, whereas factor B (fB) provides an essential component of the AP convertase. As seen in Fig. 4A, C3 activation was greatly diminished when FIGURE 1. PFC nanoparticles activate C in vitro. Human serum was incubated with the indicated NPs for 5 min, the NPs were separated from the reaction mixtures by centrifugation, and the reaction supernatant added at increasing volume to EA. The percent lysis reflects the residual C activity in the reaction supernatant. GdDOTA-PE and GdDOTA-PE CH3 NPs significantly deplete serum of C activity, thereby reducing the lysis of EA to Ͻ25%, whereas NHS or control (Ctrl NP, ␣ v ␤ 3 -targeted nanoparticles) minimally affects C activation, hence preserving all residual C activity (100% lysis). Values represent mean Ϯ S.E. derived from three independent experiments. FIGURE 2. Generation of C3 activation fragments. During C activation, C3 is first cleaved by the C3 convertase, producing C3a (an anaphylatoxin) and C3b (␣Ј ϩ ␤ chain). C3b is then cleaved twice by factor I in the presence of a cofactor (e.g. the plasma protein factor H), forming iC3b. The ␣1 and ␣2 fragments, diagnostic of C activation, can be seen under reducing conditions. serum was depleted of C1q, an essential component of C1 (lane 4). Supplementation with purified C1 restored C activation/cleavage (lane 5), whereas depletion of or supplementation with fB did not affect C3 activation (lanes 7 and 8). These findings strongly implicate the CP in GdDOTA-PE-dependent C activation while leaving no requirement for the AP.
Although CP facilitated most of the NP-induced C activity, some C3 activation fragments were also observed, albeit greatly reduced, in EDTA reactions (Fig. 3A, lane 6) and in reactions utilizing heat-inactivated serum (data not shown). In most cases, this effect was relatively minor but given the stringent Mg 2ϩ requirement for the three C activation pathways, this observation did suggest the possibility that certain NPs could induce a nonstandard C3 reaction. C3 is known to undergo extensive conformational changes during its transition to C3b (25). Moreover, C3 itself can adopt C3b-like function in the fluid phase (26) or when adsorbed to certain surfaces (27). Thus, we conjectured that when C3 binds to the NP surface, it might tend to take on its C3b-like conformation and become sensitive to factor I (fI) cleavage. This possibility was supported by subsequent in vitro analysis that indicated that PFC NPs bind purified C3, which is then readily cleaved by the fI protease in the presence of the factor H (fH) cofactor (Fig. 4B). C3 fragments generated in this manner may then dissociate from the NP surface into the fluid phase (Fig. 3B,  lane 6).
Nanoparticles Activate Mouse C System in Vivo-To confirm and extend the in vitro studies, we examined NP-dependent C activation in the mouse as this animal model provides the best characterized and most often employed in vivo model for C activation/regulation and complement-dependent disease studies. Mice were injected intravenously with GdDTPA-BOA, GdDOTA-PE, or control NPs at 10 l/g of body weight (ϳ2-5Eϩ12 particles). This dose was chosen because it consistently activated C, although it exceeds the therapeutic doses used in some previous in vivo studies (28,29). Animals were sacrificed at 30 min, and C activation was monitored quantitatively by a modified CH50 assay (Fig. 5) or qualitatively by Western blot analysis of C3 cleavage in the fluid phase (Fig. 6A). As seen in Fig. 5, injection with GdDOTA-PE NPs depleted serum of C activity by ϳ70%. In contrast, injection with PBS, control, or GdDTPA-BOA NPs did not significantly deplete serum of C activity (close to 100% lysis of EA). We also analyzed the sera of treated animals for the presence of the ␣2 fragment, a C3 derivative that is generated during complement activation. Our results indicated that all the NP formulations activated C to some degree (Fig. 6A). However, C activation was significantly more intense following the injection of GdDOTA-PE NPs (Fig. 6A, lane 4), whereas the extent of C activation elicited by GdDTPA-BOA formulation was comparable with that observed with control NPs (Fig. 6A,   FIGURE 3. In vitro C activation depends on surface functionalization of NPs. A, GdDOTA-PE NPs were incubated for the indicated time with 20% human serum, recovered, and washed extensively; NP-bound C3 activation fragments were analyzed by Western blot. The material loaded onto the gel represents all of the washed NPs. B, the serum supernatant from the above reaction was analyzed for C3 activation fragments by Western blot. The material loaded onto the gel represents ϳ1% of each reaction supernatant. Note that the intact ␣-chain is not readily recognized by the detection antibody and is only discernable with overexposure (A). In contrast, the ␣1 and ␣2 fragments were readily visible after 2 min of incubation, and intensity peaked ϳ15 min. The ␤-chain was not cleaved. The ␣2 doublet in A is generated by different factor I-cleaving sites. Addition of EDTA (E) inhibits C activity and C3 cleavage. DVGB 2ϩ buffer (D) supports C activity. C, NPs (10% v/v) were incubated in 20% NHS for 5 min, washed, and analyzed for surface-bound C proteins. The ␣1 and ␣2 fragments were prominent on the surface of the GdDOTA-PE NPs but were much less pronounced on the control and GdDTPA-BOA NPs. D, incubation of NHS with NPs functionalized with GdDOTA-DAG, a GdDOTA-PE without the phosphate group, resulted in slightly diminished C activation. Ctrl, control. FIGURE 4. NP-dependent C activation proceeds through the CP and surface-bound C3 is cleaved by factor I. A, GdDOTA-PE NPs were incubated in NHS, C1q, or fB-depleted (dpl) serum, in EDTA (E) buffer that chelates Mg 2ϩ or DVGB 2ϩ (D) buffer. Where shown, reactions were supplemented with C1 (25 g/ml) or fB (40 g/ml). Generation of C3 activation products ␣1 and ␣2 requires the CP protein C1 but not the AP protein fB. B, C3 protein (200 g/ml) was incubated in 20 l GVB 2ϩ buffer for 5 min at 37°C with the indicated NP (10% v/v) and, where shown, with fH (100 g/ml) and fI (7 g/ml). NPs were washed extensively and surface-bound C3, and C3 fragments were identified by Western blot analysis. Purified C3 and iC3b, lanes 1 and 2, respectively, served as controls. Ctrl, control.  3). No C3 activation products were seen in the mice injected with PBS alone (Fig. 6A, lane 1). These results are consistent with those obtained with the CH50 assay (Fig. 5).
Next, several C protein-deficient strains were utilized to determine the C activation pathways involved in NP-dependent C activation in vivo. As seen in Fig. 6B, GdDOTA-PE-dependent C3 cleavage was greatly diminished in C1q-deficient (lane 5) and C4-deficient (lane 7) mice, whereas the absence of fB did not prevent C activation (lane 9). The dependence on C1q for C activation implicates CP and is consistent with the in vitro experiments with human sera (see above). Activation of the CP was also observed with administration of the GdDOTA-DAG NPs (Fig. 6C). In contrast, the AP does not appear to play an essential role in this process.
Nanoparticle-dependent C Activation Requires Natural IgM Antibody-Although C1 is usually activated upon binding to antibody in the antibody-antigen complex, it can also become activated independently of antibody by directly binding to microbial surfaces or to C-reactive protein prebound to target (9). To determine whether antibody-nanoparticle recognition was required for C activation, we first examined the capacity of NPs to activate C in B-cell-deficient (B Ϫ/Ϫ ) mice, animals in which antibodies are absent. As seen in Fig. 6C, the GdDOTA-PE NPs did not activate C in B Ϫ/Ϫ mice (lane 4). Next, we examined the capacity of NPs to bind antibodies. Wild type mice were injected with control NP, GdDTPA-BOA, or GdDOTA-PE formulations as above, and NPs were recovered from peripheral blood and washed extensively. The NPs were examined for surface-bound IgM or IgG by Western blot analysis. Significant amounts of IgM were found on the surface of all NPs (Fig. 7A), whereas only trivial amounts of IgG were detected (Fig. 7B). Reconstitution with IgM in B Ϫ/Ϫ mice (Fig. 7C, lane 3) fully restored C3 activation in the plasma to wild type level (Fig. 7D, lane 3), whereas reconstitution with IgG (Fig. 7E, lane 3) only led to partial C3 activation in B Ϫ/Ϫ mice (Fig. 7F, lane 3). These findings strongly suggest that antibody-nanoparticle complexes are necessary to activate the complement CP in vivo.
Nanoparticles Bind C Regulatory Proteins-It should be noted that although antibody was required for C activation in vivo, the binding of antibody to the NP surface was not sufficient to initiate the CP, as in the case of control or GdDTPA-BOA NPs (Fig. 7A). Therefore, additional mechanism(s) may be at work. We considered that NP-dependent C activation in  some cases might be tempered by the inhibitory activity of fluid phase C regulatory proteins binding to the NP surface (12). To test this possibility, we incubated NPs in human sera and examined the surface-bound proteins. We observed that fH and C4BP, two C regulatory proteins, bound well to certain NPs. Factor H bound to NPs bearing GdDOTA-PE (Fig.  8A, lane 5). Little fH binding was seen with the GdDTPA-BOA or control NPs (Fig. 8A, lanes 3 and 4). In addition, we found that C4BP bound well to NPs bearing GdDTPA-BOA and, to a lesser extent, GdDOTA-PE (Fig. 8B, lanes 4 and 5). The binding of fH and C4BP to NPs was independent of C activation as it was seen in NHS and C3-depleted serum (Fig.  8, C and D).

DISCUSSION
Previous in vitro studies have shown that negatively charged liposome surfaces activate the CP, whereas positively charged liposome surfaces activate the AP (30,31). Additional in vitro studies have also shown the influence of varied surface chemistry on the C response (32). However, without an animal model, it remains unclear how these findings relate to the actual mechanisms at work in vivo. Here, we describe variable C activation by certain functionalized nanoparticle surfaces. Our results indicate that C activation is not a unique complication of liposomal applications but can be extended to PFC (and likely all lipid-encapsulated) NPs. We demonstrate the essential involvement of the CP in an in vitro and an in vivo model system, wherein CP proteins C1 and C4 (but not AP protein fB) are required for C activity. We also provide evidence that the CP is directly initiated by natural antibody (as opposed to direct C1-nanoparticle interactions or C-reactive protein-mediated C1-nanoparticle complexes) by showing that C activation in vivo is dependent on the presence of IgM.
Using an in vitro human system and an in vivo animal model, we found complete agreement between the mouse and human systems in their capacity to activate C. Although little C activation was observed with the control NP whose surface was composed mostly of phosphatidylcholine (99 mol %) and an ␣ v ␤ 3 -integrin targeting ligand (0.1 mol %), PFC NPs that also incorporate 30 mol % of the negatively charged GdDOTA-PE (Ϫ60 mV) or the O-methylated more neutral form of phosphatidylethanolamine GdDOTA-PE CH3 (Ϫ20 mV) rapidly activated the complement CP, leading to the cleavage of fluid phase C3, the opsonization of the NP surface with C3 fragments, the accumulation of fluid phase C3 fragments and the depletion of residual serum C activity. Complement activation, however, was significantly reduced with another negatively charged Gd chelate, GdDPTA-BOA (Ϫ39 mV). Activation required the CP (C1q-and C4-dependent), whereas the AP and the lectin pathway did not appear to play significant roles in these reactions. Moreover, in B Ϫ/Ϫ mice that were resistant to NP-dependent C activation, this process was restored when IgM and (to a lesser extent) IgG antibodies were supplied to the animals. These results suggest that the GdDOTA-PE-dependent C activation in vivo is initiated by natural antibody-nanoparticle complexes. Antibody bound to the NP is then recognized by the C1 subunit, C1q, causing the activation of C1 and leading to the assembly of surface-bound CP C3 convertase. Although there is insufficient data to identify the major epitope(s) involved in initiating C activation, it is apparent that switching the negatively charged Gd-DOTA phophatidylethanolamine group with an O-methylated, more neutral form did not diminish the intensity of CP activation. Similarly, removal of the phosphate group from GdDOTA-PE did not abolish C activation. These results suggest that the C system recognizes certain functionalized lipids as foreign. Moreover, the intensity of C response is dependent on the chemical structure of the lipid-anchored chelates and not the zeta potential effects alone. We expect the magnitude of C response would also be dependent on the surface load of the reactive lipid chelate conjugates and that lower incorporation rates may ameliorate the C response without precluding MRI diagnostic potential of the NP. We are currently testing this hypothesis.
Host cells and certain microbes avoid C activation in part by recruiting endogenous fluid phase C inhibitors to their surfaces (12,33). To further understand the specificity of NPdependent C activation, we looked for evidence that NP-dependent activation may have been tempered by similar regulatory activities. Factor H is a C regulatory protein that promotes dissociation of the AP convertases (decay acceleration) and serves as a cofactor for fI-mediated cleavage of C3b. We observed that fH strongly binds to NPs that incorporate the GdDOTA-PE chelate and, therefore, could mediate AP inhibition. Similarly, C4BP is a CP regulator that promotes dissociation of the CP/lectin pathway convertases and serves as cofactor for fI-mediated cleavage of C4b. We observed that C4BP strongly binds to NPs that incorporate GdDTPA-BOA and, therefore, could dampen CP activation. In the end, NPdependent C activity may reflect the net balance of two opposing types of reactions, those that activate C and those that inhibit C activation.
In summary, we developed a quantitative approach to examine complement-nanoparticle interactions. We used a mouse model to identify and evaluate the in vivo consequences of these interactions that are not evident from in vitro analysis alone. Our findings and those of the liposome studies raise the possibility that serious side effects may arise from a variety of nanomaterial-based therapies. Unless C activation is terminated by soluble and cell-bound C inhibitors, it will result in the systemic generation of C3a and C5a anaphylatoxins, potentially leading to adverse cardiovascular and pulmonary responses (13-15, 34, 35). Opsonization of NPs with C3 activation fragments could, in principle, accelerate their clearance and thereby diminish intended pharmaceutical effects. In addition, unregulated C activation could also cause the systemic depletion of C3, leaving the host susceptible to natural pathogens. The introduction of NPs into the body may also elicit immune responses in addition to C activation. Studies have shown that NPs can enhance or suppress both humoral and cell-mediated immune responses, depending on nanomaterial structure and composition (2,3,36). Given the current difficulties in predicting the bioreactivity of nanomaterials, a systematic and comprehensive investigation of NP structure-activity relationships with immune cells and blood protein components would be critical to the future development of biocompatible NPs.