Variability of Complement Response toward Preclinical and Clinical Nanocarriers in the General Population

Abstract

Opsonization (coating) of nanoparticles with complement C3 component is an important mechanism that triggers immune clearance and downstream anaphylactic and proinflammatory responses. The variability of complement C3 binding to nanoparticles in the general population has not been studied. We examined complement C3 binding to dextran superparamagnetic iron oxide nanoparticles (superparamagnetic iron oxide nanoworms, SPIO NWs, 58 and 110 nm) and clinically approved nanoparticles (carboxymethyl dextran iron oxide ferumoxytol (Feraheme, 28 nm), highly PEGylated liposomal doxorubicin (LipoDox, 88 nm), and minimally PEGylated liposomal irinotecan (Onivyde, 120 nm)) in sera from healthy human individuals. SPIO NWs had the highest variation in C3 binding (n = 47) between subjects, with a 15-30 fold range in levels of C3. LipoDox (n = 12) and Feraheme (n = 18) had the lowest levels of variation between subjects (an approximately 1.5-fold range), whereas Onivyde (n = 18) had intermediate between-subject variation (2-fold range). There was no statistical difference between males and females and no correlation with age. There was a significant correlation in complement response between small and large SPIO NWs, which are similar structurally and chemically, but the correlations between SPIO NWs and other types of nanoparticles, and between LipoDox and Onivyde, were not significant. The calculated average number of C3 molecules bound per nanoparticle correlated with the hydrodynamic diameter but was decreased in LipoDox, likely due to the PEG coating. The conclusions of this study are (1) all nanoparticles show variability of C3 opsonization in the general population; (2) an individual's response toward one nanoparticle cannot be reliably predicted based on another nanoparticle; and (3) the average number of C3 molecules per nanoparticle depends on size and surface coating. These results provide new strategies to improve nanomedicine safety.

Figure 1.

Nanoparticles used in the study. (A) TEM images of nanoparticles were obtained in-house (L-SPIO NWs and S-SPIO NWs) by a third party (Feraheme) or were reported previously (Cryo-TEM images of Doxil, which is similar to LipoDox used in this study). The size bar is 50 nm for SPIO NWs and 100 nm for Feraheme and Doxil. There are no published images of Onivyde liposomes. (B) Schematic representation of nanoparticles based on the previous literature^,, and TEM and DLS measurements (Table 1). SPIO NWs and Feraheme are coated with dextran chains (yellow). Liposomes are loaded with doxorubicin hydrochloride (red) or irinotecan hydrochloride (green mesh). LipoDox (Doxil) is a highly PEGylated liposome, whereas Onivyde is a minimally PEGylated liposome. The size bar is 100 nm.

Figure 2.

Between-subject variation of C3 levels and the association of C3 and C5a levels after exposure to S-SPIO and L-SPIO nanoworms. (A) Distribution of batch-adjusted C3 levels across subjects after exposure to L-SPIO NWs. A batch-adjusted estimate of C3 levels (molecules per milligram of Fe) for each sample is represented by a bar. Red bars indicate samples with C3 levels that were significantly lower than the model average (p < 0.05), and green bars indicate samples with C3 levels that were significantly higher than the model average. Sample labels indicate their gender (F or M) and their age at time of collection. Sample labels that end in “A” or “B” indicate that more than one sample had that combination of age and gender. (B) Association between batch-adjusted C3 levels and C5a concentration after exposure to L-SPIO NWs. Both C3 levels and C5a. Concentrations are plotted on the log base 10 scale. Each dot represents a separate subject and the color of the dot indicates the comparison to the model average as in panel A. The reported correlation coefficient and p-values are based on the Pearson correlation of the log base 10 values. (C) Distribution of batch-adjusted C3 levels across subjects after exposure to S-SPIO NWs. See the description for panel A. (D) Association between batch-adjusted C3 levels and C5a concentration after exposure to S-SPIO NWs. See the description for panel B.

Figure 3.

Between-subject variation of C3 levels after exposure to Feraheme, LipoDox, and Onivyde. A batch-adjusted estimate of C3 levels (molecules per milligram of drug) for each sample is represented by a bar separately for (A) Feraheme, (B) LipoDox, and (C) Onivyde. Red bars indicate samples with C3 levels that were significantly lower than the model average (p < 0.05), and green bars indicate samples with C3 levels that were significantly higher than the model average. Sample labels indicate their gender (F or M) and their age at time of collection. Sample labels that end in “A” or “B” indicate that more than one sample had that combination of age and gender. Fewer serum samples were tested with LipoDox than with Feraheme and Onivyde due to sample availability. (D) Sources of variation in C3 levels after exposure to different nanoparticles. The proportion of the total variance that can be attributed to variance between subjects (green), variance between replicate preparation of the same subject (blue), and the residual variance (red) that includes technical variation is indicated by the height of the bars and grouped by nanoparticle. SPIO NWs show the highest proportion of variance attributed to between-subject effects. LipoDox shows the lowest proportion of variance attributed to between-subject effects and the highest proportion of variance due to preparation effects.

Figure 4.

Association between C3 levels across nanoparticles. The correlation between C3 levels derived from different nanoparticles was calculated using log base 10 values and a Pearson correlation coefficient. The color of cells within the heat map represents the negative log base 10 p-values (higher value and smaller p-value). Comparisons to a p-value greater than 0.01 are white. The correlation coefficient is the first number reported in each cell followed by the p-value in parentheses. The p-values are not adjusted for multiple testing.

Figure 5.

Comparison of C3 opsonization efficiency per nanoparticle. (A) Number of C3 molecules per nanoparticle was calculated as “number of C3 molecules per milligram of Fe or drug” divided by “number of nanoparticles” as described in the Materials and Methods section and in Table 1. The number was plotted against average diameter for each nanoparticle (Table 1). Excluding LipoDox, there was a significant correlation between size and number of C3 molecules per particle. LipoDox does not fit the trend, likely due to a highly PEGylated coating. (B) Dimensions of the C3b molecule. (C) C3b and the alternative pathway convertase (C3b-green, Bb-red, and properdin-blue) in relation to the sizes of nanocarriers. The size of C3b and C3 convertase is comparable to the size of Feraheme, which could explain the low number of C3 per particle. LipoDox also showed low number of C3 per particle despite having much-larger diameters than Feraheme, most likely due to the PEGylated coating. For simplicity, the protein corona, which is critical for complement binding and activation, is not shown here.