Conjugation Chemistry Markedly Impacts Toxicity and Biodistribution of Targeted Nanoparticles, Mediated by Complement Activation

Abstract

Conjugation chemistries are a major enabling technology for the development of drug delivery systems, from antibody-drug conjugates to antibody-targeted lipid nanoparticles inspired by the success of the COVID-19 vaccine. However, here it is shown that for antibody-targeted nanoparticles, the most popular conjugation chemistries directly participate in the activation of the complement cascade of plasma proteins. Their activation of complement leads to large changes in the biodistribution of nanoparticles (up to 140-fold increased uptake into phagocytes of the lungs) and multiple toxicities, including a 50% drop in platelet count. It is founded that the mechanism of complement activation varies dramatically between different conjugation chemistries. Dibenzocyclooctyne, a commonly used click-chemistry, caused aggregation of conjugated antibodies, but only on the surface of nanoparticles (not in bulk solution). By contrast, thiol-maleimide chemistry do not activate complement via its effects on antibodies, but rather because free maleimide bonded to albumin in plasma, and clustered albumin is then attacked by complement. Using these mechanistic insights, solutions are engineered that reduced the activation of complement for each class of conjugation chemistry. These results highlight that while conjugation chemistry is essential for the future of nanomedicine, it is not innocuous and must be designed with opsonins like complement in mind.

Keywords: complement system; conjugation chemistry; drug delivery; nanomedicine; targeted nanoparticles.

Figure 1

Production and characterization of antibody‐liposome conjugates. (a) Schematic for the two primary conjugation chemistries in this study: DBCO‐azide (a copper‐free click chemistry) and SATA/thiol‐maleimide. (b) Method for producing antibody‐nanoparticle conjugates. Antibodies were modified with a reactive moiety (DBCO or SATA) and fluorescent tag via NHS ester addition. Modified antibodies were conjugated to liposomes containing the appropriate reactive lipid (azide or maleimide). Size‐exclusion chromatography (SEC) was used to purify antibody‐liposome conjugates. The fluorescent tag was used to quantify the antibody‐liposome conjugation efficiency by comparing fluorescent signal in fractions corresponding to unconjugated antibody with fluorescent signal in fractions corresponding to liposomes. Fractions containing antibody‐liposome conjugates were used for subsequent studies. (c) Nanoparticle Tracking Analysis (NTA) was used to measure the size distribution of antibody‐liposome conjugates. Solid lines indicate the mean, and dotted lines indicate standard error of the mean (SEM) (n = 3 batches per condition). Note the similar size distribution for all conjugation chemistries. (d) Mean size of antibody‐liposomes conjugates, determined from NTA (n = 3 batches per condition). Conjugation of IgG increases the liposome size by 10 to 25 nm, as expected due to the 10 nm size of IgG. Note the similar size of DBCO‐azide and SATA‐maleimide conjugation chemistries. (e) Calculated number of IgG molecules per liposome for each conjugation chemistry (See Figures S3 and S4 (Supporting Information) for detailed methods and calculations). 200 IgG molecules per liposome was the target reaction yield, and particles were only used if the actual number of IgG per liposome was within 15% of the target (170 to 230 IgG per liposome).

Figure 2

Conjugation chemistry affects the biodistribution and toxicity of antibody‐liposome conjugates, due to complement activation, especially in the setting of background inflammation. (a) The impact of conjugation chemistry on the biodistribution of antibody‐liposome conjugates was investigated in mice. Non‐specific IgG was modified with a 3‐fold or 12‐fold molar excess of DBCO (IgG(DBCO‐3); IgG(DBCO‐12)), or SATA (IgG(SATA‐3), IgG(SATA‐12)), then conjugated to liposomes. IgG‐liposome conjugates were injected intravenously into mice and biodistribution was measured via radioactivity. PEGylated liposomes (with no IgG conjugated to the surface) were included as a control. (b) In healthy mice, IgG‐liposomes primarily distributed to the liver and spleen, the primary clearance organs. The distribution did not change for different conjugation chemistries. (c) However, in mice with systemic, acute inflammation induced by intravenous‐LPS, liposomes had a 4‐fold increase in lung accumulation, which was further influenced by conjugation chemistry. Increased modification of IgG with DBCO significantly increased lung uptake. (d) Complement activation is necessary for lung uptake, as demonstrated by complete abrogation of lung uptake in mice with knockout of complement protein C3 (C3‐KO). For (b‐d), n = 3‐4 mice per group. (e–g) IgG‐liposomes cause complement activation and associated toxicities in vivo. Mice were exposed to nebulized‐LPS then treated with IgG‐liposome conjugates, and whole blood was collected 10 min later. (e) The concentration of C5a, a direct product of complement activation, dramatically increased after liposome administration. (f,g) A significant decrease in platelet count and increase in hematocrit, toxicities associated with complement activation, also occurred. For (e‐g), comparisons made by ordinary one‐way ANOVA with Dunnett's multiple comparison test, with “Naïve, Sham (PBS) treatment” as the control (n = 3–4 mice per group).

Figure 3

In vitro complement activation profiles further elucidate the disparate mechanisms by which DBCO‐azide and SATA‐maleimide induce complement activation. IgG‐liposome conjugates, with varying amounts of IgG per liposome, were incubated with serum, and C3a concentration was measured via ELISA. Studies were performed for both DBCO (a) and SATA (b) chemistries (n = 2–4 replicates per condition). The data were fit to the Hill Equation and best‐fit parameter values are presented (c). Increased modification of IgG with DBCO (12 vs 3 DBCO per IgG) leads to significantly higher complement activation, as seen in the higher estimate of C_max for IgG(DBCO‐12) (a,c). This observation suggests that DBCO‐azide induced complement activation is driven by DBCO moiety. However, increased modification of IgG with SATA does not significantly change the complement activation profile nor the estimate for C_max (b,c), indicating that SATA‐maleimide induced complement activation is driven by residual free maleimide groups, not the SATA modification. For SATA‐maleimide, maximum complement activation is reached even at low IgG densities, as demonstrated by the low EC50 values. This suggests the main driver of complement activation is free maleimide reactive groups on the surface of the liposome, as opposed to the SATA moiety or IgG conjugated to the liposome. (d,e) The complement activation assay used in (a,b) was performed in the presence of Gelatin Veronal Buffer (GVB) to inhibit the classical pathway of complement activation (n = 2–4 replicates per condition). C3a concentration in the presence of GVB was divided by C3a concentration without GVB to calculate the percent of complement activation due to the alternative pathway. For SATA‐maleimide chemistry, the majority of complement activation (≈80%) is due to the alternative pathway. This is similar for all IgG densities tested (6 to 200 IgG per liposome) and for both SATA‐3 and SATA‐12 chemistries, again suggesting that the maleimide group, not SATA/thiol, drives complement activation. DBCO‐azide chemistry has a larger contribution from the classical pathway compared to SATA‐maleimide. Increased modification with DBCO leads to an increased contribution from the alternative pathway, but only for low IgG densities.

Figure 4

DBCO‐azide conjugation chemistry induces complement activation via protein aggregation on nanoparticle surfaces. (a) Schematic depicting our hypothesis that DBCO‐modified IgG exhibits more pronounced aggregation compared to SATA‐modified IgG, due to the greater hydrophobicity of DBCO moiety. (b) DBCO is 32‐fold more hydrophobic than SATA, per model predicted logP. (c,d) The aggregation of modified IgG in solution (not conjugated to a liposome) was assessed using HPLC‐SEC (n = 3 replicates per condition). (c) The chromatograms show a similar elution profile for all conditions (unmodified, DBCO‐modified, and SATA‐modified IgG). As a positive control, IgG was heat‐treated to cause aggregation, resulting in earlier elution time. (d) The bar graph quantifies the percent of IgG aggregated in each condition. All conditions have <2% aggregates, indicating that DBCO and SATA modification alone do not induce IgG aggregation. Comparisons made by ordinary one‐way ANOVA with Dunnett's multiple comparison test, with unmodified IgG as the control. (e) IgG aggregation on the liposome surface was measured utilizing fluorescence quenching. When aggregation occurs, the fluorescent and quencher molecules come within close proximity, resulting in a decrease in fluorescence intensity. (f) DBCO‐modified IgG showed a significant decrease in fluorescence signal, and therefore significantly more aggregation compared to SATA‐modified IgG. As a positive control, Bis‐NHS Ester reagent was added to crosslink IgG after conjugation to the liposome surface. This control resulted in a decrease of fluorescence of SATA chemistry to an equal level as DBCO chemistry without the Bis‐NHS Ester reagent. Comparisons made by ordinary one‐way ANOVA with Tukey's multiple comparison test (n = 3–9 replicates per condition).

Figure 5

SATA/thiol‐maleimide conjugation chemistry induces complement activation via maleimide first binding to albumin in plasma, with subsequent attack of the immobilized albumin by C3. (a) Schematic depicting our hypothesis that maleimide reacts non‐specifically with thiol‐containing proteins in the blood, mainly albumin. (b) Liposomes were incubated with fluorescent mouse albumin, then analyzed via nanoparticle tracking analysis (NTA), allowing visualization of the fluorescence on individual particles. Example primary data shows total particle count (top row) and particles bound to fluorophore‐labeled albumin (bottom row). A fluorescent signal is visible in liposomes containing PEG2k‐maleimide, but not those with PEG2k‐azide. (c) Bar graph quantifying the fraction of liposomes that are fluorescent, indicates that liposomes with PEG2k‐maleimide conjugate to fluorescent albumin (n = 3 replicates per condition). (d) Liposomes containing a varying mol% of reactive lipid were incubated with serum and the concentration of C3a, a product of complement activation, was measured (n = 4–5 per condition). As the mol% of PEG2k‐maleimide increases, there is a corresponding rise in C3a concentration. In contrast, increasing mol% of PEG2k‐azide shows no significant change in C3a concentration. (e) The data in (d) were fit to the Hill Equation. The low R² value for PEG2k‐azide indicates a poor fit and suggests complement activation is not affected by mol% of PEG2k‐azide.

Figure 6

Rational choice of conjugation chemistry based on mechanisms of complement activation. Based on the mechanisms elucidated above, we devised engineering solutions to reduce complement activation for DBCO (a–c) and SATA (d,e) chemistries. (a) An alternative click chemistry, TCO‐tetrazine, utilizes reactive groups that are significantly less hydrophobic than DBCO. (b) Aggregation caused by TCO was measured using the fluorescence quenching method described in Figure 2c. TCO induced minimal quenching, which indicates minimal IgG aggregation, on par with SATA and significantly less than DBCO chemistry. Comparisons made by ordinary one‐way ANOVA with Tukey's multiple comparison test (n = 5–9 per condition). (c) Biodistribution of IgG‐liposomes using the same methods as Figure 1b, showing that TCO chemistry caused significantly less lung uptake compared to DBCO chemistry. Comparisons made by ordinary two‐way ANOVA with Dunnett's multiple comparison test, with IgG(DBCO‐3) liposomes as the control. Statistical significance for comparisons to PEGylated liposome control is not shown. (n = 3–4 mice per condition). (d) To block the reactivity of free maleimide, cysteine was added to liposomes. Cysteine addition reduced the binding of BSA to maleimide‐liposomes in a dose‐responsive manner (n = 3–4 per condition). (e) IgG‐liposome conjugates were reacted with varying amounts of cysteine, then incubated with serum and analyzed for C3a concentration (n = 4 per condition). The addition of cysteine also reduced complement activation in a dose‐responsive manner.