Forming next-generation antibody–nanoparticle conjugates through the oriented installation of non-engineered antibody fragments† †Electronic supplementary information (ESI) available. See DOI: 10.1039/c7sc02747h

Enabling oriented installation of non-engineered antibody fragments on nanoparticle surfaces to create next-generation antibody–nanoparticle conjugates.


Introduction
The application of nanoparticles as drug delivery vehicles has attracted signicant interest owing to potential benets of these platforms, such as improved pharmacokinetic and safety proles of encapsulated cargo. Several nanoformulations are now marketed, whilst numerous others are under clinical evaluation. 1,2 Many nanoformulations are developed for oncology applications in particular. 3 Nanoparticles can 'passively' accumulate within tumours by exploiting defects in neovasculature endothelial junctions and impaired lymphatic drainage, a phenomenon known as the enhanced permeability and retention (EPR) effect. However, surface functionalisation of nanoparticles with targeting ligands has the potential to signicantly enhance cellular uptake and retention at the tumour site in a concept referred to as 'active' targeting. [4][5][6] A variety of ligands have been explored for such purposes, including aptamers, peptides and carbohydrates, although antibodies are perhaps the most frequently employed. [7][8][9][10] Numerous bioconjugation methods exist to gra these ligands to the surface of nanoparticles, with common approaches including carbodiimide and maleimide chemistries. 11,12 Carbodiimide coupling involves the derivatisation of a carboxyl group with cross-linking agents such as 1-ethyl-3-(-3dimethylaminopropyl) carbodiimide (EDC), followed by their direct conjugation to amines resulting in the generation of an amide bond. Depending on the direction of the conjugation, antibodies can be coupled to nanoparticles by virtue of free amine or carboxyl functionalities on lysine or aspartic/glutamic acid residues, respectively. However, this approach is encumbered by low reaction efficiencies and generates highly heterogeneous nanoconjugates, where optimal orientation and functionality of paratopes cannot be guaranteed due to the high abundance of reactive amine-and carboxyl-containing residues throughout antibodies. Alternatively, maleimide chemistry can in theory facilitate site-specic conjugation to cysteine residues liberated by the reduction of the inter-strand disulde bonds of antibodies. [12][13][14] However, multiple reports have questioned the cysteine selectivity of maleimide conjugation under commonly employed conditions, 15,16 and the resultant bioconjugate would bear a thioether bond that has been shown to be inherently unstable in vivo. [17][18][19] Furthermore this strategy results in the loss of a covalent link between antibody chains. Nonetheless, reports throughout the literature demonstrate the advantages of utilising other site-specic chemistries for the generation of nanoconjugates, including improved antigen binding and greater product homogeneity. 9,12,20,21 Many of these approaches involve the installation of site-selective amino acid residues using site-directed mutagenesis, followed by subsequent pairing with an appropriate reactive group on the nanoparticle surface. 9,14,[22][23][24][25][26] Whilst this approach allows for the oriented presentation of antibody ligands, it is restricted by the need for time consuming and expensive antibody engineering.
Given these challenges, novel strategies are required to rene antibody conjugation to nanoparticles, with emphasis on: (i) optimising the presentation of the antibody for maximal interaction with cognate targets; and (ii) improving the homogeneity and efficiency of conjugation of antibodies in their native state to aid manufacturability. Previously, we described a novel approach for the functional re-bridging of native inter-strand disulde bonds of full antibodies and their constituent fragments, i.e. resulting in no loss of covalent linkage between antibody chains and modifying at positions that are distal from the antibody binding site. [27][28][29][30][31][32][33] This involved the selective insertion of pyridazinedione moieties bearing reactive handles into reduced disulde bonds, thus enabling site-specic incorporation of 'click' domains without impinging upon antibody functionality. Here we describe the application of this disulde re-bridging technology to site-selectively modify trastuzumab (TRAZ) F(ab) to bear a strained alkyne handle distal to the paratope and conjugate it to azidefunctionalised nanoparticles. This novel approach is compared to conventional methods in order to determine the importance of chemical strategy on the performance of nanoconjugates (Fig. 1). We demonstrate superior binding to HER2 of nanoconjugates formed by our new method versus NHS ester conjugation, highlighting the potential of this approach to overcome the shortcomings of conventional coupling chemistries employed in many targeted nanoformulations to date.

Results and discussion
Our study began with the synthesis of a novel heterobifunctional linker that would facilitate conjugation to an antibody disulde at one end and attachment to a nanoparticle at the other. For attachment to the nanoparticle surface, we decided to incorporate the strained alkyne BCN due to its ability to engage in copper free strained-promoted alkyl-azide cycloaddition (SPAAC) reactions and our condence in being able to formulate azide-functionalised nanoparticles (discussed in detail below). Despite wide-spread use within biomedical research, SPAAC has remained largely unexplored for the generation of nanoconjugates and we took this opportunity to appraise it in this light. To impart site-selective protein reactivity to the linker, the BCN moiety was linked to a dibromopyridazinedione, chosen for its exquisite disulde reactive selectivity and the excellent stability prole of antibody conjugates formed thereof. Synthesis of the strained alkyne functionalised pyridazinedione 3 proceeded from readily available starting materials in a facile manner over three steps (Scheme 1, further details on the synthetic methods can be found in S1-S5 in the ESI †). The monoclonal antibody TRAZ was chosen as a targeting ligand due to its clinical relevance as an approved therapeutic against HER2+ breast cancers. 34 The choice to utilise the F(ab) domain of TRAZ as the targeting component was driven primarily by the growing volume of evidence suggesting antibody fragments provide multiple benets to the overall performance of nanoconjugates when compared to full antibodies; F(ab)s retain the binding component of a full antibody. 9 Additionally, the use of the F(ab) domain is ideal in that it only contains a single solvent accessible disulde bond to ensure homogeneous modication and that there is only a single site from which the antibody ligand can be attached to the nanoparticle. Furthermore, F(ab) domains can be readily expressed and/or obtained from native full antibody scaffolds via simple enzymatic digestion procedures. In this particular case, we obtained TRAZ F(ab) from native TRAZ via enzymatic digestion (pepsin followed by papain, further details provided in the ESI †). Site-selective modication of TRAZ F(ab) with strained alkyne-pyridazinedione 3 was achieved according to previously reported protocols 31 ( Fig and S11 †). As an appropriate control to determine the overall effect of the site-directed chemistry, a strained alkyne was incorporated into TRAZ F(ab) using non-site-selective lysine-NHS chemistry (modied TRAZ F(ab) [lys]) ( Fig. S13 and S14 †). Additionally, the F(ab) fragment of cetuximab (CTX) was site-selectively modied to provide a control for antigen target specicity (modied CTX F(ab) [disulde]) ( Fig. S17 and S18 †). CTX was considered a suitable targeting ligand control in this context, given that it does not engage HER2 but rather binds to EGFR, another member of the ErbB family of receptor tyrosine kinases. 35 Following preparation of the various F(ab) domains, we next explored the potential for site-specic functionalisation of nanoparticles with modied TRAZ F(ab) [disulde] 5. To facilitate 'click' conjugation to this antibody fragment, a novel PLGA nanoparticle incorporating a complementary azide moiety was developed. A single emulsion solvent evaporation approach was employed to generate a homogeneous population of azide-terminated nanoparticles from a 25% : 75% polymer blend of PLGA-PEG-azide and PLGA RG502H (nude azide NP) ( Table 1). These nanoparticles were subjected to stability assessment over several months, with no signicant change in physicochemical characteristics observed upon storage at 4 C or À20 C (Fig. S19 †). Conjugation of modied TRAZ F(ab) [disulde] 5 to azide-functionalised nanoparticles was then enabled by incubating both components for 2 h under ambient conditions, yielding a nanoconjugate with a protein loading of 193.1 AE 49.9 pmoles per mg polymer (modied TRAZ F(ab) NP [disulde]) ( Table 1). Contrary to our approach, similar reports of the 'click' functionalisation of nanoparticles via alkyne-azide cycloaddition most oen involve copper catalysis, which can impart toxicities that ultimately limit the biomedical application of the nanoconjugate. [36][37][38][39][40] A range of equimolar controls were also formulated in parallel, which included: (i) native TRAZ F(ab) 4 conjugated to NHS-functionalised nanoparticles  conjugated to NHS-functionalised nanoparticles (native CTX F(ab) NP) and (iv) modied CTX F(ab) [disulde] conjugated to azide-functionalised nanoparticles (modied CTX F(ab) NP [disulde]) ( Table 1). Quantication of F(ab) content within these nanoformulations revealed that conjugation via the strained alkyne proceeded with much greater efficiency compared to NHS ester chemistry (Table 1). These ndings are consistent with the enhanced reaction kinetics of SPAAC click chemistry over NHS ester chemistry and also the improved stability of the surface bound azide when compared to the activated carboxylic acid. Further characterisation experiments included ESEM imaging, which revealed the spherical morphology and uniform size distribution of selected nanoformulations, with similar diameters to those acquired via dynamic light scattering (DLS) in Table 1 (Fig. 3).
Having successfully developed modied TRAZ F(ab) NP [disulde], we next explored the ability of the nanoconjugate to bind to the HER2 target receptor. Initial studies employed surface plasmon resonance (SPR) to examine binding activity towards a HER2 fusion protein immobilised on a carboxymethylated dextran chip. Although binding of native TRAZ F(ab) NP was detected via SPR, an equivalent polymer concentration of modied TRAZ F(ab) NP [disulde] showed a signicantly enhanced binding prole (Fig. 4A). No appreciable binding of nude NP (both NHS and azide) or CTX F(ab) NP (both native and modied) controls was observed, con-rming the dependence of the interaction on TRAZ F(ab). Moreover, HER2 binding of modied TRAZ F(ab) NP [lys] was also negligible despite highly efficient coupling of the fragment to nanoparticles (see Table 1), indicating that this conjugation approach restricted paratope accessibility. This comparison demonstrates that the site-selective nature of pyridazinedione conjugation to F(ab) plays a critical role in the observed improvements in antigen binding; indicating improved paratope accessibility granted by the oriented display of the fragments on the nanoparticle surface.
To exclude the possibility that enhanced binding of modied TRAZ F(ab) NP [disulde] is a consequence of free F(ab) complexation rather than direct coupling to nanoparticles, we also analysed HER2 binding capacity by modied ELISA.
Rhodamine 6G was encapsulated within the various nanoformulations to enable a uorescent readout of binding to immobilised HER2 fusion protein. These studies produced  similar trends to earlier SPR analyses, demonstrating superior binding to HER2 of modied TRAZ F(ab) NP [disulde] versus native TRAZ F(ab) NP and associated controls (Fig. 4B). Consistent with our ndings, direct comparisons of carbodiimide and click chemistry-based approaches for nanoparticle functionalisation have also been described in the literature with a similar enhancement in targeting efficiency conferred by the latter. [41][42][43][44][45] Several of these reports employ full antibodies as targeting ligands whereas our approach offers a distinct advantage through the use of F(ab) fragments comprising a sole disulde bond. This ensures that the click-reactive handle is exclusively installed at a single site located distal from the paratope. The HER2 targeting specicity of modied TRAZ F(ab) NP [disulde] and native TRAZ F(ab) NP was next validated via modied ELISA, where pre-incubation of HER2-coated wells  with an excess of TRAZ full antibody signicantly impeded nanoparticle binding (Fig. 5Ai and 5Aii). These ndings were further bolstered by competition modied ELISA formats, where simultaneous addition of TRAZ full antibody and modi-ed TRAZ F(ab) NP [disulde] to HER2-coated wells inhibited nanoparticle binding in a concentration-dependent manner (Fig. 5B). Collectively, these results demonstrate the successful coupling of TRAZ F(ab) domains to NHS-and azidefunctionalised nanoparticles by distinct approaches, generating active nanoconjugates with retained binding capacity for the cognate HER2 antigen.
We next examined the basis for the enhanced HER2 binding activity of the nanoconjugate generated using the site-selective chemistry, modied TRAZ F(ab) NP [disulde]. To assess whether this effect was simply attributed to enhanced F(ab) loading on the nanoconjugate rather than optimised paratope orientation, we formulated both native TRAZ F(ab) NP and modied TRAZ F(ab) NP [disulde] using various input amounts of the antibody fragments, ranging from approximately 210 to 2100 pmoles per mg polymer. Higher loadings of modied TRAZ F(ab) [disulde] 5 on azide-functionalised nanoparticles were observed, with stepwise increases in coupling that correlated with the initial amount of F(ab) added, highlighting the unique degree of control afforded by this conjugation approach (Fig. 6A). As before, enhanced binding of modied TRAZ F(ab) NP [disulde] was observed upon SPR analysis, even when the F(ab) loading was almost half that of native TRAZ F(ab) NP (Fig. 6A, nanoformulations 5 and 8). This suggests that signicant benets can be achieved even in the case of lower F(ab) loadings, establishing the positive effect of using more controlled (enabling orientation) chemistries. Intriguingly, this data also revealed that HER2 binding activity diminished with higher loadings of modied TRAZ F(ab) [disulde] 5 on nanoparticles, suggestive of potential steric hindrance effects leading to suboptimal paratope display. Using uorescently labelled nanoparticles, these studies were replicated via modied ELISA, with comparable ndings to SPR analyses (Fig. 6B).
Binding of modied TRAZ F(ab) NP [disulde] to HER2 was then evaluated in a more biologically relevant context using cellbased assays. The nanoconjugate was uorescently labelled via encapsulation of nile red and incubated with the HER2-positive HCC1954 breast cancer line (Fig. S20 †), with confocal microscopy demonstrating a clear association of modied TRAZ F(ab) NP [disulde] with these cells (Fig. 7A). Co-incubation with an excess of TRAZ full antibody markedly ablated uorescent labeling of the cells, indicating that the binding of the modied TRAZ F(ab) NP [disulde] was HER2-dependent.
In a nal series of studies, we examined the therapeutic effects of the TRAZ F(ab) nanoconjugates in vitro. These experiments could not be undertaken with the HCC1954 cell line used for confocal analyses, given that it shows limited sensitivity to TRAZ. However, numerous reports have demonstrated that TRAZ full antibody or its constituent fragments can reduce the viability of BT474 breast cancer cells and so this line was deemed an appropriate model following conrmation of HER2 expression (Fig. S20 †). [46][47][48][49] Here, we assessed whether TRAZ F(ab) could induce a similar reduction in viability of BT474 cells when presented in a nanoparticle-bound format. Whilst treatment with free native TRAZ F(ab) 4 led to a gradual reduction in cell viability over time as anticipated, this effect was much less pronounced for the corresponding nanoconjugate (Fig. 7Bi). However, upon treatment with modied TRAZ F(ab) NP [disul-de], the reduction in cell viability was much more comparable to free modied TRAZ F(ab) [disulde] 5 (Fig. 7Bii). These ndings are consistent with enhanced paratope accessibility on modied TRAZ F(ab) NP [disulde] conferred via the site-specic conjugation approach. Importantly, these studies also conrmed that the installation of a pyridazinedione linker did not adversely affect the functionality of TRAZ F(ab). Previous work by ourselves and others has shown that antibody display on nanoparticle surfaces can enhance receptor cross-linking; thus mimicking ligand interactions to enhance downstream biological effects. 8,50,51 However, as HER2 is a ligand-less receptor, it is not surprising that no enhancement of the effect on cell viability was observed with these particular nondrug loaded nanoconjugates.
To conclude, we have described a novel strategy for the site-specic functionalisation of nanoparticles that promotes the uniform and outward projection of paratopes for maximal target interaction. Using TRAZ F(ab) as a model platform, we demonstrate the successful re-bridging of the inter-chain disulde bond with a heterobifunctional linker and subsequent 'click' coupling to nanoparticles bearing complementary azide moieties. The controlled orientation of TRAZ F(ab) afforded by this approach leads to superior binding to HER2 when compared with conventional NHS ester coupling chemistry, with retention of F(ab) functionality in cell-based assays, demonstrating the importance of controlled chemical ligation for nanoconjugate performance. Crucially, these ndings also extend to other receptor-ligand pairings including EGFR-cetuximab (Fig. S21 †), highlighting the versatility of this novel approach and how it may be tailored for diverse applications. This work offers a signicant contribution to the ongoing endeavor to rene nanoconjugate design, with the aim of generating more controlled, homogeneous systems that can be readily translated.

Conflicts of interest
There are no conicts to declare.