Controlling colloidal stability of silica nanoparticles during bioconjugation reactions with proteins and improving their longer-term stability, handling and storage

Please note that technical editing may introduce minor changes to the text and/or graphics, which may alter content. The journal’s standard Terms & Conditions and the Ethical guidelines still apply. In no event shall the Royal Society of Chemistry be held responsible for any errors or omissions in this Accepted Manuscript or any consequences arising from the use of any information it contains. Accepted Manuscript Journal of Materials Chemistry B


Introduction
The rapid growth in nanotechnology research over the past two decades has led to the development of a 'nano-Particle Zoo', housing a myriad of nanomaterials (such as nanoparticles, nanotubes, nanoclusters, nanobubbles, nanostars etc.) with exciting ranges of properties. Though these materials can sometimes be difficult to predictably engineer 1-3 , many have shown highly desirable characteristics for industrial applications that promise to influence our daily lives 4-6 . Unfortunately, very few have been converted from the laboratory benchtop into commercial success. Importantly, two provocatively-titled contributions by Juliano 7 and Venditto 8 outstandingly described the difficult path for the clinical development of nanoparticle-based therapies and warned against 'overpromising' the benefits of this technology. This slow progress can be attributed to a number of factors, however some stand out as the major stumbling blocks: i) lack of robustness and reproducibility in nanoparticle surface functionalization, (ii) the use of expensive, specialized reagents iii) inadequate protocols for the nanomaterial handling that result in excessive manipulation of the samples by the end-user and (iv) limited scope for large scale production. Herein we attempt to address the first three issues by efficiently generating colloidally stable antibody-coated fluorescent silica nanoparticles (Ab-NP) through the use of multivalent linker molecules. The most common method of antibody-nanoparticle conjugation is through covalent bond formation via a linker molecule. The use of a linker creates a space between the protein and the nanoparticle surface, thus avoiding biomolecule denaturation or undesirable biomolecule-nanoparticle interactions. One of the major problems with the published protocols describing the bioconjugation reactions with nanoparticles is that they ignore the effect of the crosslinker introduction, which alters the surface chemistry of a given colloid and often leads to particle aggregation. Conventional colloid science prevents aggregation of 'plain' NPs by introducing capping agents to the NP surface or by tailoring surface charges to separate NPs via electrostatic repulsions 9 . Bagwe et al 10 , for example, preserved the colloidal stability of silica nanoparticles by coating their surface with 'chemically inert', negatively charged phosphonates alongside 'reactive' -NH 2 groups. There has, however, been little focus on how the characteristics of the crosslinker influence nanoparticle stability, and whether this, in turn, can play a role in antibody binding efficiency and Ab-NP polydispersity. Generating Ab-NPs is usually a two-step process: the attachment of crosslinkers to the 'plain' NP surface (Fig 1B, Step 1), followed by the attachment of antibodies to the NPs via the crosslinker (Fig 1B, Step 2). It is sensible to assume that sample monodispersity is required throughout these steps and appears to be a natural extension of conventional methods of inhibiting 'plain' NP aggregation. This implies that linker's physiochemical properties must be carefully considered before performing the chemical reactions. The lack of attention paid to this issue is surprising, especially when the nanoparticle size and its surface properties are considered to be the two most important parameters for the mediation of biocompatibility of new material used in nanomedicine 11 . This view was reinforced in several excellent reviews by Stellacci 12 , Tan 13 and Thompson 14 , in which the authors summarized the main forces governing the interfacial interactions between nanoparticles and variety of biological systems. Previously, we showed the benefits of multivalent, hyperbranched polyamidoamine (PAMAM) dendrimers as crosslinkers 15 . The use of such macromolecules improved antibody binding efficiency and subsequently enhanced Ab-NP binding kinetics in immunoassays 16 . Persisting with this multivalent approach, we have reviewed and further optimized the surface conjugation strategies to graft full size model proteins (goat anti-human IgG and Bovine Serum Albumin (BSA)) and also single chain antibody fragments (recombinant anti-CRP ScFv) onto the surface of dye-doped NPs. We have employed three different multivalent crosslinkers: generation 4.5 PAMAM dendrimer and two additional inexpensive macromolecules: 40kDa dextran and BSA (Scheme 1). Dextran and BSA are more than 10 times less expensive than PAMAM dendrimers and therefore contribute to the reduction of the cost of generating Ab-NPs. The different multivalent linkers present important consequences on the colloidal stability of the NPs and are accountable for the overall efficiency and robustness of the bioconjugation protocols. The prepared antibody-coated nanoparticles were scrutinized in a simple fluorescence-linked immunoassay to qualitatively assess their active surface area (i.e. area of the nanoparticle surface that is covered with antibodies in their active state). We also highlight the misleading depictions of Ab-NPs in many articles that regularly show tethered antibodies oriented such that their epitopes are free and available for binding. Although the scenario of predominantly randomly positioned antibodies on the particle surface is more realistic and appropriate (Fig 1C), many scientists opt to use the 'ideal' and quite unrealistic illustrations. For this reason, we stress the importance of Ab-NP's 'active surface area' throughout this article. Importantly, we have also investigated the effect of cryoprotectants 17 during freeze-drying to improve the long-term storage and colloidal stability upon re-constitution of Ab-NPs. Apart from the traditional sugar-based protectants, we have also used BSA in attempt to reduce the different physical and chemical factors that may destabilize the system, notably the freezing and dehydration stresses during the phase separation into ice and cryo-concentrated solution. Two commonly available analytical techniques, dynamic light scattering and fluorescence spectrophotometry, were used to characterize and analyze the prepared nanomaterials. Other techniques such as transmission electron microscopy, scanning electron microscopy, and confocal scanning laser microscopy have been used to support and corroborate the results.

Using G4.5 PAMAM for bioconjugation optimization
Dye-doped silica nanoparticles were prepared using a wellknown, reverse microemulsion method. The synthesis was adapted from Bagwe et al 10 . The synthetic protocol includes nanoparticle surface modification step using siloxane derivatives with both 'inert' phosphonate groups and 'reactive' amino groups in a 10:1 ratio (Fig. 1b). Such specific composition of the NP-shell displays unique electrostatic properties (strongly negative zeta potential and relatively low isoelectric point), due to which the colloid maintains its stability in aqueous solutions. Because of the limited amount of the reactive amino groups on the NP surface, multivalent linkers are employed to increase the number of potential binding sites for biomolecule attachment. We have previously reported on the effects of different half-generations of PAMAM dendrimer with COOH surface groups to couple proteins onto silica nanoparticles 16,18 . To successfully perform this reaction, the COOH groups of the crosslinker must be activated. This activation is routinely accomplished using a mixture of NHS/EDC to convert the carboxylic acids to amine-reactive NHS-esters. Such activated dendrimer is added in large excess (~10 6 PAMAM : 1 NP) to the NPs solution to avoid unwanted cross-linking of individual particles. As illustrated in scheme 1, NHS-esters are neutral, uncharged species. We hypothesized that the activation of PAMAM dendrimers (containing 128 terminal carboxy groups) with EDC/NHS renders it neutral, and upon attachment to the plain NP surface its neutrality is inherited by the negatively charged NPs, thus lowering the stability of the resultant NP-linker colloidal system. This could in turn lead to nanoparticle polydispersity during antibody attachment (Fig 1B, Step 1 and 2). However, a second form of succinimide is available for COOH activation: sulfo-NHS, which is known to be more water soluble and stable against hydrolysis than NHS 19 . The presence of the negatively charged sulfonates (-SO 3 -) (Fig. 2, bottom) on the activated dendrimer could maintain the necessary electrostatic repulsion to keep the particles highly monodispersed during NP -linker attachment ( Fig 1B, Step 1). This was a prerequisite for the subsequent, successful protein coupling. We tested this hypothesis using dynamic light scattering (DLS) to monitor the particle size evolution during the reaction between NPs and the activated dendrimers (Fig 2A) and the succeeding protein reaction with the dendrimer-coated NPs (Fig 2B, 2C). Prior to the addition of the activated dendrimer, the NP solutions were monodisperse, as shown by TEM (Fig S.1A) and DLS ( Fig. S.4). Upon mixing the NPs with the activated dendrimers (using either EDC/NHS or EDC/sulfo-NHS), significant difference in particle aggregation was observed between the two samples ( Fig. 2A). Substantial clusters of NPs were detected in the case of NHSester activation during the course of 30 minutes (Fig 2A, top). This broadening in size distribution is attributed to the attachment of the uncharged PAMAM dendrimer and increase in colloidal instability. This was confirmed by monitoring NP zeta potential. Plain NPs exhibited a zeta potential of -42.2mV and upon binding PAMAM activated by EDC/NHS the resultant linker-coated NP had a zeta potential of -22.5mV ( Fig  S.1B). In this case, this drop in zeta potential of the colloid, often described as 'charge', was insufficient for retaining particle stability in solution and therefore led to partial sample aggregation (Fig 2A, top, Scan 3). On the contrary, by activating PAMAM with EDC/sulfo-NHS and reacting it with NPs, the resultant NP-linker colloidal system retained its stability ( Fig. 2A, bottom). We attribute this to the fact that the dendrimer activation strategy generates a negatively charged macromolecule, and its physicochemical properties are inherited by the NP. This ultimately benefits the NP-linker colloidal system in terms of its overall stability and size population distribution. Unreacted dendrimer was removed from the NP-linker mixture and the colloids were incubated with the protein (Fig 2B). As a model, we have used Cy5labeled BSA, with an isolelectric pH of 4.7 20 . The isoelectric point of the dye-doped silica nanoparticles depends on the fluorophore and it is estimated between 2.0 -6.3 (data not shown). Notably, the hydrolysis of both NHS-esters and sulfo-NHS esters is much slower at lower pH. We have therefore chosen 0.1 M 2-(N-morpholino)ethanesulfonic acid (MES), pH=4.2 as the optimal buffered solution for the protein conjugation. Aggregation occurred rapidly during the protein conjugation reaction with the NHS-ester activated dendrimer-NPs (Fig. 2B, top). A very broad NP size distribution evolved over the course of the reaction, which carried through to the purified NP product (Fig. 2C, top). However, the sulfo-NHS activation preserved the stability of the colloid during conjugation (Fig. 2B, 2C, bottom). We attribute this to the state of the linker-coated NPs before being reacted with the protein. Importantly, the differences in the colloidal stability of the samples were visible to a naked eye after only 12 hours at 4˚C (Fig. S.1D). The monodisperse sulfo-NHS activated dendrimer-NPs were 40% more efficient in the coupling reaction with proteins compared to that of its neutral analogue, as determined by fluorescence measurements (Fig. S.1C). Such increased binding efficiency is presumably due to the charge nature of the overall linker-coated NPs when reacted with proteins. By incorporating the anionic nature of an EDC/sulfo-NHS activated PAMAM dendrimers, the linker-coated NPs displayed a zeta potential of -33.5mV (Fig. 3) and were therefore highly electrostatically attractive for the positively charged BSA molecules (at pH 4.2). Due to the high probability of BSA being proximal to the linker-coated NPs in solution, there was therefore an increased likelihood of covalent bond formation to generate Ab-NPs. It is also important to note that the use sulfo-NHS-ester activated NPs in this bioconjugation reaction with proteins allowed for the formation of Ab-NPs of narrow size distribution. For example, in nanomedicine, it is important to develop nanomaterials within a given size range (typically <200 nm) such that they can avoid clearance by the reticuloendothelial system, reduce toxicity and successfully locate their target site of interest via the EPR effect 11 . In our case, this strategy generated Ab-NPs with an average size of 135 ± 11nm in diameter ( Figure 3).

Conjugation of anti-CRP, a single chain variable fragment (scFv): A demonstration of 'Active Surface Area' of NPs.
To demonstrate the robustness of the sulfo-NHS conjugation strategy and to qualitatively assess that the protein on the NP surface is in fact immobilized in its active, native form, we decided to conjugate small antibody fragments to NPs and show their retained function in a simple aggregation-induced experiment ( Fig. S.2A). These fragments, usually just the active domains of the full proteins, are of <30 kDa in size ( Fig. 1A) and can suffer from relatively fast aggregation and denaturation 21 . Stripped of the glycosylated and other heavy parts of the typical full size antibody, the fragments tend to be immobilized onto surfaces in conformations that compromise their ability to specifically recognize their corresponding antigens. We speculate that the 3-dimensional shape, size and the flexibility of the G4.5 PAMAM dendrimer introduces a small degree of roughness on the NP surface. As a result, the antibody fragments are 'anchored' on the surface through fewer points, thus are able to retain their specific activity and binding capacity. In our experiment, the dendrimer-activated NPs (via sulfo-NHS-ester) were incubated with anti-CRP scFv to produce anti-CRP coated NPs. The colloidal stability of such antibody-conjugated NPs, and their antigen-reactivity were observed by DLS. Fig. S.2B-left shows that the anti-CRP-NPs were stable over the course of 3 hours of the experiment in the absence of CRP, although later measurement confirmed stability over several months. The 'active surface area' of the NPs (surface of the NPs covered by active anti-CRP) was qualitatively tested by measuring the evolution of particle size over time after addition of the CRP. Upon addition of a pentavalent antigen, C-reactive protein (CRP), aggregation of the particles was induced (Fig. S.2B-right). The resulting, timeevolving cluster size distribution increased exponentially with time (not shown), as expected for reaction-limited aggregation. To support this interpretation, the anti-CRP-NP sample showed negligible aggregation when random, non-specific protein, human chorionic gonadotropin (hCG) wad added into the sample instead of CRP ( Fig. S.2C).

Alternative multivalent crosslinkers: Dextran and BSA.
Despite the robustness of G4.5 PAMAM dendrimer in bioconjugation reactions, the cost of this well-defined macromolecule is considerable. Therefore, we have searched for alternative hyperbranched molecules that are readily available and inexpensive, but with reasonably predictable chemistries. Dextran and BSA were chosen as appropriate models. Dextran is a polysaccharide already used in medicine, with well-documented chemistry. It is neutral, water-soluble and possesses some favorable properties such as biocompatibility and biodegradability. The use of 40kDa dextran molecules to attach antibodies onto a flat surface of biosensor devices has been well described in the literature [22][23][24] . Jonsson et al presented an interesting approach to graft antibodies onto plastic substrates using dextran 'activated' by oxidation using by NaIO 4 25 . We have adapted this method for the silica NPs. Oxidation of dextran using NaIO 4 generates polyaldehyde dextran that can then act as a crosslinker with the primary amines of the NPs, and those of proteins, via Schiff base formation. The detailed optimization of this reaction including the effects of reaction variables on the reaction efficiency and colloidal stability of the NPs are summarized in the Supporting Information. BSA and other albumins are also widely studied supramolecules and are the most abundant proteins in plasma 26,27 . Albumins are characterized by high content of charged amino acids, aspartic and glutamic acids, lysine, and arginine 28 . Activation of the surface carboxylic acids with EDC/sulfo-NHS leads to rapid gelation of a BSA solution (not shown). Alternatively, additional steps employing glutaraldehyde were introduced to obtain the BSA-coated NPs, onto which antibodies could be immobilized. To compare the bioconjugation efficiency of the various multivalent linkers and the effect of the reaction conditions, we have grafted full size antibodies (Alexa Fluor 647 goat anti-human IgG) onto the surface of silica NPs as illustrated in Scheme 1. The number of antibodies immobilized on the NPs was quantified by using fluorescence measurements and described in Table 1 The incremental increase in the size of the NPs using PAMAM and Dextran method corresponded well to the size of the material that was being immobilized in each reaction step. PAMAM and dextran both have individual sizes of approximately 4-5 nm. The changes in the zeta potential (as an indicator of charge) of the NPs were as expected for both PAMAM and dextran-coated NPs. The zeta potential of the dextran-coated NPs changed to -22.5mV, which confirmed that such neutral hyperbranched glucan was firmly immobilized on the NP surface. The value rose to −32.5V after the reaction with antibodies, indicative of successful conjugation. Though the dextran-coated NPs exhibited a zeta potential similar to that of PAMAM-coated NPs activated by EDC/NHS, dextran acts as a steric barrier to the aggregation of adjacent NP and they ultimately remain stable in solution during the bioconjugation reaction. Importantly for both PAMAM and dextran, no visible or measurable aggregation of the NPs was detected during any of these steps. On the other hand, the BSA procedure showed signs of small degree of aggregation. The original size of the plain NPs was 106 nm; the BSA-coated NPs 'grew' to ～186 nm and the size of the final Ab-NPs for the BSA method reached nearly 228 nm. This undesired effect was attributed to the introduction of the additional steps, in which glutaraldehyde (monovalent, homo-functional crosslinker) was employed to produce the BSA-coated NP intermediate. We have reported previously on the negative effect of monovalent linkers in bioconjugation reactions 18 . To this end, all three methods yielded in Ab-NPs with different surface coverage of antibodies (Table 1). The robust and reliable conditions for the PAMAM dendrimer reaction resulted in the highest yield of antibodies per NP. The yield for the reaction using multivalent dextran, on the other hand, was the lowest. This is presumably because PAMAMcoated NPs are more negatively charged than dextran-coated NPs, thus the former NPs electrostatically attract more antibodies in solution (which in turn leads to higher probability of covalent bond formation). This is in agreement with our previously published data on the use of different generations of PAMAM dendrimer (antibody-coated surface area of NPs decreases with decreasing generation of PAMAM dendrimer). On the whole, PAMAM and dextran crosslinking strategies appear to be very robust in that they generate stable, monodisperse Ab-NPs. Although BSA method bound a relatively high number of antibodies per NP (78 per NP), the fact that the overall NP size was in excess of 200nm suggests that this method of grafting antibodies to NPs results in a small degree of NP aggregation. High antibody surface-coverage does not necessarily equate to high antibody activity. In the last section of this manuscript, the performance of Ab-NPs prepared by these methods will be evaluated in a simple fluorescencelinked immunoassay.

Evaluation of NP stability upon storage and long distance transportation through interactions studies with in vitro cell cultures.
Nanomaterials are usually prepared by expert material scientists, who often rely on biologists and clinicians to apply such prepared materials in many in vitro and/or in vivo experiments. It is therefore imperative to control the conditions of the nanoparticle storage so that the samples retain their unique properties and function over long period of time.
Crucially, the storage method should also enable simple means of transportation of the nanomaterial. The most convenient methods to preserve the nanomaterial and extend its shelf-life include storing the samples in a solution at either +4˚C or -20˚C or in a dried state by a method known as lyophilisation (or freeze-drying). Lyophilisation might be the preferred method because it reduces the cost and simplifies even long distance transportation. However, the freeze-drying process can add extensive strain on the nanomaterial, which often leads to particle aggregation. This phenomenon is particularly notable with antibody-coated nanoparticles. Before lyophilisation, it is therefore a common practice to suspend the NPs in formulations containing 'cryoprotectants', which are typically glycols or sugar-containing solutions. Common examples of the dessication-stress protectants that are widely used in the industry are trehalose, glucose, sucrose, fructose, dextran, mannitol and sorbitol. 29 The choice of the cryoprotectant will significantly influence the nanoparticle aggregation, which in turn presents important consequences on the interactions between the NPs and a biological material, such as cells. In the initial experiment, we have qualitatively investigated the effects of trehalose (2% (w/v)), a commonly used cryoprotectant, on the colloidal stability of silica NPs (plain FITC-doped and human IgG-coated) and their interactions with HeLa cells.
The results obtained using CLSM and SEM, shown on Fig. 4, provided very important information. Firstly, the optical properties of NPs were not affected by the storage procedure and due to their green fluorescence emission they were clearly visible under a CLSM (Fig 4A). NPs can be observed either internalized in the cytoplasm or associated to the cell surface (arrows in Fig 4A). Secondly, upon reconstitution in the cell culture medium and 60 minutes incubation with the cells, the IgG-coated NPs were observed to form aggregates at the micrometer scale and adhered to the surface of the cell (Fig. 4B and a magnified area containing NP aggregates is shown in Fig.  4C). Presumably, such large NP clusters were formed as a result of structural changes of the NP-bound proteins during the freeze-drying process, during the subsequent samplereconstitution method, imbalanced charge effects of the Ab-NPs in the high-salt concentration aqueous medium or a combination of these effects. By contrast, the unmodified NPs, freeze-dried from the same solution, appeared monodisperse, and due to their small size (~100 nm) were unable to be seen (Fig. 4d) at the same magnification, in which aggregates were observed. Moreover, these NPs were homogenously distributed and adhered over the cell surface (Fig. 4e). These simple, qualitative experiments critically highlighted the need for a simple screening protocol that would allow for the identification of the most reliable cryoprotectant and storage conditions, especially for protein coated-NPs. Without a storage method that does not influence the Ab-NP performance, it is unrealistic to envisage a successful translation of such nanomaterial into a biological setting.

The effect of cryoprotectants on the long-term storage and manipulation of Ab-NPs
In light of the cellular uptake experiments, we considered it absolutely necessary to prepare the Ab-NPs under conditions that allowed for long-term storage without any damage to the expensive antibodies on the NP surface. Equally important, the storage method must enable cost-effective transport of the nanomaterial and easy reconstitution of the Ab-NPs in the desired medium. The specific condition for the latter is that the end-user should be able to re-dissolve the Ab-NPs without excessive manipulation with the sample (e.g. ultrasonication, vortexing etc.).
In an ideal scenario, a solid formulation of the Ab-NP sample should be reconstituted in the desired solvent by simply shaking it by hand. Such reformed Ab-NP solution should maintain excellent colloidal stability for the minimum of 2 hours, which was arbitrarily chosen as sufficiently long working-window frame for the end-user to manipulate with the NPs (i.e. for sample dilution, partition, sterilization, etc.). We believe this to be a frequently overlooked facet of nanomaterial preparation and it is in fact of paramount importance. Addressing these storage issues means scientists could then confidently know that the re-dispersed material is as close to identical as its prestorage form as possible. Several useful articles review the freeze-drying process 17,30,31 . This simple method is based on removing water from a frozen sample by sublimation and desorption under vacuum. It is argued though that this process generates various stresses during freezing and drying steps. Therefore, cryoprotectants are usually added to the formulation to protect the nanoparticles from freezing and desiccation stresses. The most popular cryoprotectants encountered in the literature for freeze-drying nanoparticles are sugars: trehalose, sucrose, glucose and mannitol 17 . In the cell-interaction experiment, trehalose was identified as an inefficient protectant for the Ab-NPs system. We have therefore investigated the effect of another common stabilizer -sucrose and one unusual protectant -BSA on the colloidal stability of the reconstituted NPs and Ab-NPs. The experimental design is illustrated on Fig. S.5. Initially, considering the protocols published by industrial laboratories, plain NPs were freeze-dried in solutions containing either sucrose (20% w/v) or mixtures of BSA and sucrose (1-6% w/v BSA + 20% w/v sucrose). The samples were freeze-dried and stored for several days before they were reconstituted in 0.01 M PBS solution and also in a commonly used cell culture medium called Minimum Essential Medium (MEM). MEM is an enriched basal medium that enhances growth of sensitive or difficult-to-culture mammalian cells. The only user manipulation involved the addition of the desired medium to the solid-form nanoparticles and shaking the nanoparticles into solution by hand (see movie at: https://www.youtube.com/watch?v=HczrDJCaC38). The size of the NPs upon re-dispersion was measured by DLS, as was the evolution of the NP size over the course of 4 hours. The results, summarized on Fig. 5 show that the addition of 20% w/v of sucrose had a destabilizing effect on the colloid in both PBS and MEM. However, the addition of BSA had a stabilizing influence in both media (Fig. 5A left & centre, S5A, S5B). When BSA (1-6% w/v) was solely used as the cryoprotectant the colloid reconstituted in MEM retained its excellent stability for minimum of 2 hours (Fig. 5A right). 1% w/v BSA solution was identified as the optimal concentration to protect the solid formulation of the NPs. Addition of the MEM medium to the NP pellet containing 1% w/v BSA stabilized the NP solution for 4 hours (Fig. 5A right). In addition, a similar experiment was performed upon re-dissolving the NP pellet in Dulbecco's Modified Eagle's medium (DMEM). DMEM is an example of even more complex cell culture medium that contains approximately four times more vitamins and amino acids and two to four times more glucose than the MEM formula used. As seen on Fig. S.6B, the NP pellet containing 1% w/v BSA retained excellent colloidal stability upon re-dispersion for up to 2.5 hours. However, in order to demonstrate the real practicality of the freeze-drying approach employing cryoprotectants, we examined the colloidal stability of antibody-coated NPs. Firstly, goat anti-human IgG was conjugated to the NPs using the three multivalent-linkers strategies described in this study. All prepared Ab-NP samples were tested by DLS and showed well-defined sizes (Fig 3) before they were lyophilized from a solution containing 0, 1, 3 and 6% w/v BSA. After several days of storage, they were reconstituted both in water and PBS. PBS usually has a destabilizing effect on Ab-NPs colloidal stability due to the high content of NaCl (0.137 M), and also it is a good example of isotonic medium widely used both in immunodiagnostics and nanomedicine applications. Fig. 5B illustrates that yet again, 1% w/v BSA concentration as the cryoprotectant has stabilizing effect on the Ab-NP solution in both water and PBS. In the absence of the stabilizers (i.e. freeze dried in DI water), it was not possible to re-dissolve the Ab-NPs in aqueous solvent by shaking it by hand without high degree of aggregation (0% w/v BSA). Such samples would be practically unusable without ultrasonication. Overall, we demonstrated that it is possible to precisely control the freeze-drying of colloidal systems in order to reach a long shelf-life of the Ab-NPs and to empower the end-user to exploit the full potential of the nanomaterial without unnecessary manipulation of the samples that could potentially damage proteins bound to the surface of the NPs.

The function of Ab-NPs measured by immunoassay
High surface coverage of antibodies in Ab-NPs does not automatically translate to better performance of the Ab-NPs in applications such as biomedical sensing and drug delivery. Specific to these practices, one critical parameter that measures the function of the Ab-NPs is their signal-to-noise ratio (S/N). In diagnostics, this relates to the 'positive' signal generated by the specific Ab-NPs/antigen binding compared to the signal obtained in the absence of the antigen (i.e. non-specific binding). For targeted drug delivery and in vivo imaging, the S/N denotes the selective binding of the nanocarrier (e.g. Ab-NPs) to the desired cells/tissues with minimal collateral exposure to other cells. The function of the Ab-NPs prepared by the methods presented in this article was evaluated in simple fluorescence-linked immunoassay. The schematic of the assay format is illustrated on Fig. S.7. All Ab-NPs were freeze-dried from a solution containing 1% w/v BSA as the cryoprotectant. The S/N value for Ab-NPs is underpinned by the 'active surface area' (ASA), i.e. surface area of NPs covered by antibodies in their active, native state, available to bind to their target. We hypothesized that the higher the ASA, the higher is the S/N. The ASA for our Ab-NPs was modulated by the different multivalent likers and the underlying conjugation chemistries. However, in nanomedicine, the ASA can also be impeded by the protein adsorption layer (known as a "protein corona") that forms on the surface of colloidal nanoparticles in biological media such as serum or plasma 32 . The Dawson group has shown that this corona effect plays a key role at the nanoparticle-tissue/cell interface. A corona of adsorbed proteins on the nanoparticle surface is in fact what interacts with the biological machinery of cells rather than the carefully engineered particle surfaces created by materials scientists 33 . We have therefore probed the performance of Ab-NPs in immunoassays using PBS, 10% fetal bovine serum (FBS) and whole FBS as reaction media. As before, the solid formulations of Ab-NPs used in the immunoassays, were re-dispersed in these media by shaking by hand. To determine the specific vs non-specific binding phenomenon, goat anti-human IgG (hAb-NPs) and goat anti-rabbit IgG (rAb-NPs) antibody-coated NPs were prepared using the three multivalent linkers. Both hAb-NPs and rAb-NPs were incubated in 96-well plates modified with either human IgG or BSA respectively. Only the hAb-NPs were expected to bind to the human IgG antigen present in the wells. We have also compared the S/N ratios of these Ab-NPs to that of free dye-labelled goat anti-human and goat anti-rabbit IgG to gauge the reporting capacity of the FITC-NPs versus free dye. The fluorescence signal in each well was recorded using a fluorescence plate reader; the immunoassays results are displayed in Fig. 6. The results suggest that all of the Ab-NPs outperformed the free-antibody. Although this is in agreement with our previous published work 18 , such results were not so easily predicted for the assays performed in 10% and whole serum. What is evident though, is that the 'protein corona' effect has played a significant role in the performance of the Ab-NPs. While the specific signal (Fig. 6, red stripy bars) was relatively high in PBS, it significantly dropped when the buffer was being replaced by FBS (red striped bars in 10% and whole serum). We presume that a corona occurs on the Ab-NPs and therefore leads to a decrease in 'positive' binding. The effect of the different media played little role for Ab-NPs using PAMAM, which justifies the use of such linkers and the reliability of the method. Also, the PAMAM hAb-NPs showed the highest specific signal (red stripy bars) across all tested media. Interestingly though, the non-specific binding (grey bars) of PAMAM hAb-NPs was relatively high in PBS but improved when introduced to complex serum conditions: S/N of 5.0 in PBS compared to 5.8 and 6.1 in 10% and whole serum respectively. The S/N ratio of the Ab-NPs in PBS decreased in the following order: dextran>BSA>PAMAM. This was surprising finding for the dextran method. As presented earlier in this article, the Ab-NPs prepared by the dextran-mediated method yielded the lowest quantity of antibodies per one NP (Table 1). However, because of its low non-specific binding (grey bars), the dextran hAb-NPs showed the highest S/N ratio in PBS. This outcome is attributed to the combination of the following two parameters: 1. high active surface area -high % of active antibodies on the NP surface and 2. the consequences of dextran's hydration and swelling ability, which helps to suppress non-specific binding of molecules on its surface. The S/N ratio of dextran hAb-NPs dropped in serum but it was still comparable to the PAMAM nanoparticles, which is remarkable considering the differences in cost between the two hyperbranched linkers. On average, the BSA molecule proved to be the least effective crosslinker for the conjugation of antibodies onto the NP surface. Its S/N ratio was the lowest in serum, which was ascribed to the issues related to the low degree of aggregation specific to this sample. All things considered, we have identified an inexpensive alternative to the robust PAMAM-conjugation approach to immobilize antibodies on the NP surface. The use of readily available dextran proved to be straightforward with evident benefits for the nanoparticle's colloidal stability and its function

Conclusions
It can be argued that the surface chemistry strategies presented in this article are not universal and adaptable to every type of known nanomaterial. In this work however, we aimed to empower scientists with a robust and reliable 'recipe' or reference point for producing protein-coated nanomaterial that will enable the translation of the "on-bench potential" towards commercially attractive products or applications. We tried to clarify some currently used misconceptions and formulate some of the unaddressed challenges related to the active surface area of antibody-coated nanoparticles and the most appropriate means of preserving and handling them. We have investigated the optimal conditions for bioconjugation strategies to immobilize three different proteins on the surface of silica nanoparticles. We have probed the function of the proteincoated nanoparticles using mainly two analytical tools, usually readily available to material scientists: DLS and fluorescence spectrophotometry. We have shown that the physicochemical properties of the chosen crosslinker play a fundamental role in nanoparticle colloidal stability and antibody binding. In particular, we have shown that when activated using EDC/sulfo-NHS that PAMAM mediated immobilization method is very robust and reliable for both full size antibodies and also their single chain variable fragments. Significantly though, the dextran method represents a very good and cost effective alternative to PAMAM. It is quite clear that to translate the many formulations of nanomaterials, such as drug delivery systems or signal reporters into the clinical market, expertise from a multitude of disciplines including chemistry, materials science, engineering, biology, pharmaceutical sciences, and clinical research is required. Therefore, we have crucially established the most adequate protocols for storage and reconstitution of the nanomaterial with minimum end-user manipulation. To this end, freeze-drying as the optimal method and BSA (1% w/v) as the optimal example of cryoprotectant were identified to preserve the colloidal stability of our Ab-NPs in biologically relevant media. However, a number of challenges remain before such nanomaterials can be produced industrially; for example, would the conjugations reactions be as efficient if performed on a larger scale? We believe that the work presented in this article will help to advance the multidisciplinary efforts for the common objective of clinical benefits of nanotechnology. Aldrich. Absolute ethanol, 1-ethyl-3 (3dimethylaminopropyl) carbodiimide HCl [EDC], phosphate buffer saline tablets (one tablet dissolved in 200mL DI water yields 0.01M phosphate buffer, pH7.4) were purchased from Fisher Scientific. Here after, the use of 'PBS' refers to 0.01M PBS, pH7.4. N-hydroxysulfosuccinimide sodium salt (>98%) was purchased from TCI. AlexaFluor®647 labelled goat antihuman and goat anti-rabbit IgG (H+L, 2mg/mL), minimum essential medium were purchased from Life Technologies. Transparent Nunc Maxisorb 96 well plates used in the immunoassay were purchased from Fisher Scientific. Deionised water (< 18 MΩ) was obtained from a Milli-Q system from Millipore. The dye used in this work, fluorescein isothiocyanate isomer (95%) was purchased from ABCR.

GENERAL PROCEDURE FOR THE SYNTHESIS OF NANOPARTICLES
The dye-doped silica nanoparticles were synthesised using a quaternary reverse microemulsion method 34 . Briefly, FITC dye (2.5mg) and APTMS (5.6µL) were conjugated with stirring in a dried glass vial containing 2mL of 1-hexanol for 2hrs. After 2hrs, cyclohexane (7.5mL), 1-hexanol (1.133mL), Triton® X-100 (1.894g), DI water (0.48mL), dye conjugate solution (0.666mL) were added to a 30mL plastic bottle under constant stirring. 100µL TEOS was then added. After 30mins, 40µL of ammonium hydroxide solution was added to catalyse the nanoparticle synthesis reaction and was left to stir for 24hrs. Subsequently, TEOS (50µL) was added to the reaction mixture. After 30mins, THPMP (40µL) was added and, 5mins later, APTMS (10µL) was introduced. The solution was stirred for a further 24hrs. The microemulsion was broken through the addition of 30mL ethanol. The nanoparticles were then purified by centrifugation in ethanol (14,000rpm, 7mins, ×3). The centrifugation steps led to the formation of a NP pellet. The supernatant was removed and the pellet was re-suspended in fresh ethanol, ultrasonicated and then re-centrifuged. After purification the nanoparticles were stored in ethanol solution at 4°C.
CONJUGATION OF PROTEINS TO NPS USING G4.5 PAMAM DENDRIMER 1µmol (526µL) of PAMAM dendrimer was used to dissolve 73.6mg EDC. This solution was then added to 42.6mg of sulfo-NHS or 23.0mg NHS. 16µL of 1N HCl was added to the solution, and was topped up to 1mL using DI water. The sample was allowed to shake for 20-25mins at 600rpm using Stuart SA8 vortex mixer. The activated dendrimer solution was then added to 2mg of NP pellet. The NPs were redispersed into the dendrimer solution via ultrasonication. The sample was shaken for 15mins. After 15mins the sample was centrifuged (14,000rpm, 8mins). The supernatant contained unbound dendrimer and was discarded. The pellet contained dendrimercoated NPs and was redispersed (via ultrasonication) in 968.5µL MES buffer (pH4.2, 0.1M). 31.5µL of 2mg/mL AlexaFluor647 goat anti-human or goat anti-rabbit was added to the dendrimer-coated NP solution and allowed to shake for 4 hours at 600rpm. Goat anti-human and anti-rabbit IgG contained 4 moles of dye per antibody. The concentration of protein added was calculated using Bangs Lab Technical Note 205. The amount of proteins needed to achieve high nanoparticle surface coverage was determined from the formula: where S = amount of protein required to achieve surface saturation (mg protein/ g nanoparticle), ρ s is solid sphere density (2.4g/cm 3 for silica), d is the mean particle diameter (which, from TEM, was calculated to be 80nm) and C is the capacity of sphere surface for a given protein (mg protein/ m 2 of sphere surface), which is approximately 2.5mg/m 2 for IgG molecules and 3mg/m 2 for BSA. For purification, the sample was centrifuged for 8mins at 14000rpm, sonicated and redisssolved in 1mL PBS (x4).

CONJUGATION OF PROTEINS TO NPS USING DEXTRAN
500µL of 40mM of sodium periodate (in DI water) was added to 1µmol 40kDa Dextran in a 2mL centrifuge tube and shaken at 600rpm for 1.5hrs. The dextran solution was then added to 1mg of NP pellet, ultrasonicated and shaken for 1hr. 500µL of 40mM of sodium borohydride (in DMF) was added directly into the NP-dextran solution, which was then shaken for 25mins. The sample was then centrifuged (14000rpm, 8min, x2). The NP pellet was dispersed (via ultrasonication) in DI water after the first centrifugation. This procedure was shown to be most effective way of conjugating dextran to the NP surface ( Fig. S.3). After the second centrifugation, the pellet was redissolved in 500µL of 40mM of sodium periodate and shaken for 30mins (600rpm), and then centrifuged (14,000rpm, 8mins, x2). After the second centrifugation, 484µL 0.1M MES pH 4.2 was added to the pellet and sonicated. 16µL of AlexaFluor647 antibody was added and shaken for 4hrs (600rpm). 500µL of 40mM sodium borohydride (in DMF) was added directly into the tube and allowed shake for another 40mins. The sample was purified as per the PAMAM section.
CONJUGATION OF PROTEINS TO NPS USING BSA 1mL of 10% (w/v) glutaraldehyde in PBS was added to 2mg of NP pellet. The sample was ultrasonicated and shaken for 45mins at 800rpm. The sample was centrifuged and pellet was resuspended in 900µL PBS, followed by the addition of 100µL BSA solution (30mg/mL). The sample was allowed to shake for 2hrs at 800rpm. The sample was centrifuged (14000rpm, 5mins) and the pellet was redispersed in 1mL 10% glutaraldehyde in 0.01M PBS pH7.4 and allowed to shake for 75mins (800rpm). The sample was centrifuged (14000rpm, 5mins) and the pellet was redispersed in 968µL 0.01M PBS, pH7.4. 32µL of AlexaFluor647 antibody (2mg/mL) was added to the sample and allowed to shake for 2hrs. After 2hrs, the sample was centrifuged (14000rpm, 5mins) and the pellet was redispersed in 1mL of 1% BSA, 40mM aminohexanoic acid in 0.01M PBS pH7.4 for 15 mins. The sample was purified as per the PAMAM section. BSA was used as it exhibits low non-specific binding, thus serving as a background reference since low number of Ab-NPs would bind to this surface. The plate was then cleaned by washing the wells with 120µL 0.2% (v/v) Tween in PBS and then PBS (x2). All wells were then blocked with 120µL of 1% (w/v) BSA solution in PBS and incubated at 37°C for 2hrs. The same well cleaning procedure was used as before. Solid form Ab-NPs (goat anti-human or anti-rabbit IgG coated nanoparticles, freeze dried in 1% BSA) were redispersed in either 0.01M PBS pH7.4, 10% FBS in PBS or whole FBS such that their concentration was 0.4mg/mL. 70µL of Ab-NP sample was added to both types of well (human IgG coated and BSA coated). 70µL of free AlexaFluor647-labelled goat anti-human and goat anti-rabbit IgG was equal to the number Ab-NPs introduced per well. The plate was then incubated overnight at 4°C. The wells were cleaned as before and dried with a stream of nitrogen. 60µL of PBS was added to the wells and fluorescence readings for FITC-doped NPs and free antibody (AlexaFluor®647) were carried to investigate their respective signal-to-noise ratio.

CELL CULTURE MAINTENANCE
Human Epithelial Cervical Adenocarcinoma Cells (HeLa, ATCC) were cultured in 1X Minimal Essential Medium (MEM, Gibco by Life Technologies) supplemented with 10% w/v Fetal Bovine Serum (FBS, Biowest) and 2 mM L-Glutamine (L-Glut, Life Technologies), subsequently referred to as supplemented MEM. Cell culture maintenance was done in T25 or T75 flasks (Thermo Scientific) and cultures were kept in an incubator (Hera Cell 150, Heraeus) at standard conditions (37ºC and humidified atmosphere of 5% CO2) during incubation and postincubation times.

CELL CULTURE FOR SPECIFIC ANALYSIS
To carry out analysis through Scanning Electron Microscopy (SEM), round glass coverslips (Knittel Glass) were placed into 4-well plates (Nunc, Thermo Scientific) in order to make cells grow on the coverslips. To perform Confocal Laser Scanning (CLSM) studies, cells were directly cultured in 35 mm diameter glass bottom petri dishes (MatTek Corporation). In both cases, a solution containing supplemented MEM with 1.5 x 10 4 cells was added in each well or petri dish and then cultures were placed in an incubator at standard conditions. Assays were carried out after complete cell adhesion and once cultures had reached exponential growth, normally one day following cell seeding.

STUDY OF THE INTERACTIONS OF NPS WITH HELA CELLS AND ANALYSIS BY CLSM
Plain, modified NPs and human IgG-coated NPs were freeze dried in 100µL aliquots (2mg/mL) in a 2% (w/v) trehalose solution. The IgG-coated NP were prepared similarly to the AlexaFluor 647-labelled anti-human/rabbit IgG coated NPs described in the 'Conjugation of proteins to NPs using G4.5 PAMAM dendrimer' methodology. A 50 µg/mL solution of Silica NPs prepared in supplemented MEM (1 mL final volume) was added into each petri dish and incubated for 60 min. After incubation, NP solution was removed from each petri dish and 1 mL of fresh supplemented MEM was added. To check the interaction of the NPs (green fluorescence emission) with cells, they were labelled with CellMaskTM Deep Red plasma membrane Stain (Life Technologies, far-red fluorescence emission) to set the cell limits and Hoechst to stain the nucleus (Life Technologies, blue fluorescence emission). Images in random areas of the cultures were acquired with a Leica SP5 Confocal Laser Scanning Microscope and interaction of NPs with cell cultures was analyzed by means of localization within the cells.

STUDY OF THE INTERACTIONS OF NPS WITH HELA CELLS AND ANALYSIS BY SEM
A 50 µg/mL solution of Silica NPs prepared in supplemented MEM was added into each well and let incubate for 60 min. After incubation, the NP solution was removed from each well and 0.1 M Cacodylate buffer (TAAB) was added. After 2 min, TAAB was removed and 2.5% Glutaraldehyde (Merck) was added and incubated for 45 min in order to carry out the fixation of the cells. The fixative was removed and cells were dehydrated through series of increasingly graded ethanol (Merck Millipore), in which each step (50%, 70% and 90% once, and 100% twice) lasted 8 min. Finally, cells were incubated with Hexamethyldisilazane (Electron Microscopy Sciences) to completely dry the samples. Finally, coverslips were mounted onto SEM stubs and analyzed using a FEI Magellan 400L SEM.