Reversible Control of Nanoparticle Functionalization and Physicochemical Properties by Dynamic Covalent Exchange

Abstract Existing methods for the covalent functionalization of nanoparticles rely on kinetically controlled reactions, and largely lack the sophistication of the preeminent oligonucleotide‐based noncovalent strategies. Here we report the application of dynamic covalent chemistry for the reversible modification of nanoparticle (NP) surface functionality, combining the benefits of non‐biomolecular covalent chemistry with the favorable features of equilibrium processes. A homogeneous monolayer of nanoparticle‐bound hydrazones can undergo quantitative dynamic covalent exchange. The pseudomolecular nature of the NP system allows for the in situ characterization of surface‐bound species, and real‐time tracking of the exchange reactions. Furthermore, dynamic covalent exchange offers a simple approach for reversibly switching—and subtly tuning—NP properties such as solvophilicity.


Reversible Control of Nanoparticle Functionalization and PhysicochemicalProperties by Dynamic Covalent Exchange**
Flavio della Sala and Euan R. Kay* Abstract: Existing methods for the covalent functionalization of nanoparticles rely on kinetically controlled reactions,a nd largely lackt he sophistication of the preeminent oligonucleotide-based noncovalent strategies.H ere we report the application of dynamic covalent chemistry for the reversible modification of nanoparticle (NP) surface functionality,c ombining the benefits of non-biomolecular covalent chemistry with the favorable features of equilibrium processes.Ahomogeneous monolayer of nanoparticle-bound hydrazones can undergo quantitative dynamic covalent exchange.The pseudomolecular nature of the NP system allows for the in situ characterization of surface-bound species,a nd real-time tracking of the exchange reactions.F urthermore,d ynamic covalent exchange offers as imple approach for reversibly switching-and subtly tuning-NP properties such as solvophilicity.
Despite tremendous advances in the preparation of nanoparticles (NPs) from arange of materials, [1] manipulation and characterization of NP surface functionality remains acrucial challenge in the quest to exploit the often remarkable properties observed within this newfound region of chemical space.D irect incorporation of surface-bound functional molecules during NP synthesis is intrinsically restrictive, demanding compatibility with the synthesis conditions.P ostsynthetic substitution of temporary surface species in a" ligand exchange" process can facilitate the introduction of aw ider range of surface-bound functionalities,i ndependent of the NP synthesis methods. [2] Yet, such processes are often irreversible,i nefficient, and can lead to NP surface reconstruction or etching.
Ag eneralizable synthetic approach whereby as et of NP "building blocks" may be predictably functionalized, manipulated, and assembled is therefore highly desirable.T he potential of such ac oncept is well exemplified by oligonucleotide-functionalized nanomaterials. [3] Yet, biomolecular methods only operate within tightly defined conditions and offer limited scope for customization. On the other hand, nonbiomolecular strategies will allow the full gamut of synthetic chemistry to be exploited in the optimization of nanomaterial structure,f unction, and properties.I nnovative designs exploiting noncovalent interactions for NP functionalization, [4] aggregation, [5] and surface immobilization [6] have recently been explored, but these non-biomolecular systems cannot yet match the stability,s pecificity,a nd selectivity of oligonucleotide hybridization. Postsynthetic covalent modification of NP-bound monolayers is an attractive alternative, but traditionally this strategy has relied on kinetically controlled reactions, [7] which at best produce statistical mixtures of products and only offer one-shot opportunities for functionalization. Thermodynamically controlled covalent bond-forming reactions instead combine the error-correcting and stimuli-responsive features of equilibrium processes with the stability and structural diversity of covalent chemistry. [8] Thefirst examples of dynamic covalent exchange taking place on 2D surface-confined molecular monolayers, [9a-c] or at the surface of self-assembled phospholipid bilayers, [9d,e] have recently been reported. We now show that such reactions may also be successfully performed on 3D NP-bound monolayers.W epresent prototypical "dynamic covalent NP building blocks": gold nanoparticles (AuNPs) bearing ah omogeneous monolayer of N-aroylhydrazones (Figure 1), through which reversible control of NP functionalization and properties can be achieved. Hydrazones display stability under aw ide range of conditions,y et undergo covalent exchange reactions in the presence of acid or nucleophile catalysts, [10] making them particularly useful for creating dynamic covalent systems with good differentiation between kinetically labile and locked states. [11] This combination of behaviors likewise appeared ideal for ar obust but exchangeable linkage for the construction of dynamic covalent AuNPs. [12] Ligand 1 bears an Naroyl hydrazone unit, connected through an aliphatic linker to at hiolate functionality for binding to AuNP surfaces (Figure 2a). [13] Thea lkyl linker encourages the formation of awell-ordered surface monolayer,maximizing van der Waals interactions between neighboring chains, [2a] whereas the outer tetraethylene glycol unit confers compatibility with polar solvents and conformational flexibility at the dynamic covalent reactive site. [14] Gold nanoparticles bearing ahomogeneous monolayer of 1 were prepared in aone-step,single-phase process, [15] which consistently yielded NPs of mean diameters in the range of 2.8-3.4 nm, with dispersities < 20 % (Figures 2d and S6), and exhibiting aw ell-defined surface plasmon resonance (SPR, Figures 2e and S6). Thea bsence of surfactants or temporary ligands facilitated the preparation of single-component monolayers,w hile all unbound contaminants could be removed by NP precipitation and washing. Verification of both comprehensive purification, and the structural integrity of NP-bound 1, [16] were essential for being able to unambiguously characterize the surface-confined dynamic covalent processes.A uNP-1 displays characteristically broad 1 Ha nd 19 FNMR spectra (Figure 2b and c) consisting only of the resonances expected for asingle-component monolayer of 1. Thea bsence of nonsolvent unbound contaminants was confirmed by T 2 -filtered 1 HNMR spectroscopy using the recently developed CPMG-z pulse sequence (Figures 2b,  bottom, and S3). [17] Corroboration of the surface-bound molecular structure was provided by LDI-MS,w hereby all major ions could be assigned as originating from desorbed 1 (Figures 2fand S4). [18] Finally,only products consistent with ah omogeneous monolayer of 1 were detected after iodineinduced oxidative ligand stripping from AuNP-1 ( Figure S5).
Directly tracking reactions that occur on molecules confined to non-uniform faceted surfaces,within ah eterogeneous population of NPs,p resents several challenges.I nherently low concentrations,f ast transverse relaxation, and chemical shift heterogeneity combine to yield broad, weak 1 HNMR spectra for NP-bound molecules (Figure 2b, middle), making quantitative deconvolution of resonances from structurally similar species extremely challenging. [19] Incorporating fluorine labels allowed us to exploit the significant chemical shift dispersion and excellent sensitivity of 19 FNMR spectroscopy to interrogate the composition of hydrazone-bound monolayers before and after dynamic covalent exchange reactions ( Figure 3).
As table colloidal suspension of AuNP-1 in 10 %D 2 O/ [D 7 ]DMF was treated with an excess of p-(trifluoromethyl)benzaldehyde (5)a nd CF 3 CO 2 H. [20] After 16 ha t5 0 8 8 C, 19 FNMR spectroscopy showed that the signal for NP-bound p-fluorobenzylidene hydrazone 1 had decreased in intensity and two new resonances had appeared:one corresponding to free p-fluorobenzaldehyde (6), and another corresponding to NP-bound p-(trifluoromethyl)benzylidene hydrazone 2 (see Figure S10 for full sweep width crude and purified spectra). Unbound molecular species (released 6,e xcess 5,a nd CF 3 CO 2 H) were removed by NP precipitation and washing with nonsolvents,y ielding aN Ps ample with am ixed monolayer comprising 90 %h ydrazone 2 and 10 %h ydrazone 1 (AuNP-1 0.1 2 0.9 ,F igure 3b). By subjecting this sample again to the same exchange conditions,f ollowed by purification as before,y ielded ap ure sample of AuNP-2 (Figure 3b). A homogeneous monolayer of 2 was confirmed by 19 Fa nd 1 HNMR spectroscopy (Figures 3b,m iddle,S 9, and S11), LDI-MS (Figures 3c,m iddle,a nd S12), and oxidative ligand stripping ( Figure S13). [21] Thed ynamic covalent exchange process is entirely reversible.T reatment of AuNP-2 with 6,u nder identical exchange conditions to before,produced asample displaying am ixed monolayer of the two hydrazones in the ratio 1:1 (AuNP-1 0.5 2 0.5 , Figure 3b). Subjecting this sample to afurther  (Figure 3c,b ottom) indicate the intimate mixing of hydrazones on the NP surface, [18] whereas the lower extent of exchange in this reverse process is in line with the greater stability expected for the p-(trifluoromethyl)benzylidene hydrazone. [10c] Importantly,t hese mixed hydrazone samples allowed us to confirm that quantification of the monolayer composition by integrating the broad NP-bound 19 FNMR signals was consistent with the results of iodineinduced oxidative ligand stripping and subsequent analysis of the released molecular species ( Figure S17).
Thea bility to quantify both NP-bound and unbound species using 19 FNMR spectroscopy allowed us to track hydrazone exchange in real time and explore the effects of surface confinement on reactivity.T he concentrations of all four fluorinated species (AuNP-1,A uNP-2, 5, 6)w ere monitored during the exchange of AuNP-1 with aldehyde 5. [22] Comparing the resulting kinetic profile to that of afreely dissolved model compound under the same conditions (Figure S20) indicates aclear kinetic inhibition for the NP-bound reaction. Fitting to derive apparent rate constants (Table S1), [22] counterintuitively revealed the inhibitory effect to be stronger in one reaction direction (k NP /k MOL -(F!CF 3 ) = 0.2) than the other (k NP /k MOL (CF 3 !F) = 0.5), corresponding to an equilibrium endpoint that favors AuNP-1 more strongly than predicted by the model reaction in bulk solution. Slower kinetics for the NP-bound process might be expected on the basis of simple steric arguments. However,itisunclear whether the very small increase in size on converting 1 to 2 can explain the differential effect on the exchange rates,o rw hether other intra-monolayer interactions or local concentration effects are also at play.
Mild and reversible methods for postsynthetic NP modification are highly desirable and would have significant benefits for nanomaterial property control, handling,a nd processability.F or example,t uning solvent compatibility is often required to match an optimized NP synthesis route with as pecific end application, [23] yet existing methods involve either encasing an anoconstruct in ap olymeric modifier, encapsulation in micelles,orcompletely replacing the surface ligands.T he latter strategy may be considered as ad ynamic exchange of the Au À Sbond. [2] However, completely replacing the stabilizing monolayer is ar elatively harsh and slow process.W hereas hydrazone exchange at 50 8 8C( as described in Figure 3) reaches 90 %e xchange within 24 h, the ligand exchange of AuNP-1 with disulfide 2 2 takes several days to reach an endpoint exhibiting far lower conversion (63 %) under analogous conditions ( Figure S21), and does not proceed at all at ambient temperatures. [24] By exchanging only simple units on the periphery of the stabilizing monolayer, dynamic covalent exchange occurs rapidly under mild conditions;i tf urthermore avoids the necessity for multistep synthesis of several thiol-containing ligands,o ffers simple purification of the modified NPs from the molecular exchange species,and is entirely reversible. [24] To demonstrate the potential of dynamic covalent exchange for reversible property control, we sought to introduce simple aldehyde exchange units,c hosen to confer different solvophilic characteristics on our dynamic covalent AuNP building blocks (Figure 4). AuNP-1 functionalized with p-fluorobenzylidene hydrazone showed good colloidal stability only in polar aprotic solvents such as DMF and DMSO (Figure 4, top). Tr eating AuNP-1 with an excess of hydrophobic aldehyde 7 and CF 3 CO 2 Hi n1 0% D 2 O/ [D 7 ]DMF at 50 8 8Cr esulted in complete precipitation of the NPs within 1.5 h. Thes olid was easily recovered by centrifugation, and then purified from all molecular species by redispersion in methanol followed by precipitation with hexane.T he resulting residue exhibited markedly different solubility properties to AuNP-1 and could be readily reredispersed in organic solvents of intermediate polarity,such as chloroform or tetrahydrofuran (AuNP-3,F igure 4, left). Analysis of the reaction supernatant by 19 FNMR spectroscopy indicated > 95 %conversion of the starting NP-bound pfluorobenzylidene hydrazones ( Figure S23). On the other hand, 1 HNMR and LDI-MS analysis of the new NP sample ( Figures S22 and S24) were consistent with the expected malkoxybenzylidene hydrazone.C learly,d ynamic covalent exchange of NP-bound hydrazones occurred to give AuNP-3 and the consequent marked change in solvent compatibility.
In as imilar manner, AuNP-3 could be converted to AuNP-4,w hich showed excellent colloidal stability in water (Figure 4, right). o-Sulfonylbenzylidene hydrazone was confirmed as the major constituent of the NP-bound monolayer by ac ombination of 1 HNMR spectroscopy and LDI-MS ( Figures S28-S30). Each of these exchange reactions proved to be entirely reversible,s uch that any of the three AuNP systems,e xhibiting markedly different solvophilicity properties,c ould be accessed from either one of the other two by treatment with the appropriate aldehyde exchange unit (Figure 4a nd Scheme S2). Interestingly,d uring the conversion of AuNP-3 to AuNP-1,asample was obtained exhibiting solubility properties that were intermediate between the two extremes (Scheme S2). That this state arises from am ixed monolayer of hydrazones 3 and 1 was confirmed by LDI-MS analysis,which presented ion fragments originating from both possible hydrazones in roughly equal intensities ( Figure S27). Subjecting this material to af urther round of exchange with aldehyde 6 then yielded asample displaying indistinguishable physical and chemical properties to AuNP-1 produced by all other routes.T hus,i ti sp ossible to access ac ontinuum of AuNP solvophilicity characteristics across aremarkably wide range by fine-tuning the monolayer composition through the appropriate choice of exchange conditions.
Controlling the molecular details of NP surface functionality will be critical for realizing the full technological potential of nanomaterials.D ynamic covalent NP building blocks now offer ag eneralizable strategy for achieving this, using simple molecular designs and mild processes that are independent of the underlying NP material. Thea bility to reversibly tune surface functionality raises the prospect of smart NP-based devices with environment-responsive properties,o rr econfigurable self-assembly capabilities.T he pseu-domolecular nature of 3D NP-bound monolayers allows the direct characterization of surface-bound chemical processes, offering fundamental insights into the influence of crowded environments on reactivity,w hich have not been so readily accessible from analogous 2D surface-bound systems. [8b] Determining the complex influence of nanoscale features, such as surface curvature and monolayer composition, on reactivity is the next step that can now be addressed in the development of dynamic covalent NP building blocks to become flexible and versatile nanomaterial synthons. [25] Keywords: dynamic covalent chemistry ·gold nanoparticles · hydrazones ·s upramolecular chemistry