Anisotropic functionalization of upconversion nanoparticles

Ligand competition directs heterogeneous bio-chemistry surface and self-assembly for upconversion nanoparticles.

Doped by rare-earth ions, hexagonal-phase (b) NaYF 4 upconversion nanoparticles (UCNPs) are a new generation of nanomaterials featuring step-wise photon anti-Stokes emission: pumped by near infrared laser to emit visible uorescence, 1 tunable lifetime values from tens of microseconds to several hundred microseconds 2 and low toxicity in biological systems. 3 Owing to such exceptional optical properties, a diversity of applications have been realized using UCNPs, for instance, background-free biomolecular detection, [4][5][6] in vivo bio-imaging, 7,8 forensic applications, [9][10][11] anti-counterfeiting applications, [12][13][14] super resolution imaging, 15,16 and nanoscale thermometry. 17,18 Towards bio-related applications, the key is to functionalize the surface of UCNPs by specic biomolecules modication and transfer them from organic solvent to aqueous phase. Up to now, a variety of strategies have been devoted to modifying the surface of UCNPs, including amphiphilic polymer interaction, 7,19 silica coating, [20][21][22] surface ligands oxidization 23 and ligand exchange; 24,25 nevertheless, facet-selective functionalisation of upconversion nanoparticles has seldom been reported.
DNA appears to be one of the most popular biomolecules for surface functionalisation of nanoparticles, due to its commercial availability, low cost, excellent stability and specicity that allows direct recognition of complementary sequences. In 2005, Costa and co-workers reported that the backbone of DNA molecules can bind to the lanthanide ions, 26 which suggests a new way to directly conjugate DNA onto UCNPs. Based on this nding, researchers have developed a one-step conjugation technique to attach DNA onto the surface of UCNPs; 24,25 nevertheless, such a method still treats UCNPs as spherical nanoparticles and overlooks the fact that UCNPs are hexagonal prism nanoparticles with two (001) facets on the tips and six (100)/ (010) facets on the lateral surface, and these facets have different charge distribution and are capped by different ligand molecules. 27 Hence, we hypothesize that the binding strength of DNA could be varied on the (001) and (100)/(010) facets which may lead to selective molecule binding on the different facets of UCNPs. If this is true, in-depth understanding and proper control of anisotropic surface properties will lead to a new scope for bio-/nano-interface chemistry and applications.
In this paper, we utilize DNA to investigate the facet-selective binding to the surface of UCNPs. We nd that the binding affinity of phosphodiester bonds on the backbone of DNA is stronger than oleic acid (OAH) on (001) facets but weaker than oleate anions (OA À ) on (100)/(010) facets, resulting in selective binding to the two ends of UCNPs; whereas the phosphate group on the end of DNA shows the strongest affinity to replace all the surfactant molecules on UCNPs which creates hydrophilic surface. The location of DNA molecules is experimentally conrmed by analytical chemistry methods and directly visualized by the stochastic optical reconstruction microscopy (STORM). The facet-selective functionalization of UCNPs not only provides insights into the understanding of bio-/nano-interface reaction but also has potential application in selfassembly of structures of nanoparticle building blocks.

DNA ligand exchange and quantication
To amplify the anisotropic surface properties of UCNPs, instead of using small nanoparticles with small aspect ratios, we choose nanorods of $170 nm in length and $35 nm in diameter for the study. The ligand exchange reaction is taken place by mixing the rods suspended in chloroform and DNA aqueous solution followed by gentle shaking for 3 hours. Fig. 1a illustrates the ligand competition process with two kinds of hydrophilic molecules: single strand DNA molecules with and without a phosphate group on the 5 0 terminus. The rods are transferred from chloroform into the upper aqueous phase by replacing the surfactants, i.e., oleic acid molecules (OAH) and oleate anions (OA À ) on the (001) and (100)/(010) facets of the particles, respectively.
We compute the binding energies of four different chelating moieties, e.g. oleic acid, oleate anion, phosphate group and phosphodiester bond, onto the two kinds of facets of the hexagonal prism-like UCNPs based on density functional theory (DFT) simulation (Fig. 1b, table). The binding strength of phosphate groups to the surface of UCNPs is remarkably stronger than that of the surfactant molecules, which results in the fact that the phosphorylated DNA can replace the initial surfactant molecules on all the facets of the particles; whereas phosphodiester bonds is not strong enough to compete with OA À on (100)/(010) facets thus only replace the OAH molecules on the (001) facets.
We propose three scenarios of how the different ligands compete to attach to the anisotropic surfaces of UCNPs ( Fig. 1c-e), which are experimentally veried from the locations of the UCNPs aer the ligand competition and exchange process. Aer completing reaction with DNA ligands without the 5 0 -terminus modied with phosphate groups, it is shown that UCNPs featuring a mixture of hydrophilic and hydrophobic surface properties are suspended between chloroform (oil) and water (Fig. 1d). In contrast, by using DNA ligands with 5 0 terminus modied with phosphate moiety groups that display the strongest binding to both (100)/(010) and (001) facets, upconversion nanorods are completely pulled into the aqueous suspension (Fig. 1e). Due to the varied surface quenching effects on the green and red emission bands of UCNPs, 28,29 the colour of UCNPs was slightly changed aer being transferred from the organic phase to the aqueous phase ( Fig. S3 †). The degree of anisotropic surface properties can be ne-tuned by decreasing the pH value of DNA solution, which induces more DNA molecules to be bonded onto the side surfaces (100)/(010) facets of the UCNPs, see ESI Section 3. †

Determine the location of DNA molecules on UCNPs
To quantitatively evaluate the selective binding of single strand synthetic DNA ligands, we design and synthesize two types of nanorods with different length ($70 and $135 nm, TEM images shown in Fig. 2b), and conduct a set of comparison experiments using the same weight to ensure the same volume, but different areas of (001) facets. In this way, the area of (001) facets of 70 nm nanorods is twice that of the 135 nm ones (illustrated in Fig. 2a). We prepare DNA solution at pH 7 to rule out the inuence of hydrino (H + ) or hydroxyl (OH À ) in the amounts of DNA conjugated to the particles. Aer ligand exchange, the amount of DNA  is quantied by checking the absorbance intensity at 260 nm (see ESI Fig. S2f †). Fig. 2c shows that the amount of DNA on the 135 nm nanorods is about half that on the 70 nm nanorods, indicating that DNA mainly replaces OAH on the (001) facets.
In 2016, Su et al. employed a super-resolution localization and defocused imaging approach to locate the uorescent dye molecules on the tips of gold nanorods. 30 To visualize the selective binding of DNA molecules on the end (001) facets of the UCNP nanorods, we conduct stochastic optical reconstruction microscopy (STORM) to resolve the location of the ATTO-550 labelled DNA molecules conjugated on the UCNP nanorods (170 nm in length, as TEM image shown in ESI Fig. S6 †). Consistent to the TEM measurement, the distance between a typical pair of ATTO-550 single molecule clusters is determined to be 176 nm by STORM (Fig. 3b). The Gaussian t to the histogram distribution of the distance of the pairs of dye clusters reveals a mean value of 170 nm (Fig. 3a), clearly indicating the locations of DNA molecules mainly on the end (001) facets. In contrast, the isotropically modied nanorods display multiple and random uorescent clusters (Fig. 3d) but with a broader paired distance distribution of ATTO-550 dyes (Fig. 3c). As the controls to conrm that the uorescent signals on the end of nanorods indeed come from ATTO-550 single molecule uorophores, the as-synthesized nanorods and non-uorophore DNA modied nanorods display no uorescence (see ESI Section 4 †).

Investigate the activity of single strand DNAs on the nanocrystals
We employ a hairpin structure DNA 31 to probe the affinity and activity of single strand DNA on the nanocrystal surface, as shown in Fig. 4. To investigate the anisotropic surface properties for UCNPs, two types of nanocrystals of different aspect ratios ($55 nm long Â $30 nm in diameter, and $50 nm long Â $80 nm in diameter) are used. Fig. 4a schematically shows that at pH 5.5, both DNA molecules, with and without phosphate groups on the 5 0 terminus, can bind to (001) and (100)/(010) facets of UCNPs. When adding the hairpin probe into the system, phosphorylated DNA molecules strongly bind to the surface of UCNPs, thus no obvious uorescence singles can be detected either on the nanoparticles sample or the supernatant.
Nevertheless, DNA molecules without phosphate groups can be physically inserted into the hydrophobic OA À surfactants on side surfaces at low pH, although unstable. They can be further released to trigger the hairpin DNA probe to uoresce. This is veried by the supernatant of rod-shape nanocrystals showing much stronger uorescent signals than the supernatant of plate-shape nanocrystals. It explains our earlier observation (see ESI Fig. S2b and c †) that low pH values would increase absorption of DNA molecules onto the nanocrystal surfaces, and it is caused by weak physical absorption on the (100)/(010) facets. Fig. 4 Investigation of the bio activity of single strand DNA on the nanocrystal surface. (a) Illustration of the mechanism: DNA molecules can only replace the OAH on (001) facets of UCNPs but only insert into the OA À molecules on (100)/(010) facets at low pH value; while phosphorylated DNA molecules can replace both OAH and OA À on the surface of UCNPs. It is difficult for the DNA/phosphorylated DNA molecules to hybridize with the hairpin probe on the surface of UCNPs, but inserted DNA molecules would be replaced by hairpin probes and then hybridize with the rest hairpin probes in the supernatant to recover the fluorescent. (b) and (c) are the rod-shape and plate-shape UCNPs used in the experiment. (d) The DNA concentration on the rod-shape UCNPs and plate-shape UCNPs is calibrated to be the same. Larger (100)/(010) area of the rod-shape sample releases more DNA molecules to recover higher fluorescent signals in the supernatant than the plate-shape UCNPs. (e) Phosphorylated DNA molecules bind strongly on both (001) and (100)/(010) facets so very low fluorescent signals are observed for both of the samples. Scale bar: 100 nm.

Anisotropic functionalization directed self-assembly of UCNPs
Different self-assembly formats of nanorods can be achieved by controlling the concentration of molecules on the surface. 32 In this work, the successful control in facet selective functionalization of DNA molecules, either on the (001) facet of UCNPs or on all surfaces of UCNPs, can result in UCNPs with either anisotropic or hydrophilic surface properties. By dispersing the above two kinds of UCNP nanorods (2.5 mg mL À1 ) in water and preparing them onto the copper grid, only aer 5 minutes, two self-assembly patterns, side-by-side (Fig. 5a-d) and end-to-end ( Fig. 5e-h), can be formed with the efficiencies of 100% and 53.8% respectively. We ascribe these distinct self-assembly behaviours to the theorem of achieving minimum surface energy. DNA molecules are negatively charged because of the existence of phosphodiester bond on the backbone. When the oleic acid molecules of the side facets of UCNPs are not exchanged by the DNA, the side surfaces are inherently hydrophobic. The UCNP nanorods prefer to assemble in a side-byside manner owing to the mutual attraction between the hydrophobic facets. In contrast, the phosphorylated DNA modied UCNPs have hydrophilic surfaces. The large area of side facets with negatively charged DNAs provides stronger electrostatic repulsion that tends to keep each nanorod away from each other, and the ends with lower energy tend to connect each other forming the end-to-end pattern.

Conclusion
The key to equipping inorganic nanocrystals with reliable and versatile biomolecular functions lies in the degree of biochemistry control at the bio-/nano-interface, which ultimately determines their stability, specicity, selectivity, and biocompatibility. This work suggests a new dimension in surface modication and functionalization of UCNPs to either have isotropic surface groups or anisotropic surface properties by applying facile DNA ligand exchange method. For the rst time, we have shown that we are able to tailor the binding of surface capping ligands based on the facet specic properties of UCNPs, which has been supported by analytical chemistry experiments and super resolution imaging. Our results open a new avenue of selective biomolecule functionalization for nanoscale surface biochemistry, which is beyond the size and morphology controls of nanocrystals. Furthermore, controlled self-assembly of UCNPs enabled by tailored DNA chemistry suggest the promise of using UCNPs as building blocks to construct more sophisticated functionalized nanostructures.

Synthesis of nanocrystals
The nanocrystals were synthesized according to our previously reported method. 27 Full method regarding the synthesis of the nanocrystals of multiple morphologies are given in ESI. † Briey, NaYF 4 :Yb,Er nanocrystals were synthesized by thermal solvent method. By tuning the ratio of chemicals, we obtained spherelike nanocrystals and nanoplates. The nanorods were synthesized by over-growth onto the sphere-like nanocrystals.

DNA functionalization of nanocrystals by ligand exchange method
Typically, 50 mL of 10 mg mL À1 UCNPs cyclohexane suspension was mixed with 400 mL chloroform in a small glass vial. Aer that 300 mL of 5 mM DNA solution with certain pH values were added to the vial. The UCNPs chloroform suspension and DNA water solution would form two phases. Aer incubation at 600 rpm on a vortex machine for 3 hours, the UCNPs transferred from chloroform to water phase. It should be noticed that aer reaction most of the UCNPs would stay in the interface of water and chloroform if the pH value is high, so all the liquid in the water phase and interface were taken out to centrifuge and the participated nanoparticles were puried by ethanol rst to remove the organic solvent and then water. The products were nally suspended in 200 mL distilled water.

STORM set up, imaging and data analysis
The Stochastic Optical Reconstruction Microscopy (STORM) imaging of UCNPs was carried out with Olympus cellTIRF-4Line system (Olympus IX83 motorized inverted microscope; UPlan-SApo TIRF 100 Â 1.40 oil; Photometrics EMCCD 512 Â 512; CellSens Soware; HP Z840 Work Station). Aer conjugated with ATTO-550 labelled DNA molecules, the UCNPs water suspension was diluted for 1000 times (2.5 Â 10 À4 mg mL À1 ) and a 20 minute ultrasonication was applied before dropping 10 mL into a LabTek 8-well chamber immediately for air-drying. The super-resolution images of ATTO-550 conjugated DNAoligo labelled UCNPs were acquired at 40 Hz for up to 20 000 frames under the excitation of 561 nm laser (10 kW cm À2 at the sample) and activation of 405 nm laser (#5 kW cm À2 at the sample). The excitation beams were reected by a customdesigned polychroic mirror (z405/488/561/640, Chroma). Fluorescence emissions from ATTO-550 were ltered by a bandpass lter (605/70, Chroma). An imaging buffer (100 mM Tris/HCl pH 8.0, 20 mM NaCl and 10% glucose) with an oxygen scavenger system (60 mg mL À1 glucose oxidase, 6 mg mL À1 catalase) was used for the STORM imaging. STORM images were analyzed using Insight3 (provided by Dr Bo Huang, UCSF) for singlemolecule localization and custom-written Matlab codes for cluster analysis based on nearest-neighbour algorithm.

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