The principle and procedure of the proposed approach are schematically illustrated in Fig. 1. To endow nanoparticles to prepare with protein targeting capability, a N- or C-terminal nonapeptide, which is a characteristic fragment of a target protein, was selected as an epitope. The selected epitope was grafted with a hydrophobic fatty acid chain of appropriate length (with a C13 chain in this study), which was used as the imprinting template. C-terminal epitopes were directly grafted with a fatty acid chain. While for N-terminal epitopes, a lysine (K) was first introduced to the C-terminal of the epitopes and then grafted with a fatty acid chain. For the imprinting and cladding, a reverse microemulsion formed with an appropriate surfactant and an oil phase was constructed as a nanoscale reaction cell to confine nanospheres to generate in the aqueous phase. When preparing core/shell cMIPs, nanoparticles to be encapsulated were dispersed into the microemulsion; otherwise, the generated cMIPs were coreless. Due to the presence of a hydrophobic chain on the template, the imprint epitope was anchored at the aqueous/oil interface of the microemulsion, with the imprint epitope protruding in the confined aqueous phase. After the silylating reagents dissolved in the oil phase (cyclohexane in this study) diffused into microemulsion, they hydrolyzed and then polymerized in the aqueous phase. After polymerization for adequate period (usually 24 - 48 h), a certain amount of tetraethyl orthosilicate (TEOS) was added to the microemulsion mixture. With TEOS molecules diffusing into the microemulsion, a hydrophilic silica cladding thin-layer was in situ formed on the surface of the imprinted nanospheres. Finally, the imprinted and cladded nanospheres were washed with an appropriate solution that can disrupt microemulsion and extract the template out of the polymeric shell. This processing generated cMIPs with well-formed cavities complementary to the template in aspects of shape, size and functionalities but with non-specific binding sites diminished by the cladding thin-layer. Therefore, the prepared core/shell nanoparticles not only maintained the original function of the core nanoparticles but also gained antibody-comparable affinity and specificity.
In order to obtain high affinity, it is essential to use multiple functional monomers with different functionalities capable of interacting with the target molecules via various interactions. The selection of silylating reagents monomers relies on the property classification of amino acids and functional monomers reported previously38,39. As shown in Supplementary Fig. 1, amino acids were classified into five classes: I) acidic, II) basic, III) aromatic, IV) hydrophobic, and V) others. Accordingly, four kinds of functional monomers were selected. Aminopropyltriethoxysilane (APTES), which contains an amino group, was used to interact with class I amino acids through electrostatic attraction as well as class V amino acids via hydrogen bonding. 3-Ureidopropyl-triethoxysilane (UPTES), which contains a carbamido moiety, was used to bind with amino acids of class II and class V via mainly hydrogen bonding. Benzyltriethoxysilane (BnTES), which contains a phenyl moiety, was selected to interact with class III amino acids via π-π stacking interaction. Isobutyltriethoxysilane (IBTES), which contains a hydrophobic moiety, can interact with class IV amino acids via hydrophobic interaction. TEOS does not contain any functional moieties, but it can function as a crosslinker to form a silica skeleton as well as a hydrophilic cladding thinlayer to cover non-imprinted area. Since different epitopes contain different numbers of different type of amino acids, the type of the monomers to be used for a specific epitope should be selected according to the kinds of amino acids presented in the epitope while the ratio of the monomers should be optimized experimentally.
Method development
To demonstrate the general applicability of the proposed method, three important protein disease biomarkers were used as the test targets, including B2M, HER2 and GPNMB. B2M is a biomarker for multiple myeloma while HER2 and GPNMB are two biomarkers for breast cancer. Particularly, GPNMB is a newly discovered biomarker for triple negative breast cancer. The amino acid sequences of the epitopes of these targets are shown in Supplementary Fig. 2. It should be noted that due to the different sources of GPNMB, two slightly different peptide sequences were used as the epitopes to fabricate cMIPs for recognizing GPNMB. Using the C-terminal nonapeptide of B2M as the epitope to prepare coreless cMIP, we first developed the imprinting and cladding method and confirmed the expected binding specificity.
The composition of the silylating reagents may greatly influence the rate of hydrolysis, which may affect the material formation and ultimately affect the imprinting effect. As the diffusion of the silylating reagents into the microemulsion took time, the reaction time required by the ROSIC method was generally long (typically 24 - 48 h). Besides, aqueous ammonia acts as both a reactant (H2O) and a catalyst (NH3) for the hydrolysis of silylating reagents. On one hand, a sufficient amount of ammonia was needed to ensure hydrolysis. On the other hand, too much ammonia would increase the size of the aqueous domain, which might also affect the material formation and ultimately affect the imprinting effect. To facilitate the reaction rate, a certain amount of ammonia was added to the aqueous phase. Therefore, providing that the reaction time was long enough, it was no need to optimize the imprinting time. We optimized the specific ratio of monomers and ratio between total monomers and crosslinking agent (TEOS) in terms of imprinting factor (IF). IF is an essential parameter that reflects the imprinting effect, which is calculated by the ratio of the amount of template molecules captured by the prepared MIPs and cMIPs over that by non-imprinted polymers (NIPs) and cladded non-imprinted polymers (cNIPs).
As shown in in Fig. 2 and Supplementary Fig. 3, among the specific monomer ratios investigated, the ratio of APTES/UPTES/BnTES/IBTES at 20:20:50:10 yielded the best imprinting effect for both MIP and cMIP, giving the IF value of 6.7 and 13.7 for MIP and cMIP, respectively. This suggests that the best specific monomer ratios for a given imprint was the same for both MIP and cMIP. However, the optimal ratio of total monomers/TEOS was different for MIP and cMIP. The highest IF value for MIP was found at the monomers/TEOS ratio of 20:80, whereas the best ratio of monomers/TEOS for cMIP shifted to 30:70. It can be seen that under all the ratios investigated, the amount of template captured by the NIPs changed apparently, while the amount of template captured by the cNIPs was greatly reduced to a low and constant level, which suggests that the cladding process was effective in all cases. Clearly, the improved imprinting effect in the MIC process was due to the fact that the template adsorption capability of non-imprinted surface was greatly suppressed by the cladding thin-layer.
On the other hand, it can be seen that under all specific monomer ratios investigated, the amount of template molecules captured by MIPs and cMIPs increased as increasing the total ratio of the monomer over TEOS within the range from 5:95 to 30:70. However, further increase in the ratio resulted in dramatic drop in the amount of template captured. To elucidate the mechanism, using SEM and TEM, we characterized the MIP, cMIP, NIP and cNIP nanoparticles prepared at all the monomers/TEOS ratios. The results are shown in Supplementary Fig. 4-11. From Supplementary Fig. 4-9, it can be seen that the prepared nanomaterials showed uniform spherical morphology, with a diameter of about 45 nm. The sizes of nanomaterials prepared by MI and MIC were similar, which indicates that the cladding layer was very thin. We found that when the ratio of monomers/TEOS was 40:60 (Supplementary Fig. 10), the prepared materials became less uniform, smaller in size, and aggregated. When the ratio was increased to 50:50, this phenomenon became more obvious (Supplementary Fig. 11). This indicates that the backbone structure and morphology of the prepared materials were mainly determined by the crosslinker TEOS and thereby a high crosslinker/monomer ratio was required.
Property Characterization
Using fluorescence-labeled B2M C-terminal epitope (FITC-KIVKWDRDM) as the test compound, the affinity of the materials prepared by MI and MIC are comparatively investigated. The binding isotherms of B2M C-terminal epitope-imprinted MIPs and cMIPs prepared under above optimized conditions were established by plotting the fluorescence intensity for the test compound captured by the materials against the logarithmic concentration of the test compounds. Using corresponding data, the dissociation constants (Kd) of B2M C-terminal epitope-imprinted MIPs and cMIPs were measured according to the Scatchard equation. As shown in Supplementary Fig.12, the Kd values were found to be 1.39±0.39 and 1.41 ±0.39 × 10-8 M for MIP and cMIP, respectively, suggesting that high affinity was maintained in MIC.
The specificity of the above prepared MIP and cMIP was comparatively investigated at the peptide level and the protein level. As shown in Fig. 3, the MIP showed apparently poor specificity, yielding cross-reactivity £ 16.9% towards the interfering peptides and cross-reactivity £ 19.1% towards the interfering proteins. As comparison, the cMIP exhibited much improved specificity, giving cross-reactivity £ 8.3% at the peptide level and cross-reactivity £ 8.1% at the protein level. Thus, the expected specificity enhancement by the cladding processing was experimentally confirmed.
Controllable engineering of core/cMIP shell nanoparticles
By virtue of the unique nano-confinement effect of reverse microemulsion, we further explored the possibility of the ROSIC approach for controllable engineering of core/cMIP shell nanoparticles. As shown in Fig. 4a, the process included five steps: 1) ligand exchange, 2) addition of aqueous phase, 3) ligand exchange again and adding silylating monomers, 4) phase transfer, and 5) imprinting and cladding. The key step of the reverse-microemulsion-based functionalization is to introduce nanoparticles into the nanometer-sized droplets. One simple way is in situ synthesis of nanoparticles and then direct functionalization. However, most nanoparticles need to be pre-synthesized to obtain particular physical and chemical properties, which can also be functionalized by reverse microemulsion method. The mechanism for the functionalization of nanoparticles stabilized with hydrophilic ligands is straightforward. Due to the hydrophilicity, such nanoparticles can enter the aqueous phase directly. Recently, the silica-based functionalization mechanism of pre-synthesized nanoparticles stabilized with organic ligands by reverse microemulsion method has been well elucidated by many researches40-43. Usually, organic ligands, like alkyl amino chains or alkyl acid chains, on the surface of nanoparticles are labile, can be easily replaced by both hydrolyzed silylating monomers and surfactant molecules, which facilitates the incorporation of the nanoparticles into the reverse micelles (Fig. 4a).
In this work, we successfully achieved controllable functionalization of various dual-functional size-tunable core/shell nanoparticles by ROSIC. As show in Fig. 4b, the prepared QD@cMIP, Ag@cMIP, SPMNP@cMIP and UCNP@cMIP exhibited uniform and single-cored morphology, with a uniform diameter. This confirms that the ROSIC process truly achieved precisely controllable engineering functionalization of nanoparticles.
The functional shell of different thicknesses has its own specific function, and its control is very important for different biological applications. It has been reported that the thickness of silica shells could be tuned by changing the amount of silylating monomers like TEOS, but the tuning is very limited and free-core silica or multicore silica constantly appeared because of the heterogeneous or homogeneous nucleation of SiO2. The use of multiple silylating monomers in this work could also make its controllability more challenging. Previous work44 has found that the SiO2 shell thickness basically remains constant as the TEOS content increases up to a certain extent, which is due to confinement of aqueous domain size on the overall size of core/shell nanoparticles. This is also the reason to ensure the effectiveness of our ROSIC approach. Inspired by this, it is a suitable way to control the thickness of the overall size of core/shell nanoparticles by controlling the size of the aqueous phase (Fig. 4c). The size control of the aqueous phase can be achieved by simply adjusting the composition of microemulsion, like turning the ratio of surfactant to water or change the type of surfactant. In this work, single-cored and size-controllable QD@cMIP NPs with size of 30, 40 and 50 nm were successfully prepared by the ROSIC method (Fig. 4c).
Fluorescence Imaging of Triple-Negative Breast Cancer Cells
TNBC refers to a subtype of breast cancer that does not express three major therapeutic targets including estrogen receptor (ER), progesterone receptor (PR) and HER2. So far, effective treatment of TNBC still remains lacking and challenging. GPNMB is a type I transmembrane glycoprotein and expressed in the tumor stroma of 64% of human breast tumors and in the tumor epithelium of an additional 10% of tumors45. Recent studies45,46 have identified high GPNMB expression in TNBC and suggested GPNMB as a potential biomarker for clinical research47,48. In this work, using QD520 (with emission maximum at 520 nm) and QD620 (with emission maximum at 620 nm) as fluorescent nanocores, GPNMB-specific QD520@cMIP and HER2-specific QD620@cMIP were prepared by the proposed ROSIC method for targeting TNBC cells (GPNMB+) and HER2+ breast cancer cells, respectively.
We first optimized the imprinting conditions for the preparation of anti-GPNMB QD520@cMIP. According to above established knowledge, the ratio of monomers/TEOS was fixed at 30:70. As shown in Supplementary Fig. 13a, the best imprinting was obtained when the ratio of APTES/UPTES/IBTES/BnTES was 10:30:50:10, giving a high IF value (12.3). Because a single QD was introduced into the cMIP, we further investigated the imprinting effect within a higher monomers/TEOS ratio range (from 30:70 to 50:50) at the optimal specific monomer ratio to check if the previous conclusion was still valid. From Supplementary Fig. 13b, we found that as increasing the monomers/TEOS, the IF value decreased too. This is in agreement with above knowledge of the requirement of high TEOS/monomers ratio for good material structure. Thus, whether or not the cMIP contains a nanocore, the ratio of monomers/TEOS should be fixed at 30:70. The specificity of anti-GPNMB QD520@cMIP was investigated. The cross-reactivity was less than 8.9% at the peptide level and less than 9.3% at the protein level (Supplementary Fig. 13c and 13d).
Keeping the ratio of monomers/TEOS at 30:70, we further optimized the imprinting conditions for the preparation of anti-HER2 QD620@cMIP. Since the structure of N-terminus of HER2 contains no class III amino acids, BnTES was not used for the preparation of QD620@cMIP. As shown in Supplementary Fig. 14, the best monomer ratio was found at APTES/UPTES/IBTES of 10:30:60, giving an IF value of 11.8. The cross-reactivity was found to be less than 9.0% at the peptide level and less than 9.7% at the protein level.
The prepared nanomaterials were characterized by TEM and EDS elementary mapping. As show in Supplementary Fig.15, the prepared QD520@cMIP and QD620@cMIP exhibited uniform and single-cored morphology, with a uniform diameter of about 27 nm. It was found that the MIC process did not significantly affect the fluorescence properties of quantum dots (Fig. 5b-e).
Finally, the target recognition and fluorescence imaging capability of the QD520@cMIP and QD620@cMIP was investigated. QD520@cMIP, QD520@cNIP, QD620@cMIP and QD620@cNIP were incubated with human TNBC cells MDA-MB-157 (GPNMB+, HER2-), human breast cancer cells MDA-MB-361 (GPNMB+, HER2+) and MCF-7 cells (GPNMB-, HER2-). As shown in Fig. 5, GPNMB N-terminal epitope-imprinted QD520@cMIP showed very strong fluorescence intensity towards MDA-MB-157 and MDA-MB-361 cells, but no obvious fluorescence signals towards MCF-7 cells. HER2 N-terminal epitope-imprinted QD620@cMIP showed very strong fluorescence intensity to MDA-MB-361 cells, but no obvious fluorescence signals to MDA-MB-157 and MCF-7 cells. As a comparison, QD520@cNIP and QD620@cNIP had no fluorescence signals to the above three cells (Supplementary Fig.16). These results indicate that the prepared GPNMB-specific QD@cMIP could distinguish TNBC cells from other breast cancer cells, showing the great potential of QD@cMIP NPs in biomedical applications.