Engineering of Hollow Mesoporous Silica Nanoparticles for Remarkably Enhanced Tumor Active Targeting Efficacy

Hollow mesoporous silica nanoparticle (HMSN) has recently gained increasing interests due to their tremendous potential as an attractive nano-platform for cancer imaging and therapy. However, possibly due to the lack of efficient in vivo targeting strategy and well-developed surface engineering techniques, engineering of HMSN for in vivo active tumor targeting, quantitative tumor uptake assessment, multimodality imaging, biodistribution and enhanced drug delivery have not been achieved to date. Here, we report the in vivo tumor targeted positron emission tomography (PET)/near-infrared fluorescence (NIRF) dual-modality imaging and enhanced drug delivery of HMSN using a generally applicable surface engineering technique. Systematic in vitro and in vivo studies have been performed to investigate the stability, tumor targeting efficacy and specificity, biodistribution and drug delivery capability of well-functionalized HMSN nano-conjugates. The highest uptake of TRC105 (which binds to CD105 on tumor neovasculature) conjugated HMSN in the 4T1 murine breast cancer model was ~10%ID/g, 3 times higher than that of the non-targeted group, making surface engineered HMSN a highly attractive drug delivery nano-platform for future cancer theranostics.

H ollow mesoporous silica nanoparticle (HMSN), with a large cavity inside each original mesoporous silica nanoparticle (MSN, a well-known nanocarrier which has been intensively investigated for drug delivery applications since 2000 1,2 ), has recently gained increasing interests due to their tremendous potential as more advantageous nano-platform for cancer imaging and therapy 3,4 . The last several years have witnessed a rapid development in engineering of functionalized HMSNs (f-HMSNs) with various types of inorganic functional nanocrystals integrated into the system for imaging and therapeutic applications [5][6][7][8][9] . For example, Gd 31 -doped upconversion nanoparticle encapsulated HMSN has been developed for the integration of optical imaging, magnetic resonance imaging (MRI), chemotherapy and radiotherapy 10 . Another fascinating study is the combination of ultrasound imaging and MRI within one single HMSN, where MRI could potentially be employed as the pre-surgical evaluation tool while ultrasound can be exploited for real-time guidance during high intensity focused ultrasound (HIFU) therapy 11 .
In spite of the progress, most of the existing reports have been focused on the design of sophisticated f-HMSNs without fully examining their potential for actively targeted cancer imaging and/or therapy after systemic intravenous (i.v.) administration. Currently, very few reports about the quantitative assessment of in vivo tumor uptake, biodistribution and clearance studies of HMSN exist in the literature. Applying HMSN to in vivo targeted imaging and drug delivery is still considered as one of the major challenges in this field, possibly due to the lack of efficient in vivo targeting strategy and well-developed surface engineering techniques.
Recently, we reported the first example of in vivo tumor targeting and enhanced drug delivery of TRC105 (a human/murine chimeric IgG1 monoclonal antibody, which binds to both human and murine CD105 12 ) antibody conjugated MSN, and demonstrated the promising potential in applying tumor vasculature targeting strategy to functionalized nanoparticles for possible better therapeutic outcome 13 . However, the drug loading capacity of MSN needs to be further improved and the highest tumor uptake was not ideal (,6 percentage injected dose per gram of tissue [%ID/g]).
Here, we report the in vivo tumor targeted positron emission tomography (PET)/near-infrared fluorescence (NIRF) dualmodality imaging and enhanced drug delivery of HMSN by using a generally applicable surface engineering method. The anti-cancer drug loading capacity of HMSN was found to be 3-15 times higher than previously reported MSN. By using CD105 (also known as endoglin, which is a validated marker for tumor angiogenesis and almost exclusively expressed on proliferating tumor endothelial cells 14,15 ) based tumor vasculature targeting strategy, we achieved over 3-fold higher tumor uptake (,10%ID/g at 4 h post-injection [p.i.]) in targeted group when compared with the non-targeted group (,3%ID/g at 4 h p.i.). Systematic in vitro and in vivo studies were performed to investigate the stability, targeting efficacy and specificity, biodistribution and drug delivery capability of well-functionalized HMSN nanoconjugates. With the growing interests in using HMSN as a drug delivery nano-platform, this study may pave the way for future cancer theranostics using HMSN, since silica is generally regarded as safe (GRAS) by the Food and Drug Administration (FDA) 4 .

Results
Synthesis and characterization. Uniform HMSN with an average particle size of 150 nm was synthesized following the literature procedures with improved Na 2 CO 3 -etching process 16 . Generally, three steps were involved. Firstly, uniform ,100 nm sized dense silica (dSiO 2 ) nanoparticles were prepared as hard templates using a modified Stöber method 17 . Secondly, shell thickness controllable MSN layer was coated on each dSiO 2 using cetyltrimethylammonium chloride (i.e. CTAC) as the soft templates at 80uC, forming dSiO 2 @MSN(CTAC) (Figure 1a, dark core represents dSiO 2 core, while lighter shell is MSN). Thirdly, as-synthesized dSiO 2 @ MSN(CTAC) together with free CTAC (,66.7 mg/mL) in the solution was transferred to 50uC water bath, and kept etching for 30 min after addition of Na 2 CO 3 (21.2 mg/mL). In comparison with previously reported method 16 , which usually needs more than 10 h of etching, our modified method avoids the tedious washing steps with no re-addition of CTAC, and the etching process could be easily completed within 30 min possibly due to the increased CTAC concentration. Figure 1b shows the representative transmission electron microscopy (TEM) image of as-prepared HMSN(CTAC) with oriented mesoporous shell structures. The surfactant CTAC was later removed via an extraction process by stirring HMSN(CTAC) in 1 wt% solution of NaCl in methanol 18 . The pore size of HMSN was 2-3 nm, similar to that of MSN as we reported previously 13 . The diameter of hollow space was ,100 nm, and the thickness of MSN shell was measured to be ,25 nm. Both hollow cavity and MSN shell thickness are readily tunable. For example, using smaller size dSiO 2 (,50 nm), which could be synthesized in a water-in-oil reverse microemulsion 19 , uniform HMSN with a smaller cavity inside (,50 nm) and thinner MSN shell (,7.5 nm) outside could also be synthesized ( Figure S1).
As proposed by Zheng et al., selective etching of dSiO 2 core and redeposition of the soluble silicate species are both important steps during the formation of HMSN 16 . Cationic surfactants (in our case CTAC) play vital roles in such etching-redeposition mechanism 16 . During the etching process, positively charged cetyltrimethylammonium cations (CTA 1 ) will first absorb to the surface of dSiO 2 @ MSN(CTAC) via electronic attraction. Subsequently, with the presence of Na 2 CO 3 in the solution and CTAC in the MSN shell, selective etching of silica from the dSiO 2 core starts and will be accelerated by the surrounding free CTAC, forming uniform HMSN(CTAC) after the redeposition process.
We also found that, to selectively etch out the dSiO 2 core and leave the MSN shell intact, the CTAC concentration, etching temperature and etching time should be tightly-controlled. Only dSiO 2 will be obtained, and all the MSN shell will be etched away if the concentration of CTAC was kept at below 6.25 mg/mL ( Figure S2a). Increasing the CTAC concentration up to 66.7 mg/mL will result in uniform HMSN only when etching is done in a water bath with a mild temperature (i.e. 50uC) for a relatively short time (e.g. 30-60 min), as shown in Figure S2b, c. Too high etching temperature (e.g. 80uC) and too long etching time will easily cause over-etching of HMSN, leaving aggregated nanoparticles with highly rough surfaces ( Figure S2d,e,f).
A representative TEM image of HMSN-ZW800-TRC105 is shown in Figure 1c, where no obvious morphology change of HMSN could be observed after multi-step surface modification. The final nanoconjugate could be well-dispersed in phosphate buffered saline (PBS). Average size of HMSN-ZW800-TRC105 was found to be ,150 nm, which is slightly smaller than the Z-average size (i.e. 194.4 nm) measured by dynamic light scattering (DLS). Figure S3 also presents the size distribution of HMSN-ZW800-TRC105, acquired by DLS. Slightly negatively charged surface (25.1 6 0.9 mV) of HMSN-ZW800-TRC105 was also observed.
www.nature.com/scientificreports SCIENTIFIC REPORTS | 4 : 5080 | DOI: 10.1038/srep05080 ratio during in vivo image-guided surgery, in comparison with traditional dyes such as Cy5.5 and IRDye800CW 20 . Therefore, ZW800-NHS ester was covalently conjugated to the outer surface of HMSN-NH 2 to render the nanoparticle NIRF imaging capability. To avoid the possible dye-dye self-quenching, the optimal ratio between HMSN and ZW800 was carefully investigated (Figure 3). The lowest degree of ZW800 quenching (,4.8%) was observed when keeping the mole ratio as low as 152, where up to 81.4% conjugation efficiency was achieved with each HMSN covered by approximately 1.6 ZW800 dyes (Figure 3a, Table S1). Increasing the amount of dyes while keeping that of HMSN unchanged would increase the number of dye per HMSN to larger than 8, as shown in Table S1, however it also caused severe dye-dye quenching: About 74% and 92% degree of quenching were estimated when fixing the ratio to be 1510 and 1520, respectively (Figure 3b,c and Table S1). Therefore, the optimized ratio of 152 was used in the following in vivo tumor targeted imaging studies.
In vivo tumor targeting and PET/NIRF dual-modality imaging. PET is highly sensitive, quantitative, clinically relevant, and has excellent tissue penetration of signal [21][22][23][24] . Labeling nanoparticles with positron-emitting radionuclides has been generally recognized as the most accurate means for non-invasive evaluation of their biodistribution and pharmacokinetics [25][26][27][28] . For in vivo tumor targeted PET imaging, NOTA-HMSN-ZW800-TRC105 was labeled with 64 Cu and purified using PD-10 columns with PBS as the mobile phase. The radioactivity fractions (typically elute between 3.0 and 4.0 mL) were collected for further in vivo experiments. A typical size exclusion column chromatography profile is provided in Figure S4. Inset in Figure S4 also shows the PET and NIRF imaging of fraction 3.0-3.5 mL, demonstrating the successful synthesis of 64 Cu-HMSN-ZW800-TRC105.
Since PET imaging detects the radioisotopes (e.g. 64 Cu in this work) rather than the HMSN conjugates per se, excellent in vivo stability is a prerequisite for PET to truly reflect the in vivo distribution of 64 Cu-HMSN-ZW800-TRC105. Serum stability study ( Figure S5) of 64 Cu-HMSN-ZW800-TRC105 showed that over 99% of 64 Cu remained intact within 64 Cu-HMSN-ZW800-TRC105 over a 24 h incubation period in whole mouse serum at 37uC while shaking, which indicates high stability of as-synthesized 64 Cu-HMSN-ZW800-TRC105.
In vivo tumor targeted PET imaging were then carried out in 4T1 murine breast tumor-bearing mice, which express high level of CD105 on the tumor neovasculature 14,15 . Each mouse was injected with 5-10 MBq of 64 Cu-HMSN-ZW800-TRC105, and time points of 0.5, 4, 16, 24 h p.i. were chosen for serial PET scans to show the in vivo biodistribution patterns of mice from targeting, non-targeting and blocking groups ( Figure 5). Quantitative data obtained from region-of-interest (ROI) analysis of these PET images are also shown in Table S2-S4.
The accumulation of 64 Cu-HMSN-ZW800-TRC105 in the 4T1 tumor was found to be 8.5 6 1.1%ID/g at 0.5 h p.i., and peaked at 9.9 6 0.9%ID/g at 4 h p.i., as shown in Figure 5a,6a,S6 and Table S2 (n 5 4). In contrast, without the conjugation of TRC105 (i.e. passive targeting alone), the 4T1 tumor uptake of 64 Cu-HMSN-ZW800 was found to be only one third of the targeted group at all of the time points examined (n 5 4; Figure 5b, Table S3), indicating that TRC105 conjugation could be the controlling factor for enhanced tumor accumulation of 64 Cu-HMSN-ZW800-TRC105. To further confirm CD105 targeting specificity of 64 Cu-HMSN-ZW800-TRC105 in vivo, blocking studies were also performed. Administration of a blocking dose (1 mg/mouse) of free TRC105 at 1 h before 64 Cu-HMSN-ZW800-TRC105 injection could significantly reduce the tumor uptake to 4.1 6 1.7 and 5.5 6 2.2%ID/g at 0.5 and 4 h p.i., respectively (n 5 3, Figure 5c, Table S4), clearly demonstrating CD105 specificity of 64 Cu-HMSN-ZW800-TRC105 in vivo. Figure 6c also summarizes the comparison of 4T1 tumor uptake of 3 groups at different time points, where 64 Cu-HMSN-ZW800-TRC105 shows the highest tumor uptake throughout the study period.
Besides significantly improved tumor targeting efficacy, the tumor-to-muscle (T/M) ratios increased as well. As shown in Figure 6b, T/M value of the targeted group was estimated to be 9.0 6 1.7 at 0.5 h p.i., and increased up to 13.7 6 0.6 at 4 h p.i., which is significantly higher than that of non-targeting and blocking groups ( Figure S7, Table S6-7). Similar as what we have observed previously with 64 Cu-MSN-TRC105 13 , besides tumor accumulation, most of 64 Cu-HMSN-ZW800-TRC105 nano-conjugates were taken up by the reticuloendothelial system (RES) with liver uptake found to be 27.8 6 1.6%ID/g at 0.5 h p.i. and decreased gradually to 15.4 6 1.1%ID/g at 24 h p.i. (n 5 4; Figure 6a, Table S2), which was validated by the ex vivo biodistribution study (Figure 6d).
Engineering of HMSN to integrate two or more imaging modalities into one nanosystem has become a recent trend, which could provide complementary and more accurate information about the pharmacokinetics of HMSN in vivo 8 . Generally speaking, optical imaging is inexpensive, widely available, and highly sensitive for superficial tissues, which has been extensively used for monitoring various molecular/biological events in cells and small animal models 29,30 . Although various strategies (e.g. organic dye doping, fluorescent nanoparticle encapsulation, etc.) have been developed to engineer MSNs for optical imaging applications in vitro and in vivo 31 , very few reports about the engineering of HMSN for in vivo targeted optical imaging exist in the literature.
With the presence of optimized number of ZW800 on the surface of each HMSN nano-conjugate, the in vivo NIRF imaging was also carried out. Each 4T1 tumor-bearing mouse was first shaved and injected with HMSN-ZW800-TRC105 with 800 pmol of ZW800. Similar time points (i.e. 0.5, 4, 20, 24 h p.i.) as that of PET imaging studies were chosen for serial optical imaging using the IVIS Spectrum system (Ex 5 745 nm, Em 5 800 nm). Figure 7 shows the in vivo NIRF images of mice from 3 groups. Significantly higher optical intensity of tumor (marked with red circle) from targeted group could be clearly seen (Figure 7a-c), and was confirmed by the ex vivo imaging (Figure 7d-f, right lane). For all 3 groups, tumor signals decreased over time, possibly due to the wash-out of nanoconjugates in vivo. Ex vivo imaging of major organs, e.g. liver (Li), spleen (S), lung (Lu), kidney (K), heart (H) and tumor (T), showed similar distribution patterns as what has been shown in Figure 6d, with liver and spleen being the major nanoparticle uptake organs.
Enhanced in vivo tumor targeted drug delivery. Although MSN has been considered as a drug delivery nano-platform for over 10 years 1 , it was not until recently that the first example of quantitative assessment of tumor targeting efficacy and in vivo enhanced drug   13 . In this work, we demonstrated that HMSN could become a more attractive nanocarrier for future drug delivery applications.
In comparison with the DOX loading capacity in MSN (i.e. 71.6-481.6 mg/g), with the presence of hollow cavity inside HMSN, a significantly improved (about 3-15 times higher) drug loading capacity, up to 1129.2 mg/g, was achieved ( Figure S8a). The successful loading of DOX was further supported by the UV-vis absorbance spectra of HMSN(DOX), which exhibited the characteristic absorption peak at around 480 nm (Figure 8a, red line). DOX is known to be intrinsically pH sensitive 32 . Due to the protonation of silanols (-Si-OH) in HMSN with the decrease of pH, electrostatic interaction between DOX and HMSN will be decreased, causing the dissociation of DOX from the silica surface and mesoporous channels. A pHsensitive DOX release profile was also observed in HMSN(DOX), where shows a faster DOX release rate at acidic condition (pH 5.0) than that at neutral condition (pH 7.4, Figure 8b). Besides loading one type of hydrophilic anti-cancer drug, e.g. DOX, we also demon-strated here that HMSN could be a promising nanocarrier for loading certain hydrophobic drugs, such as Sunitilib (SUN, Figure S9), as well as double-loading with both SUN and DOX ( Figure S10).
As a proof-of-concept, we further demonstrated the feasibility of enhanced tumor targeted drug delivery in vivo using TRC105 conjugated HMSN(DOX), denoted as HMSN(DOX)-TRC105, after i.v. injection in 4T1 tumor-bearing mice (HMSN dose: 10 mg/kg, DOX dose: 6.5 mg/kg). Half an hour after injection of HMSN(DOX)-(w/ o)-TRC105 (w/o denotes with and without), the major organs were collected and imaged in the IVIS system (Ex 5 465 nm; Em 5 580 nm) to detect the presence of DOX, as shown in Figure 8c, d.
It is important to note that due to different absorption/scattering behaviors of DOX in various tissues, optical signal intensities from different organs may not accurately reflect the absolute uptake level of injected HMSN(DOX)-(w/o)-TRC105. For example, although liver is the dominant organ for HMSN nano-conjugates accumulation, as evidenced in our PET/NIRF imaging and biodistribution studies ( Figure 5, 6d, and 7), only weak optical signal could be observed based on ex vivo optical imaging because of its dark color and strong absorbance of visible DOX fluorescence (Figure 8d). In contrast, due to the much lighter color of tumor tissue, dominant optical signal from DOX could be observed in mice injected with HMSN(DOX)-TRC105, which is significantly stronger than the control group without TRC105 conjugation. Therefore, the significantly enhanced tumor targeting efficiency and drug (or multiple drugs) loading capacity will make antibody conjugated HMSN a highly attractive nano-platform for future cancer targeted imaging and therapy.

Discussion
Although molecular imaging of functionalized HMSN using imaging techniques, such as MRI 11 , optical imaging 10 and ultrasound 7 , has been reported previously 8 , to date most of the studies have been primarily focused on optimization of nanoparticle synthesis. To the best of our knowledge, no in vivo tumor active-targeted PET imaging, quantitative tumor uptake, and biodistribution studies of HMSN have been reported. By surface engineering of uniform HMSN with radioisotopes (i.e. 64 Cu), NIR dyes (i.e. ZW800) and CD105 targeting antibodies (i.e. TRC105), we reported here the first example of in vivo tumor vasculature targeted PET/NIRF dualmodal imaging of 64 Cu-HMSN-ZW800-TRC105. An average of 3-fold enhancement of tumor accumulation was achieved in targeted group when compared with the non-targeted group. The CD105 specific targeting was further confirmed with blocking studies both in vitro and in vivo. It is also worthy to mention that the highest 4T1 tumor uptake of 64 Cu-HMSN-ZW800-TRC105 (9.9 6 0.9%ID/g at 4 h p.i., n 5 4) was nearly 2 times higher than our previously reported 64 Cu-MSN-TRC105 (5.9 6 0.4%ID/g at 5 h p.i., n 5 3) 13 , as shown in Figure S8b, resulting in remarkable enhancement in targeting efficacy as well as T/M ratio (13.7 6 0.6 at 4 h p.i. vs 6.2 6 1.8 at 5 h p.i.) 13 .
Meanwhile, although tumor vasculature targeting has recently been demonstrated to be a versatile strategy for many well-functionalized nanomaterials regardless of tumor type 28 , the absolute level of tumor accumulation for most of the nanomaterials investigated to date is still relatively low, typically in the range of 3-6%ID/g 13,25,[33][34][35][36][37] , significantly lower than that of liver uptake (usually 20-30%ID/g or higher for most of the nanomaterials upon i.v. injection). To the best of our knowledge, 64 Cu-HMSN-ZW800-TRC105 is one of the very few vasculature targeted nanoparticles (another sample is RGD peptide conjugated single-walled carbon nanotubes which target integrin a v b 3 38 ) that could have about 10%ID/g of in vivo tumor uptake. Although we believe surface engineering of nanoparticles could play  a vital role, elucidating the exact reasons behind such remarkably enhanced tumor uptake requires systematic side-by-side tumor uptake comparison of well-engineered HMSN with varied particle size, morphology, surface charge, antibody density, etc., which deserves comprehensive investigation in the future. Considering the presence of large drug reservoir (i.e. hollow space) inside each HMSN, TRC105 conjugated HMSN might become a highly attractive drug delivery platform for future cancer theranostics.
In conclusion, we reported the engineering of uniform HMSN for remarkably enhanced in vivo tumor vasculature targeted PET/NIRF dual-modality imaging and enhanced drug delivery. Uniform and size-controllable HMSN was synthesized with a modified hard-templating method, which were subjected to a generally applicable surface engineering process, including amino groups functionalization, anti-cancer drug loading, NIR dye and NOTA linkages, PEGylation, antibody conjugation and radiolabeling, forming ,150 nm sized mono-disperse 64 Cu-HMSN(DOX)-ZW800-TRC105 nano-conjugates. The optimal ratio between HMSN and ZW800 was also carefully investigated to avoid the self-quenching effect, and found to be 152 with less than 5% of quenching. Systematic in vitro and in vivo studies was performed to investigate the stability, targeting efficacy and specificity, biodistribution and drug delivery capability of wellmodified HMSN nano-conjugates. Quantitative PET imaging study showed an over 3-fold higher of tumor uptake in the targeted group when compared with the non-targeted group, demonstrating that 64 Cu-HMSN-ZW800-TRC105 is one of the very few vasculature targeting nanoparticles with high (,10%ID/g) tumor uptake. Furthermore, with the presence of hollow space inside HMSN, 3-15 times higher of DOX loading capacity has also been achieved when compared with MSN, with the highest loading capacity estimated to be 1129.2 mg/g. In vivo enhanced DOX delivery was also demonstrated in 4T1 tumor-bearing mice. Take together these results, active tumor targeting and drug delivery of functionalized HMSN may pave the way for further in vivo targeted cancer therapeutic studies as well as potential clinical translational of this promising type of nano-platform, since silica is generally regarded as safe by the FDA.

Methods
Synthesis of uniform HMSN. Uniform HMSN with an average particle size of 150 nm was synthesized following the literature procedure with improved Na 2 CO 3etching process 16 . Three major steps were involved in the synthesis of HMSN. Firstly, synthesis of uniform ,100 nm sized dSiO 2 using a modified Stöber method 17 . In a typical synthesis, 35.7 mL of absolute ethanol was mixed with 5 mL water and 0.8 mL of ammonia and stirred for 5-10 minutes at room temperature. Then 1 mL of TEOS was added and the mixture was allowed to react at room temperature for 1 h. Afterward, dSiO 2 nanoparticles was washed with water and ethanol and suspended in 20 mL of water.
Secondly, synthesis of dSiO 2 @MSN. CTAC (2 g) and TEA (20 mg) were dissolved in 20 mL of high Q water and stirred at room temperature for 1 h. Then, 10 mL of www.nature.com/scientificreports SCIENTIFIC REPORTS | 4 : 5080 | DOI: 10.1038/srep05080 dSiO 2 water solution was added and stirred at room temperature for 1 h before addition of 0.15 mL of TEOS. The mixture was stirred for 1 h at 80uC in a water bath to form dSiO 2 @MSN.
Thirdly, etching of dSiO 2 @MSN to form HMSN. Both mixture and water bath were cooled down to 50uC followed by addition of 636 mg of sodium carbonate (Na 2 CO 3 ), which was under constant stirring for 30 min to form HMSN. To remove the CTAC, the product was extracted for 24 h with a 1 wt% solution of NaCl in methanol at room temperature. This process was carried out for at least 3 times to ensure complete removal of CTAC.
Synthesis of HMSN-NH 2 . To functionalize the HMSN surface with -NH 2 groups, assynthesized HMSN was first dispersed in 20 mL of absolute ethanol, followed by addition of 1 mL of APS. The system was sealed and kept at 86-90uC in a water bath for 24 h. Afterward, the mixture was centrifuged and washed with ethanol for several times to remove the residual APS. The HMSN-NH 2 could be well-dispersed in water, and the concentration of -NH 2 groups (nmol/mL) was measured using a Kaiser test kit.