Uniform Drug Loading into Prefabricated Microparticles by Freeze-Drying


  • Seo Woo Song,

    1. Department of Electrical and Computer Engineering, Seoul National University, Kwanak-gu, Seoul, South Korea
    2. Institute of Entrepreneurial Bio Convergence, Seoul National University, Kwanak-gu, Seoul, South Korea
    Search for more papers by this author
  • Hyung Jong Bae,

    1. Department of Electrical and Computer Engineering, Seoul National University, Kwanak-gu, Seoul, South Korea
    2. Institute of Entrepreneurial Bio Convergence, Seoul National University, Kwanak-gu, Seoul, South Korea
    Search for more papers by this author
  • Sudeok Kim,

    1. Department of Electrical and Computer Engineering, Seoul National University, Kwanak-gu, Seoul, South Korea
    2. Institute of Entrepreneurial Bio Convergence, Seoul National University, Kwanak-gu, Seoul, South Korea
    Search for more papers by this author
  • Dong Yoon Oh,

    1. Institute of Entrepreneurial Bio Convergence, Seoul National University, Kwanak-gu, Seoul, South Korea
    2. Interdisciplinary Program of Bioengineering, Seoul National University, Kwanak-gu, Seoul, South Korea
    Search for more papers by this author
  • Okju Kim,

    1. Department of Electrical and Computer Engineering, Seoul National University, Kwanak-gu, Seoul, South Korea
    2. Institute of Entrepreneurial Bio Convergence, Seoul National University, Kwanak-gu, Seoul, South Korea
    Search for more papers by this author
  • Yunjin Jeong,

    1. Department of Electrical and Computer Engineering, Seoul National University, Kwanak-gu, Seoul, South Korea
    2. Institute of Entrepreneurial Bio Convergence, Seoul National University, Kwanak-gu, Seoul, South Korea
    Search for more papers by this author
  • Sunghoon Kwon

    Corresponding author
    1. Department of Electrical and Computer Engineering, Seoul National University, Kwanak-gu, Seoul, South Korea
    2. Institute of Entrepreneurial Bio Convergence, Seoul National University, Kwanak-gu, Seoul, South Korea
    Search for more papers by this author


Microparticle-based drug delivery is a promising technology for small volume bioassay platforms. The general utilization of the drug-loaded microparticles in the in vitro bioassay platforms requires the drug loading method, which should impregnate the general drug types (e.g., water insoluble) with high payload into the variously designed microparticles. Loading the drug into the prefabricated microparticles using solvent evaporation satisfies the requirement. However, similar to the “coffee-ring effect,” drugs are loaded in a seriously nonuniform manner, caused by the capillary flow during the evaporation process. Here, it is presented that the freeze-drying is an efficient way to load uniform and high amount of the drug into the prefabricated microparticles. It is demonstrated that freezing solvent can block the capillary flow during the solvent removal process, improving the loading uniformity. The delivered amount of drugs is linearly proportional to the initial loading amount of drugs. Also, this drug loading method is shown to be applied to the various drug types and the prefabricated microparticles with different properties. Considering many challenges to suppress the “coffee-ring effect” that induces nonuniform impregnation/deposition, the proposed concept can be applied not only for microparticle-based drug delivery but also for uniform coating applications (e.g., thin-film coating, DNA/protein microarray).

1 Introduction

In the biochemistry research field, there is a large trend toward the scaling down of reaction volume to reduce reaction time and experimental cost. As a result, nanoliter-scale bioassay platforms based on microwell or droplet technology have been developed.[1-6] These small volume experimental platforms facilitate many different bioassay types, including immunoassays,[7, 8] enzyme activity assays,[9] bacteria-based assays,[10, 11] and cell-based assays,[12-14] with reasonable costs and in a highly parallelized manner. These small-volume bioassay platforms led to the development of high-throughput and multiplexed bioassay platforms combined with a drug delivery method that can convey small amounts of drug for individual microreaction chambers.[15-18] To deliver drugs heterogeneously to individual microwells, ink-jet-based microarray spotters[19, 20] or patterned lithography[21, 22] have typically been applied. However, we have demonstrated that a microparticle-based drug delivery system (DDS) can accomplish the delivery of drugs heterogeneously to individual microreaction chambers without expensive equipment or labor-intensive work.[23, 24] The use of encoded microparticles allows pooling of heterogeneous drug-laden microparticles into a library, and those particles can be delivered simultaneously into the individual reaction chambers by directed assembly of one microparticle per microwell.

The most important point for nanoliter-scale bioassays is that the delivered amount of drugs should be precisely controlled because even small differences can cause large errors in total drug concentration, owing to the small volumes involved. Therefore, in microparticle-based DDS, an exact drug-loading and drug-releasing method is critical for accurate results. The most common drug-loading method is simple absorption. However, the drug absorption efficiency is largely influenced by swelling ratio, drug solubility as well as drug interactions with microparticles.[25, 26] Therefore, it is hard to attain sufficient loading amounts for general drug types (e.g., poorly water-soluble drugs) using simple absorption. Solvent evaporation from droplets composed of oil and solution of prepolymer and drug is a well-established technique to generate drug-laden microspheres with high drug load.[27, 28] However, usable solvent and polymer types for stable droplet formation are restricted. Available modifications of shapes or characteristics of microparticles are also limited; thus, it is more advantageous to use prefabricated microparticles in practice. To load large amounts of drugs into prefabricated microparticles, the loading process, which is carried out by mixing prefabricated microparticles with drug solutions followed by solvent evaporation, has been developed.[25, 29-31] Because solvent evaporation enables the dissolved drug molecules in solution to be completely absorbed into the microparticle or coated on the particle surface, high loading capacity is achievable regardless of the intrinsic features of the drugs and microparticles. To the best of our knowledge, no method that can effectively realize “uniform” drug impregnation into prefabricated microparticles at the individual particle level has been developed. In the case of experiments with a large number of microparticles, nonuniform drug loading is not a serious problem because the heterogeneity of loaded drug amounts is averaged out, and it does not affect experimental results. However, it is a serious problem for maintaining experimental result accuracy in applications where small reaction volumes and a single microparticle are used.

It is difficult to achieve uniform drug loading into prefabricated microparticles with solvent evaporation because of a phenomenon similar to the “coffee-ring effect.” When a liquid drop dries on a solid surface, the liquid boundary is pinned to the substrate, and capillary flow from the center toward the boundary drives dissolved molecules to the edge.[32] As a result, suspended molecules are deposited in a ring-like pattern after complete evaporation (Figure 1b). Similarly, when solvents are evaporated from mixtures of microparticles and drug solutions, the liquid boundary is pinned to the border of aggregated microparticles, which consequently induces capillary flow that drives the drug molecules to the upper layer of the microparticles (Figure 1c). Theoretically, continuous mixing can reduce this nonuniformity, but it is practically difficult because surface tension becomes dominant as the solvent dries out. Other approaches to suppress the coffee-ring effect, such as use of surfactant,[33, 34] acoustic wave,[35] and electrowetting,[36] were also not helpful for improving the uniformity. This was because the parameters (surfactant concentration and wave amplitude) should be continuously adjusted precisely according to the solute density and/or particulate matter size. Therefore, productivity is seriously limited with those methods because of complicated equipment or procedures.

Figure 1.

Drug loading into prefabricated microparticles. a) Drug loading via solvent evaporation, freeze-drying, and vortex freeze-drying (VFD). Drug solution was added to prefabricated microparticles and the drying process was performed under reduced pressure. Conventionally used solvent evaporation-based drug loading is nonuniform. i) Drug loading by simple freeze-drying. Because freeze-drying blocks capillary flow, uniformly drug-loaded microparticles can be obtained. ii) Drug loading by VFD. Vortexing motion during freezing process enables the even suspension of microparticles in the frozen solution, and thus loading uniformity can be further improved. b) Theoretical background of the coffee-ring effect and suppression of this effect by freeze-drying (Scale bars: 0.2 mm). When a liquid drop evaporates at room temperature, capillary flow drives dissolved molecules toward the edge during the drying process. This capillary flow is blocked during solvent sublimation and thus the coffee-ring effect is suppressed. c) Images of rhodamine-B-loaded microparticles via solvent evaporation and VFD. Rhodamine-B molecules are highly concentrated in the upper layer of microparticles with solvent evaporation, but uniform loading is achieved via freeze-drying (Scale bars: 2 mm).

In this study, we first demonstrated that uniform and high amount of drug loading into prefabricated microparticles is attainable via freeze-drying. Because the fundamental cause of nonuniformity is capillary flow that occurred in a liquid drop, freezing the solvent blocked this flow, and consequently, enhanced the loading uniformity (Figure 1c). At the same time, drug loading based on solvent removal allows free selection of solvent type and complete impregnation of drug molecules. Therefore, even poorly water-soluble drugs can be loaded in high amounts into microparticles by using organic solvents. Moreover, lyophilization generally guarantees simple storage conditions and long-term stability of bioactive substances.[37] We showed that the delivered amount of drugs is directly proportional to the initial loading amount of drugs. In addition, we showed that this drug-loading method could be applied to various types of drugs and microparticles. Considering the many challenges to overcome regarding the nonuniform loading/coating issue caused by the coffee-ring effect, our approach of using freeze-drying provides a promising strategy for microparticle-based drug delivery as well as for uniform coating applications such as thin-film coating and DNA/protein microarrays.

2 Results and Discussion

We compared three drug-loading methods in this study: solvent evaporation under reduced pressure at room temperature, freeze-drying (solvent sublimation), and vortexing during the freezing process followed by freeze-drying (vortex freeze-drying, VFD) (Figure 1a). The vortex motion during the temperature drop allows the microparticles to become evenly suspended in the drug solution before and after freezing, so that more uniform drug-loaded particles can be obtained after lyophilization. However, as the solvent remains in a liquid state at room temperature, the surface tension becomes dominant as the solvent evaporates. Therefore, because the vortexing motion cannot result in a significant improvement in uniformity, we did not choose to perform it when drying under room temperature.

We first evaluated loading uniformity at the individual particle level. For visualization, we used fluorescent rhodamine-B as a model chemical and analyzed the fluorescence intensity of microparticles spread in a monolayer (Figure 2a,b). Three loading methods—solvent evaporation by simple vacuum drying, solvent sublimation by freeze-drying, and VFD—were compared. The fluorescence images of rhodamine-B-impregnated microparticles are shown in Figure 2a, and the normalized intensity distribution curves are illustrated in Figure 2b (more details are provided in the Supporting Information). As shown in Figure S1 (Supporting Information), more uniform loading was possible when water was used as a solvent. However, to utilize this loading method for general drug types including water-insoluble drugs, dimethyl sulfoxide (DMSO) is a more widely applicable solvent. Therefore, we performed all the following experiments with DMSO as the solvent. As expected, the intensity of microparticles from simple vacuum drying showed a large coefficient of variation (CV; 43.2%), whereas the CVs from the freeze-drying and VFD methods were 21.0% and 11.8%, respectively. Freezing the solvent significantly improved the drug-loading uniformity, and it was further enhanced by the vortexing motion during temperature drop. The shape of the distribution changed from a right-skewed distribution to a normal distribution by altering the drying method from solvent evaporation to VFD (Figure 2b). This showed that solvent evaporation made a substantial amount of the rhodamine molecules to be highly concentrated in the upper layer of the microparticles, but the skewness was reduced by using VFD method owing to the blocking of the capillary flow, which causes the nonuniformity.

Figure 2.

Loading uniformity of the three drug-loading methods. a) Fluorescence image of rhodamine-B-loaded microparticles spread in monolayers (Scale bars: 200 µm). b) Fluorescence intensity distribution of microparticles. Coefficient of variation (CV) values were 43.2%, 21.0%, and 11.8% for simple vacuum drying, freeze-drying, and vortex freeze-drying (VFD), respectively (n = 322, 367, and 382). c) Fluorescence image of rhodamine-B releasing in microwells (Scale bars: 500 µm). d) Distribution of fluorescence intensity of the releasing solution. Microparticles were located at the center and impregnated rhodamine-B was released into the surrounding solution in the microwell (see also the Supporting Information). Fluorescence intensity was measured from the surrounding region. CV values were 45.9%, 19.0%, and 13.9% for simple vacuum drying, freeze-drying, and VFD, respectively (n = 21, 56, and 54). e) Scanning electron microscope (SEM) images of bare microparticles and rhodamine-B-loaded microparticles by the three loading methods. Wide view images are in the first row (Scale bars: 100 µm) and magnified images of the yellow-boxed region are in the second row (Scale bars: 50 µm). Microparticles from the simple vacuum drying method showed precipitated clods (yellow arrows). This means that rhodamine-B molecules were not loaded/deposited uniformly into/on the microparticles.

To identify whether the amount of impregnated fluorescent molecules is directly related to the amount released, we performed a microwell-based releasing experiment (Figure 2c,d). To place a single microparticle in a single microwell, a previously reported technique for microparticle assembly in a microwell array was used (Figure S2, Supporting Information).[23] Briefly, rhodamine-B-loaded microparticles were spread on a “particle chip” that has a microwell array (165 µm diameter, 40 µm depth). These dimensions were just slightly larger than those of the microparticles (150 µm diameter, 38 µm depth). Then, microparticles were assembled into microwells by sweeping with inert silicone oil as a lubricant. After particle assembly, the “particle chip” was combined face-to-face with a phosphate-buffered saline (PBS)-filled reaction chip having large microwells (600 µm diameter, 360 µm depth). This interaction was maintained overnight to facilitate the release of fluorescent molecules to the end-point (the release profile of the microparticles is illustrated in Figure S3, Supporting Information). After the completion of the releasing process, the fluorescence intensity of the releasing solution was measured. Fluorescence images of the microwells are shown in Figure 2c, and the normalized distribution curve of solution intensities are represented in Figure 2d (more details are provided in the Supporting Information). Noticeably, the uniformity of solution intensity was significantly enhanced by solvent sublimation and the VFD method. The CV values were 45.9%, 19.0%, and 11.7% for simple vacuum drying, simple freeze-drying, and the VFD method, respectively.

Images of the surface of rhodamine-B loaded microparticles observed by scanning electron microscope (SEM) also showed meaningful differences between the drug-loading methods (Figure 2e). Initially, bare hydrogel particles have clean surfaces, but several precipitated molecule clods were detected after loading via simple vacuum drying (represented with yellow arrows). This precipitation is possibly because rhodamine-B molecules, driven by capillary flow, were highly concentrated to the upper layer of microparticles. Thus, it was difficult to impregnate the molecule evenly into the hydrogel matrix. Conversely, loading by lyophilization and the VFD method, as expected, showed few agglomerated molecules.

Next, we demonstrated that the delivered amount of drug could be controlled easily and precisely according to the loading amount of drug. Five model chemicals with different molecular weights and solubility including poorly water-soluble drugs (rhodamine-B, doxorubicin hydrochloride, erlotinib hydrochloride, BIBW2992, and temozolomide) were dissolved in DMSO (Table 1), and the solutions were loaded into identical amounts of microparticles (13.2 mg, 15 225 microparticles) using the VFD method. After complete drying, the drug-loaded microparticles and rhodamine-B loaded microparticles were dispersed into 1.5 mL of PBS in order to release the loaded chemicals, and the final released concentrations were measured after overnight vortexing (end-point). The concentration was measured in bulk-scale because we had already verified that impregnated- and released amounts of molecules have small CV values and normal distribution with the VFD method. First, we changed the concentration of the loading solution with the same loading volume (20 µL) (Figure 3a and Figure S5, Supporting Information). Released chemical concentrations were calculated from a reference curve based on a concentration-absorbance curve measured via ultraviolet-visible spectrophotometry (Figure S4, Supporting Information). A simple linear relationship was observed between the loading concentration and released concentration with the same loading volume. Similarly, the loading volume and released concentration with the same loading concentration also showed a simple linear relationship (Figure 3b and Figure S5, Supporting Information). Taken together, these results showed that the amount of released molecule was directly proportional to the amount of loaded molecule when VFD was used for loading method. More specifically, the amount of loaded molecules was calculated by multiplying the volume and molecule concentration of the loading solution. This was because the molecules in the loading solutions were completely absorbed into microparticles or evenly coated on their surface after solvent removal. However, not all impregnated molecules were dissolved when the chemical-laden particles were in the solution. Nevertheless, the releasing ratio is a constant if the system (particle types and chemical types) is determined. The releasing ratios of the five model chemicals are represented in Table 1. From this relationship, we can precisely calculate the required loading amount of drugs for the targeted amount of drug release. Interestingly, this linear proportionality between loading and release amount is also observed in drug loading using solvent evaporation.[23, 25] This is likely because freeze-drying is basically identical to solvent evaporation in the aspect that drug molecules are loaded into microparticles by solvent removal. In our experiment, ≈0.1 mg of the drug was loaded into 10 mg microparticles. To the best of our knowledge, this level of drug-loading amount has not been realized before with simple absorption. This high loading capacity has much potential to be applied to the in vitro bioassay platform using microparticle-based drug delivery.

Table 1. Releasing ratio and intrinsic characteristics of each drug and chemical
Drug nameDOXERLRhoTemozolomide (TMZ)BIBW2992
  • a)

    All solubility data are from Aldrich, except those of

  • b)

    whose solubility information was obtained from Santa Cruz Biotech

  • c)

    Releasing ratios were calculated from loading-release curves under two conditions: the first from various loading concentrations with the same loading volume (conc. var.) and the second from various loading volumes with the same loading concentration (vol. var.).

Molecular weight579.98429.90479.01194.15485.94
Solubility to watera) [mg mL−1]29<11b)5<1
Releasing ratioc) [conc. var./vol. var.]96.3%/93.1%74.5%/73.7%99.6%/93.4%80.5%/85.7%54.0%/54.0%
Figure 3.

Relationship between loading amount and released amount of drug with drug loading by VFD method (see also Figure S5, Supporting Information). Doxorubicin, erlotinib, and rhodamine-B were used as the model molecules. a) The relationship between the concentration of loading solution and the concentration of releasing solution when the same volume (20 µL) of the chemical solution is loaded. Molecule impregnated in microparticles was released into 1.5 mL phosphate-buffered saline (PBS) solution (n = 3). b) The relationship between the volume of loading solution and concentration of releasing solution when the same concentration of loading solution is used (n = 3). Loading process was performed via VFD with 1 × 10−3 m loading solution, except for rhodamine-B (0.5 × 10−3 m). All data showed that the released amount was directly proportional to the loading amount. The proportional constant is represented in Table 1. Abbreviations: doxorubicin hydrochloride, erlotinib hydrochloride, and rhodamine-B are DOX, ERL, and Rho, respectively.

Finally, to verify the versatility of the VFD method to various types of microparticles, we applied the method to four types of microparticles including hydrophilic, porous hydrophilic, hydrophobic, and porous hydrophobic microparticles (Figure 4). Loading uniformities of VFD method and solvent evaporation were compared in order to examine the merit of VFD method. Rhodamine-B was used as a model chemical, and different shapes were assigned to each of the different types of microparticles (for easy identification). For all the types of microparticles, VFD method showed smaller CV than the solvent evaporation method. Especially, porous hydrophobic microparticles showed higher uniformity than the other types of microparticles. Experimental results with the same circular shape of the microparticles showed similar results (data not shown). We suppose that this was because porous hydrophobic microparticles had the lowest degree of agglomeration in DMSO solution, and thus, the high uniformity was because of the well-dispersed microparticles in the frozen solution during the freeze-drying process. However, because different types of microparticles showed different release profiles (Figure S3, Supporting Information) and different releasing ratio, the choice of microparticle type required for a specific application is important.

Figure 4.

Versatility of drug loading via VFD in various kinds of microparticles. Different shapes were assigned to each type of microparticles. VFD method showed less CV values than solvent evaporation method, and those values were less than 10% for all types of microparticles. CV values are represented at the top-right of the images (n = 30, Scale bars: 200 µm).

3 Conclusion

In summary, we demonstrated that freeze-drying is a simple and productive method for drug loading into prefabricated microparticles. This method enables a high amount of drug to be loaded with high uniformity, and even poorly water-soluble drugs could be loaded using an organic solvent. We also proved that this method could be utilized for various types of microparticles. Drug impregnation uniformity was significantly enhanced by freeze-drying than it was by simple solvent evaporation and was further improved by vortexing during the freezing process. This was because freezing the solvent blocks capillary flow that induces nonuniformity, and vortexing during the temperature drop facilitated more even microparticle dispersion in the frozen solution. When we used the VFD method for drug loading, the delivered amount of the drug was directly proportional to the loading amount, and thus the dose of the drug could be controlled precisely. We expect that the method of freeze-drying for uniform drug loading into microparticles will contribute to great advances in the microparticle-based DDS field, especially as an experimental platform using microparticles as drug carriers in small reaction volumes.

Moreover, we showed that freeze-drying could successfully suppress the coffee-ring effect. The coffee-ring effect profoundly affects the solvent evaporation-based coating applications, such as thin-film coating and DNA/protein microarray. Owing to its simplicity, the proposed method can be used by any individual researcher without requiring a complicated protocol. We therefore expect that this approach can make a great impact in coating applications using solvent evaporation.

4 Experimental Section

Microparticle Fabrication: Unless otherwise stated, a 19:1 volume ratio of poly(ethylene glycol) diacrylate (PEGDA, Mn ≈ 700, Aldrich) and photoinitiator (2-hydroxy-e-methylpropiophenone 97%, Aldrich) was used as photocurable resin to fabricate hydrogel microparticles. The prepolymer resin was dispensed between a polydimethylsiloxane-coated chrome mask and a slide glass with 50 µm spacer that determined the thickness of the microparticles. The polymer resin was cross-linked using 20 mJ cm−2 UV irradiation for 5 s (UV lamp SHMA 350S1 was used). After polymerization, microparticles were gathered in a 1.5 mL test tube, washed two times with ethanol, and dried in a 60 °C oven. In a single mask, 15 225 microparticles were fabricated. For the experiment depicted in Figure 4, different compositions of photocurable resin were used. To make porous hydrophilic microparticles, photocurable resin composed of 65% (v/v) PEGDA 700, 30% (v/v) PEG 400 (Aldrich), and 5% (v/v) of photoinitiator were used. PEG 400 molecules, which were used as porogen,[38] did not crosslink with the surrounding matrix, thus uncrosslinked molecules were rinsed out during the washing step. For the hydrophobic microparticles, PEGDA 250 (Aldrich) was used instead of PEGDA 700.

Preparing Chemical-Laden Microparticles: Rhodamine-B and temozolomide were purchased from Aldrich while erlotinib hydrochloride and BIBW2992 (Afatinib) were purchased from Servlab and Santa Cruz Biotech, respectively. All chemicals were dissolved in DMSO. Then, the solution was added into the test tube filled with prefabricated microparticles. After that, the particle and solution were dried by three methods, simple vacuum drying, freeze-drying, and VFD. Speed vacuum dryer (model 4080C, Hanil Scientific) was used for the simple vacuum drying (solvent evaporation) and a freeze dryer (laboratory FD 5508, IlshinBioBase) was used for freeze-drying and VFD. For the VFD, mixing block (MB-102, BIOER) was used to make the vortexing motion during the temperature drop.

Drug Concentration Measurement: Drug concentration was measured using an ultraviolet-visible spectrophotometer (UV-1800, Shimadzu). First, peak wavelength of absorbance spectrum was measured for each drug. After determining the peak wavelength, the reference curve of the relationship between drug concentration and peak absorbance was obtained from samples of known concentration. Then concentration of unknown sample was determined from its peak absorbance and reference curve.

Microscope Image Acquisition: Fluorescence images were obtained using an inverted microscope setup (Nikon Eclipse Ti) and Charge-Coupled Device (CCD) camera (Nikon Digital Sight DS-Ri1). Nikon C-LHGFI HG LAMP was used as a fluorescence light source and Tetramethylrhodamine (TRITC) channel filter (Excitation 540 nm, Emission 605 nm) was used for fluorescence image acquisition. Two types of objective lens (Nikon 4× plan fluor, Numeriacl Aperture (NA) = 0.1, Working Distance (WD) = 172 000 µm per Nikon 10× plan fluor Differential Interference Contrast (DIC) L N1, NA = 0.3, WD = 160 000 µm) were used depending on proper magnification. SEM images were obtained using a SUPRA 55VP. All the SEM images were obtained at 300× magnification with following condition (WD = 3.2 mm, Extra High Tension (EHT) = 2.00 kV, Contrast = 23.5%, Brightness = 50.3%). Bright field microscope images in Figure 1 were obtained using Dino-Lite Digital Microscope.

Fluorescence Intensity Analysis: Fluorescence intensity of circular particles was analyzed using customized MATLAB codes. In brief, each circular particle was recognized from fluorescence images by using a circle detection algorithm, and particles that overlapped or cut at the image boundary were excluded using interparticle distance and center-image boundary distance. After filtering, average pixel intensity inside each particle area was recorded. For the noncircular microparticles, the fluorescence intensity of each microparticle was manually measured using ImageJ software.


The authors acknowledge useful discussions with Amos C. Lee and G. Y. Lee. The microparticles were kindly provided from R. H. Jung. This work was supported by a National Research Foundation of Korea (NRF) funded by the Ministry of Science, ICT and Future Planning (2012M3A7A9671610, NRF-2012-0009555, and 2015K1A4A3047345) and by the Industrial Technology Innovation Program (10050991) funded by the Ministry of Trade, Industry and Energy (MI, Korea). This work was also supported by a grant of the Korean Health Technology R&D Project, Ministry of Health and Welfare, Republic of Korea (HI13C2162).