116
Bioconjugate Chem. 2002, 13, 116−121
High Throughput Magnetic Resonance Imaging for Evaluating Targeted Nanoparticle Probes Dagmar Ho¨gemann, Vasilis Ntziachristos,§ Lee Josephson,§ and Ralph Weissleder* Center for Molecular Imaging Research, Massachusetts General Hospital, Charlestown, Massachusetts 02129. Received August 21, 2001; Revised Manuscript Received October 26, 2001
The ability to image specific molecular targets in vivo would have significant impact in allowing earlier disease detection and in tailoring molecular therapies. One of the rate-limiting steps in the development of novel compounds as reporter probes has been the lack of cell-based, biologically relevant, high throughput screening methods. Here we describe the development and validation of magnetic resonance imaging (MRI) as a technique to rapidly screen compounds that are potential MR reporter agents for their interaction with specific cellular targets. We show that MR imaging can (1) evaluate thousands of samples simultaneously and rapidly, (2) provide exceedingly accurate measurements, and (3) provide receptor binding/internalization data as validated by radioactive assays. The technique allows the screening of libraries of peptide-nanoparticle conjugates against target cells and the identification of conjugates that may be subsequently used as reporter agents in vivo. The technology should greatly accelerate the development of target-specific or cell-specific MR contrast agents.
INTRODUCTION
The sequencing of the human genome has led to a large number of potential targets for drug development and necessitated ever more efficient strategies for compound synthesis and compound evaluation. Current therapeutic drugs address roughly 500 different molecular targets (45% receptors, 30% enzymes, 25% other targets). Estimates of potential future drug targets are in excess of 5000 to 10000. The large number of potential targets can be contrasted with the handful of specific imaging agents available clinically. One of the clearly identified bottlenecks in drug development generally, and targeted imaging agents in particular, has been the development of efficacy testing methodologies that are both highly relevant to the intended use of the drug and amenable to high throughput screening. The goal of this study was to develop and validate magnetic resonance imaging (MRI) as a high throughput screening (HTS) modality for examining the interactions between superparamagnetic nanoparticles and cells in the wells of microtiter plates. To accomplish this we used a recently developed technology that allows the attachment of a variety of biological molecules to iron oxide nanoparticles with the retention of biological activity. Proteins such as transferrin (1), peptides such as the membrane translocating tat peptide of the HIV tat protein (2, 3), and oligonucleotides of various sequences (4) have been attached to an aminated, cross-linked iron oxide nanoparticle (amino-CLIO) and used in a variety of applications. We first evaluated MR imaging as an analytical method by examining the reproducibility and detectability of nanoparticles in microtiter plates using a clinical * Corresponding author: Ralph Weissleder, MD, Ph.D.,Center for Molecular Imaging Research, Massachusetts General Hospital, Bldg. 149 13th Street, #5406, Charlestown, MA 02129. Tel.: (617) 726 5788. Fax: (617) 726 5708. E-mail: weissleder@ helix.mgh.harvard.edu. § Contributed equally.
MR imaging system. Afterward we used a transferrinnanoparticle conjugate to validate the method for determining nanoparticle binding and uptake by cells expressing differing levels of the transferrin receptor. Finally, we synthesized a limited library of peptide-nanoparticle conjugates and screened the library using the MR imaging method. Use of the MR imager permitted up to 1920 samples to be imaged in about 50 min acquiring more than 38000 individual data points in multiecho trains. Using this approach it appears feasible to analyze upward of 20000 experimental samples per day, a number that is currently only limited by the design and availability of better magnetic coils and stronger gradients to achieve higher spatial resolutions. Imaging of large numbers of biological samples in HTS fashion with a clinical MR imager may accelerate the discovery of new targeted MR contrast agents. MATERIALS AND METHODS
Synthesis of Amino-CLIO Nanoparticles. Monocrystalline iron oxide nanoparticles were prepared through the reaction of iron salts with ammonium hydroxide in the presence of T10 dextrans. The dextrans were then modified in the presence of epichlorohydrin according to previous reports to yield cross-linked iron oxide nanoparticles (CLIO-10). CLIO was finally treated with ammonia to yield amino-CLIO, to obtain reactive groups for the covalent binding of bifunctional linking reagents (2, 4). Peptide Synthesis. Peptides were synthesized for attachment to the aminated nanoparticle, amino-CLIO. The peptides included the tat peptide of HIV, a wellestablished membrane translocation signal (GRKKRRQRRRGYK(FITC)C-NH2), short poly-arginyl peptides which might act as membrane translocating signals (5, 6), and control peptides. Peptides were synthesized on an automatic synthesizer (PS3, Rainin, Woburn, MA) by Fmoc chemistry and underwent reaction with fluorescein isothiocyanate (FITC) to facilitate the determination of the
10.1021/bc015549h CCC: $22.00 © 2002 American Chemical Society Published on Web 01/16/2002
High Throughput MRI Evaluating Nanoparticle Probes
number of peptides attached per nanoparticle (2). Peptides were purified by C18 reversed-phase HPLC and characterized by MALDI-MS. The sequences of the peptides used and values of their molecular weights, as M + 1 (calculated/found), were RRK(FITC)C-NH2 (951.08/951.47), GRRK(FITC)C-NH2 (1008.13/1007.97), GRRRRK(FITC)C-NH2 (1320.51/1321.77), GRRGRRK(FITC)C-NH2 (1377.54/1377.55), GRRRRGGGGK(FITC)CNH2 (1547.69/1548.66), GRKKRRQRRRGYK(FITC)CNH2 (2237.54/2238.54), CGGGPQRYGNQWAVGHLMNH2 (1931.13/1391.95), CGVYGWGMGLHGPQRNQANH2 (1931.13/1932.15), and TMG-RRRRRGYK(FITC)GCG-NH2 (1689.77/1690.81). TMG stands for tetramethylguanidium and is a capping group on the N-terminal residue of the peptide. Synthesis of a Limited Library of Peptide-Nanoparticle Conjugates. Peptides were conjugated to aminoCLIO either with long chain N-succinimidyl 3-(2-pyridyldithio)propionate (lc-SPDP, Molecular Biosciences, Boulder, CO) or with succinimidyl iodoacetate (SIA, Molecular Biosciences). For the former, 1 mL of amino-CLIO (9.25 mg iron in 5 mM citrate buffer, pH 8), 0.67 mL of PBS (pH 8), and 1.3 mL of 3.4 × 10-2 M (in DMSO) lc-SPDP were incubated for 2 h at room temperature to yield activated CLIO particles. Low molecular weight impurities were removed by filtration through Sephadex G-25 columns equilibrated with PBS, pH 8. Peptides and activated CLIO were mixed at equal molar ratios of peptides and 2-pyridyl disulfide groups in PBS, pH 8, and allowed to react for 18 h at room temperature. Nonreacted peptides were separated from peptide-nanoparticles using magnetic columns (Macs separation columns, Miltenyi Biotec, CA). For the latter, 1.5 mL of amino-CLIO (39 mg iron in 5 mM citrate buffer, pH 8) were reacted with 3 mL of 4.9 × 10-2 M (in DMSO) SIA for 2 h at room temperature. The mixture was separated over Sephadex G25, and the void volume with iron oxide particles was collected. The CLIO particles were mixed with 13 mg of GRKKRRQRRRGYK(FITC)C-NH2 dissolved in 3 mL of 0.02 M citrate, pH 6.5, and allowed to stand for 3 h at room temperature. The mixture was then applied to the column above to remove nonreacted peptide. Peptide attachment was quantified from absorption of the conjugate (extinction coefficient of 73000 M-1 cm-1 at 494 nm in 0.1 M phosphate buffer). For non-FITC labeled peptide conjugates, peptide was released by DTT and attachment was determined by HPLC. Standards of the coupled peptide were used whose mass was determined by amino acid analysis. Conjugates had between 2 and 12 mol of peptide per mole of nanoparticles based on 2064 iron atoms per crystal (7). Conjugates are denoted as (peptide sequence)n-SS-CLIO or (peptide sequence)n-SC-CLIO where the letters SS or SC indicate disulfide or thioether linking groups, respectively, and n is the number of peptides coupled per particle. Synthesis of Transferrin-Nanoparticle Conjugate. Transferrin (Tf) was attached to CLIO-NH2 according to a protocol optimized in an earlier study (1). The linker in the current study was SIA (Molecular Biosciences). In brief, 1 mL of amino-CLIO (9.25 mg iron in 5 mM citrate buffer, pH 8), 0.67 mL of PBS (pH 8), and 1.3 mL of 3.4 × 10-2 M (in DMSO) SIA were incubated for 2 h at room temperature to yield activated CLIO particles. Human diferric transferrin was activated through thiolation of primary amine groups (1). Tf-SH and iodoacetylated CLIO were mixed at equal molar ratios of transferrin and iodoacetate groups in PBS, pH 8. The mixture was incubated for 18 h at room temperature. Nonreacted Tf-SH was separated from Tf-nano-
Bioconjugate Chem., Vol. 13, No. 1, 2002 117
particles using Macs separation columns (Miltenyi Biotec). On average each superparamagnetic particle contained one transferrin molecule covalently attached. The transferrin-nanoparticle conjugate is denoted as TfSC-CLIO. Magnetic Resonance Imaging. A 1.5 T superconducting magnet (Signa 5.0; GE Medical Systems, Milwaukee, WI) was used. A standard quadrature extremity coil was chosen to obtain a homogeneous field of view over the well plates. The imaging protocol consisted of a T2-weighted spin-echo sequence (TR 3000 ms, variable TE 15-1200 ms; total of 20-24 echo times). The 1.5 mm imaging slices were carefully placed to avoid partial volume effects. At a field of view (FOV) of 12 × 9 cm and a 256 × 192 imaging matrix, each voxel had a volume of 0.33 mm3, and image acquisition time was 10.12 min per number of excitation (NEX). Data Analysis. The MR images obtained at various echo times (TE’s) were processed to obtain T2 maps. To automate the calculations a script was written in Matlab (Matlab 6.0; The MathWorks, Natick, MA). In this routine, signal intensities of each voxel were fitted for exponential decay according to the following equation: SI ) A × e(-TE/T2) + B; with SI ) signal intensity, TE ) echo time, A ) amplitude, B ) offset. Following this calculation, the T2 maps underwent semiautomatic segmentation of the areas of the different wells (IPLab 3.5.2; Scanalytics, Fairfax, VA). The first step was to employ a median filter to change each pixel value to the median value within a neighborhood of 3 × 3 pixels. This filter reduced the noise from isolated pixels without affecting the average T2 values. All regions of interest were simultaneously defined through a one step thresholding function with the threshold set to a value just above the background of the plastic material of the plate. Applying an erode function that computes the minimum value of the pixels within the neighborhood and assigns it to the target pixel made sure that indeed no voxels outside the wells were included into the measurements. The average values of the T2 relaxation times were measured automatically, and the data was exported to a spreadsheet. Evaluation of Levels of Transferrin Receptor Expression in Different Cell Lines by MR Imaging. 9L gliosarcoma clones, differing in expression of the transferrin receptor (8), were cultivated in Dulbecco’s modified Eagle’s medium (DMEM, Cellgro, Mediatech, Washington, DC) containing 10% fetal bovine serum (FBS, Cellgro) and were used to validate MR measurements of receptor binding and internalization of transferrin-CLIO. One hundred thousand cells/well of the different clones were grown in 24 well plates (Falcon, Becton Dickinson, Lincoln Park, NJ). The following day the binding and uptake of radioactive labeled transferrin and of Tf-SC-CLIO conjugate were determined. In the radioactive assay different concentrations of unlabeled diferric transferrin (1 nM to 10 µM in DMEM, 10% FBS) were added immediately prior to the addition of 125Itransferrin (0.25 nM in DMEM, 10% FBS) (1). After incubation for 2 h at 37 °C, the cells were washed three times with Hanks’s Balanced Salt Solution (HBSS, BioWhittaker, Walkersville, MD) and detached with trypsin (0.05% trypsin, 0.53 mM ethylenediaminetetraacetic acid). Cell-associated radioactivity was counted in a gamma-counter (1282 Compugamma CS, LKB Wallac, Sweden). In the MRI assay cells were incubated in the presence of different concentrations of Tf-SC-CLIO (0.0181.8 mM iron). To control for receptor specificity, cells were also incubated with nonconjugated CLIO (7.2 mM iron)
118 Bioconjugate Chem., Vol. 13, No. 1, 2002
Ho¨gemann et al.
Figure 1. Use of a MR imager for high throughput screening. Microtiter plates were imaged in a quadrature coil with different pulse sequences (variable TE) to yield a series of images for each plate. Signal intensity data were processed to yield a T2 map indicating wells with “hits”.
as well as with Tf-SC-CLIO (7.2 mM iron) in the presence of a 100-fold excess of nonconjugated transferrin. Nonspecific binding to the wells was ruled out through the incubation of the culture plates with Tf-SC-CLIO (7.2 mM iron) in the absence of cells. After incubation at 37 °C for 2 h, cells were washed three times with Hanks’s Balanced Salt Solution (HBSS, Bio-Whittaker, Walkersville, MD). Cells were lysed in 60 µL of PBS, pH 8, containing 1% Triton X-100 and an aliquot of 5 µL was used to determine the number of cells per well by protein analysis (BCA, Pierce, Rockford, IL). Lysates were transferred to the wells of 384 well plates and centrifuged at 700 rpm for 2 min (Sorvall 7 RT, Kendro Laboratory Products, Newtown, CT) to remove air bubbles. MR imaging and the determination of the T2 values were carried out as described above. The quantification of cellassociated iron concentration was determined from T2 values of nanoparticle standards (1.46 to 29.2 µM iron). Evaluation of a Limited Library of PeptideNanoparticles Conjugates for Cell Uptake by MR Imaging. Peptide-nanoparticle conjugates were evaluated for their cell-association and ability to serve as reagents for cell labeling and tracking (3). 9L gliosarcoma cells were cultivated as described above. One hundred thousand cells/well were grown in 24 well plates. The following day 9L gliosarcoma cells were incubated with the various peptide-CLIO conjugates at iron concentrations ranging from 0.0018 to 1.8 mM. After the 1 h incubation at 37 °C, the cells were washed three times with HBSS and lysed in 60 µL of PBS, pH 8, containing 1% Triton X-100. Lysates were transferred to the wells of 384 well plates and centrifuged at 700 rpm for 2 min. MR imaging and the determination of the T2 values were carried out as described above. RESULTS
We performed three types of experiments to evaluate MR imaging as a HTS method for samples in microtiter plates. They were (a) analysis of signal intensity variation, (b) analysis of T2 variation, and (c) validation of ligand/receptor interaction in a cell based system expressing different levels of receptor. Finally, with the principles of the method established, we used the method to analyze the components of a limited superparamagnetic particle library for uptake into cells. We evaluated
Table 1. Reproducibility of Signal Intensity Determination at Different Nanoparticle Concentrations (384 well microtiter plate, 16 × 20 Wells, TR/TE ) 3000/100) well no.
0.009 mM
0.018 mM
0.09 mM
0.18 mM
1 2 3 4 5 . 320 mean SD SEM 99% 95%
1125.03 1094.76 1116.99 1087.80 1082.16 . 1375.01 1106.16 79.00 4.42 11.38 8.66
921.20 922.88 946.24 945.53 1007.67 . 1078.91 952.42 91.46 5.11 13.17 10.02
322.47 314.25 284.98 264.03 254.20 . 399.06 391.41 40.13 2.24 5.78 4.40
163.56 162.21 159.03 143.14 137.18 . 130.82 154.11 10.03 0.56 1.44 1.10
two different types of microtiter plates: 384 (60 µL volume per well) and 1536 (10 µL) well plates. All data were obtained with the 384 well configuration except where noted. The general procedure for the MR imaging of iron oxide nanoparticles in microtiter plates is shown in Figure 1. Analysis of Signal Intensity and T2 Variation. Before utilizing the microtiter plates to screen for nanoparticle interactions with cells by MR imaging, we tested the reproducibility of the measurements of signal intensities and T2 values. Signal intensities were determined for wells over the entire plate at four nanoparticle concentrations. At an echo time of 100 ms an area of 16 × 20 wells provided a 99% confidence interval with individual wells being within 1-2% of the mean signal intensity at all nanoparticle concentrations (Table 1). The reproducibility of the technique was unaffected by stacking plates up to six high during signal acquisition. This enabled data to be obtained simultaneously on 1920 samples of six 384 well plates. To quantify the changes in signal intensities depending on the different concentrations of nanoparticles, we calculated the changes of the transverse relaxation times (T2). The reproducibility with which T2 could be determined was analyzed for wells over the entire plate (Figure 2A), and the chi-square of the exponential decay curve fit of signal intensities against echo times is shown in Figure 2B. The reproducibility of T2 values determined for 320 wells is shown in Table 2. Again the 99%
High Throughput MRI Evaluating Nanoparticle Probes
Bioconjugate Chem., Vol. 13, No. 1, 2002 119
Figure 2. Uniformity of T2 values determined by imaging 384 well microtiter plates. A. Images of plates contained four nanoparticle concentrations. T2 is displayed as a color map from yellow (low nanoparticle concentration, high T2) to green (high nanoparticle concentration, low T2). B. The accuracy of the fitting procedure as a function of plate geometry. Signal intensities were plotted against echo times and T2 values were calculated by curve fitting for exponential decay. The chi-square of each respective curve fit is displayed as a color map from blue (low value of chi-square) to red (high value of chi-square). Table 2. Reproducibility of T2 Determination at Different Nanoparticle Concentrations (384 well microtiter plate, 16 × 20 Wells, TR/TE ) 3000/15-1200) well no.
0.009 mM
0.018 mM
0.09 mM
0.18 mM
1 2 3 4 5 . 320 mean SD SEM 99% 95%
485.39 494.61 457.06 459.51 434.51 . 478.63 459.0322 38.4995 2.15 5.54 4.22
208.59 209.93 207.42 206.16 205.29 . 208.43 201.46 12.59 0.70 1.81 1.38
54.09 53.91 56.85 56.68 56.78 . 57.17 55.50 2.94 0.16 0.42 0.32
30.41 29.26 29.67 29.80 30.20 . 30.70 30.04 0.67 .038 0.10 0.07
confidence interval for individual T2 measurements was within 1-2% of the mean. The reproducibility of T2 determinations was further evaluated over successively smaller areas (fields of view, FOV) for a 384 well plate as shown in Table 3. This experiment, performed to determine the “sweet spot” over which variations would be smallest, evaluated FOV’s ranging from 8066 mm2 (384 wells) to 672 mm2 (32 wells). The standard deviation ranged from a high of 13.56 ms (mean 233.71 ms), when 384 wells were imaged, to 9.95 ms (mean 235.88 ms), when only the central 32 wells were imaged. Reproducibility of T2 determination is a function of FOV, though values from the entire plate should be sufficiently accurate for most compound or cell line screening experiments. The data acquisition of 1536 well plates led to similar results as the 384 well format, but required a higher spatial resolution (matrix: 512 × 512) and therefore a two times longer imaging time. The ability of the imaging method to detect nanoparticles is shown in Figure 3. Figure 3A shows the progressive decrease in T2 values as a colorized map for duplicate samples as the nanoparticle concentration was increased from 0 to 29.2 µM iron. Values of T2 were plotted against iron concentrations as shown in Figure 3B. The relaxivity, R2 was 188.5 [mM × s] -1 with a high coefficient of correlation, r2 ) 0.999. Nanoparticle concentrations in the low micromolar range of iron could be
distinguished from zero with visual or graphical analyses. Since there are 2064 iron atoms per particle (7), iron oxide particle concentrations can be readily detected in the low nanomolar concentration range. Determining Transferrin Receptor Expression Using a Transferrin-Nanoparticle Conjugate. The expression of the transferrin receptor has been used as a marker gene for imaging transgene expression (1, 8). Five 9L gliosarcoma cell lines differing in transferrin receptor expression were incubated either with the nanoparticle conjugate Tf-SC-CLIO or with 125I labeled Tf under identical conditions. As shown in Figure 4A, which summarizes the imaging results as a colorized T2 map, cell lines were evaluated in duplicate wells after incubation with Tf-SC-CLIO at concentrations ranging from 0.018 mM to 1.8 mM iron. Figure 4B compares the T2 values for duplicate samples of the five cell lines. Three data sets from separate experiments underwent t-test analysis and representative p values are shown. There was no statistically significant difference in T2 between the duplicates of each sample, however, significant differences in transferrin receptor expression were identified among different clones as indicated. ANOVA analysis confirmed the differentiation of five clones by MRI (p < 0.001). The correlation between nanoparticle uptake determined by MR imaging and the cell-associated radioactivity determined with 125I-Tf was examined as shown in Figure 4C. Cell-associated iron was determined from T2 (Figure 4B) using a standard curve of nanoparticles (Figure 3B). The binding and uptake of Tf-SC-CLIO paralleled the binding and uptake of 125I-Tf with a high coefficient of correlation, r2 ) 0.996. Screening a Limited Library of Peptide-Nanoparticle Conjugates for Cell Binding/Uptake. We then employed the HTS MR imaging technique to evaluate a series of peptide-CLIO conjugates for their binding and/or uptake by 9L gliosarcoma cells as shown in Figure 5. Conjugates were evaluated for their ability to become cell-associated and thereby serve as reagents for labeling cells for tracking by MR. Magnetic cell labeling has become an essential tool for tracking stem and progenitor cells in vivo and for isolating homed cells from tissues
120 Bioconjugate Chem., Vol. 13, No. 1, 2002
Ho¨gemann et al.
Table 3. Reproducibility of T2 Determination as a Function of Area (field of view, FOV) at 0.018 mM Iron FOV (mm2)
8066
7394
6722
6050
5377
4705
4033
3361
2689
2017
1344
672
384 plate mean SD SEM 99% 95% 1536 platea
384 233.7 13.56 0.69 1.78 1.36 1536
352 234.8 12.85 0.68 1.76 1.34 1408
320 234.1 12.79 0.71 1.84 1.40 1280
288 235.2 11.51 0.68 1.75 1.33 1152
256 235.7 10.96 0.68 1.76 1.34 1024
224 236.6 10.46 0.70 1.80 1.37 896
192 236.5 10.33 0.75 1.92 1.46 768
160 236.5 10.24 0.81 1.41 1.07 640
128 236.7 10.1 0.89 2.30 1.75 512
96 236.4 9.2 0.94 2.42 1.84 384
64 235.9 9.64 1.21 3.10 2.36 256
32 235.9 9.95 1.76 4.53 3.45 128
a
Number of wells of 1536 plate in the FOV used for imaging a 384 well plate.
Figure 3. Detection of nanoparticles by MR imaging. A. Duplicate wells with increasing concentrations of nanoparticles were imaged by MR imaging, and the T2 maps were calculated. B. 1/T2 values were plotted versus the iron concentrations to determine the relaxivity R2.
Figure 4. Imaging of transferrin receptor expression. A. Uptake of the nanoparticle conjugate Tf-SC-CLIO by clones 1-5 expressing different levels of the transferrin receptor. Cells were incubated with 0.018 to 1.8 mM iron in Tf-SC-CLIO. Data are expressed as a T2 color map. B. Comparison of T2 values of duplicate wells of clones 1-5 at 1.8 mM Tf-SC-CLIO from Figure 4A. T2’s of pairs of clones were evaluated for significance of differences. C. Correlation of receptor expression determined by cell-associated Tf-SC-CLIO (T2 values Figure 4B) or cell-associated 125I-Tf.
(3). The highest cellular uptake was observed for the conjugate (GRKKRRQRRRGYKC-N)11-SC-CLIO which led to the lowest T2 values after incubation with 1.8 mM iron (G10) and 0.18 mM iron (B9), with T2 changes still detectable at 0.018 mM iron (D7) (Figure 5). (GRKKRRQRRRGYKC-N)5-SS-CLIO (1.8 mM iron, D13 and 0.18 mM iron, D5) was the second best conjugate in terms of cell labeling. Finally, (TMG-RRRRRGYKGCG-N)5-SSCLIO also showed some minor efficacy (1.8 mM iron, D12 and 0.18 mM iron, F5) (Figure 5). DISCUSSION
We used a clinical MR imaging system (1.5 T, 64 MHz) widely available in medical institutions and a conven-
tional send-and-receive quadrature extremity coil to image superparamagnetic nanoparticles in microtiter plates. A major question regarding the use of MR imaging in this fashion is the reproducibility of the data obtained. An imaging protocol where the FOV was
