Purpose: To develop and validate an optical imaging nanoprobe for the discrimination of epidermal growth factor (EGF) receptor (EGFR)–overexpressing tumors from surrounding normal tissues that also expresses EGFR.
Experimental Design: Near-infrared (NIR) quantum dots (QD) were coupled to EGF using thiol-maleimide conjugation to create EGF-QD nanoprobes. In vitro binding affinity of these nanoprobes and unconjugated QDs was evaluated in a panel of cell lines, with and without anti-EGFR antibody pretreatment. Serial optical imaging of HCT116 xenograft tumors was done after systemic injection of QD and EGF-QD.
Results: EGF-QD showed EGFR-specific binding in vitro. In vivo imaging showed three distinct phases, tumor influx (∼3 min), clearance (∼60 min), and accumulation (1-6 h), of EGF-QD nanoprobes. Both QD and EGF-QD showed comparable nonspecific rapid tumor influx and clearance followed by attainment of an apparent dynamic equilibrium at ∼60 min. Subsequently (1-6 h), whereas QD concentration gradually decreased in tumors, EGF-QDs progressively accumulated in tumors. On delayed imaging at 24 h, tumor fluorescence decreased to near-baseline levels for both QD and EGF-QD. Ex vivo whole-organ fluorescence, tissue homogenate fluorescence, and confocal microscopic analyses confirmed tumor-specific accumulation of EGF-QD at 4 h. Immunofluorescence images showed diffuse colocalization of EGF-QD fluorescence within EGFR-expressing tumor parenchyma compared with patchy perivascular sequestration of QD.
Conclusion: These results represent the first pharmacokinetic characterization of a robust EGFR imaging nanoprobe. The measurable contrast enhancement of tumors 4 h after systemic administration of EGF-QD and its subsequent normalization at 24 h imply that this nanoprobe may permit quantifiable and repetitive imaging of EGFR expression.
One of the most promising biological targets for cancer therapy is the epidermal growth factor (EGF) receptor (EGFR), a transmembrane glycoprotein that controls pleiotropic biological phenomena, including proliferation, angiogenesis, tissue invasion, and metastasis (1, 2). Although EGFR is ubiquitously expressed in normal tissues, it is preferentially overexpressed on the surface of many tumors and downstream signaling from this receptor renders them resistant to standard therapies (3, 4). Targeted therapies that selectively inhibit this receptor have found widespread clinical applicability (4) but there are few reliable methods to predict response to therapy or gauge treatment response over time (5). Noninvasive imaging techniques that can discriminate between EGFR-overexpressing tumors and surrounding normal tissues that also express EGFR may facilitate repetitive and quantitative imaging of EGFR during a course of treatment.
Although several studies have been reported on the imaging of EGFR expression, they predominantly use radiolabeled probes (6–11). Alternatively, optical imaging using fluorescent techniques (12–14) offers a convenient means of mapping molecular profiles noninvasively as they are relatively inexpensive and rapid, involve no exposure to ionizing radiation, and provide spatiotemporal resolution with relatively small data sets compared with other conventional imaging modalities (15). Organic fluorophores with emission peaks in the visible and NIR wavelengths have been successfully used for various receptor-targeted optical imaging, including EGFR (16–26). More recently, a comparative study on visible and NIR excitable fluorescent dyes to image EGFR expression has shown the advantages of NIR excitable fluorescent dyes (27). In spite of the promising results, the irreversible photobleaching characteristics of organic fluorescent dyes require faster image acquisition rate with limitations on the number of images acquired. Hence, an alternate optical marker with more stable photophysical characteristics is needed for better imaging.
One such optical marker is the semiconductor quantum dot (QD) that possesses many advantages over organic fluorophores such as (a) size-tunable narrow (20-30 nm) emission spectra over a wide excitation spectral range; (b) relatively longer fluorescence half-lives (5-40 ns) than organic fluorophores (0.5-2 ns); (c) an inherent resistance to photobleaching; and (d) the potential for multiplexing (simultaneous excitation of multiple QDs of different emission wavelengths with a single excitation wavelength; refs. 28–43). In addition, QDs can be readily conjugated to biomolecules such as peptides and nucleic acids for biomedical imaging applications. To date, in vivo applications of QDs have been limited to imaging of tumor vasculature (29, 36, 44, 45), tumor-specific membrane antigens (33, 34), and sentinel lymph nodes (35, 46, 47). However, to our knowledge, no publication exist on the use of peptide-labeled QDs for quantitative in vivo imaging of a ubiquitously present receptor, such as EGFR, that is overexpressed on tumors compared with adjacent normal tissues. Here, we describe the development and characterization of EGF-conjugated NIR (800-nm emission peak) QDs (EGF-QD) as nanoprobes to image EGFR expression in human colon cancer xenografts. These nanoprobes could potentially be used for quantifiable and repetitive imaging of EGFR during and after a therapeutic intervention.
Materials and Methods
Materials. DMEM/Ham's F-12 50/50 mix with l-glutamine, DMEM with l-glutamine, and MEM α 1× with Earn's salt and without ribonucleosides, deoxyribonucleosides, and l-glutamine were purchased from Mediatech, Inc. (Herndon). Fetal bovine serum and penicillin-streptomycin were purchased from Hyclone and Invitrogen Corporation, respectively. The eight-well Lab-Tek II chambers with sterile no. 1.5 borosilicate coverslips were obtained from Nalge Nunc International. PBS powder was purchased from Sigma-Aldrich and constituted in triple-distilled water to get a final pH of 7.4. The 800-nm QD and the antibody conjugation kit were purchased from Molecular Probes (Invitrogen). Human recombinant EGF was purchased from BD Biosciences (Bedford). The anti-EGFR human-mouse chimeric antibody cetuximab was purchased from The University of Texas M.D. Anderson Cancer Center pharmacy. Rabbit monoclonal anti-EGFR antibody was purchased from Millipore (Billerica), and rat monoclonal anti-CD31 antibody was purchased from BD PharMingen. Secondary antibodies used were FITC-conjugated Affinipure donkey anti-rabbit IgG (Jackson Immunoresearch Laboratories, Inc.) and Alexafluor 594 goat anti-rat IgG (Invitrogen). Nonstick microcentrifuge tubes for preparing the QD conjugates were purchased from VWR International. Alfalfa-free diet for the animals was purchased from Dyets, Inc.
Cell lines. Colorectal cancer cell lines HCT116 and DiFi, which have moderate (++) and high levels of EGFR expression (+++), respectively, and the Chinese hamster ovarian cancer cell line CHO K1, which has no EGFR expression (− − −) were used in this study. HCT and CHO K1 cells were obtained from the American Type Culture Collection and DiFi cells were kindly provided by Prof. Lee M. Ellis (Surgical Oncology, University of Texas M.D. Anderson Cancer Center, Houston, TX). The cells were maintained at 37°C in a humidified atmosphere containing 5% CO2 in the appropriate growth medium supplemented with 1% penicillin-streptomycin. HCT116 and CHO K1 cells grow as a monolayer, whereas DiFi cells grow with ∼50% of the cells in suspension.
Tumor xenografts. Six- to 8-week-old immunocompromised male nude mice (Swiss nu/nu; n = 16) weighing 20 to 25 g each were purchased from the specific pathogen-free breeding colony in the Department of Experimental Radiation Oncology, M.D. Anderson Cancer Center. The animals were kept in well-ventilated polypropylene cages with a 12-h light-dark cycle and fed sterilized standard laboratory diet and water ad libitum. Approval from the Institutional Animal Care and Use Committee was obtained for all experimental procedures. Near-confluent HCT116 cells grown in culture flasks were harvested using 0.05% trypsin-EDTA, centrifuged, and resuspended in sterile PBS to get a final cell concentration of ∼2 × 106 cells per 50 μL, which was injected s.c. into the right flank of the mice. After the injection, animals were fed with a special alfalfa-free diet (to minimize the fluorescence interference from the standard laboratory diet), and tumor growth was monitored daily. In vivo imaging experiments were initiated when tumors reached 0.8 to 1 cm in diameter.
EGF conjugation with QD. The conjugation process was done in three steps. In step 1, amine-functionalized CdSeTe/ZnS (core/shell) QDs (∼2.0 μmol/L; 125 μL) with emission maximum at 800 nm were activated by reacting them with 10 mmol/L of the noncleavable and membrane-permeable heterobifunctional cross-linker, 4-(maleimidomethyl)-1-cyclohexanecarboxylic acid N-hydroxysuccinimide ester at room temperature (21°C) for 1 h. The activated QD was eluted with PBS (pH 7.4) through a gel filtration PD-10 desalting column containing Sephadex G-25 (Amersham Biosciences).
In step 2, human recombinant EGF (1 mg/mL; 300 μL) was reacted with 20 mmol/L DTT for 30 min at room temperature to obtain reduced EGF with free sulfhydryl groups. The reduced EGF was purified by elution with PBS (pH 7.4) through a gel-filtration PD-10 column containing Sephadex G-25.
In step 3, activated QD (the end product of step 1) and reduced EGF (the end product of step 2) were reacted for 1 h at room temperature to form the conjugate. The conjugation reaction was quenched by 10 mmol/L β-mercaptoethanol, concentrated by ultracentrifugation for 15 min at 7,000 rpm, and purified by eluting with PBS through a gel filtration PD-10 column containing Superdex-200. Using the molar extinction coefficient [ε550 = 17 × 105 (mol/L)−1 cm−1 and ε277 = 4.12 × 104 (mol/L)−1 cm−1] at 550 and 277 nm, the estimated concentrations of QD and EGF in the final purified EGF-QD nanoprobe were 0.75 and 2.9 μmol/L, respectively, resulting in the multivalency of four EGF molecules per QD.
Binding assay by ELISA. The binding assay was done using a RayBio human EGF ELISA kit (ELH-EGF-001; RayBiotech, Inc.). The assay uses a 96-well plate coated with an antibody specific for human EGF. One hundred microliters of unmodified human recombinant EGF and the reduced EGF (from step 2 in conjugation process) were added to each well and incubated for 2.5 h at room temperature and allowed to bind to the immobilized antibody. The wells were then washed with wash buffer, and 100 μL of biotinylated anti-human EGF antibody were pipetted into the wells and incubated for 1 h at room temperature. Next, the unbound biotinylated anti-human EGF antibody was washed with wash buffer, and 100 μL of horseradish peroxidase–conjugated streptavidin were pipetted into the wells and the plate was incubated at room temperature for 45 min. The wells were again washed with wash buffer, and 100 μL of 3,3′,5,5′-tetramethylbenzidine one-step substrate reagent were added and incubated for 30 min at room temperature in the dark. Blue color developed in proportion to the amount of sample bound to the immobilized antibody coated in the wells. Fifty microliters of stop solution (2 mol/L sulfuric acid) were added to each well to stop the reaction. The stop solution changed the color from blue to yellow, and the intensity of the color was measured at 450 nm. The binding affinity of EGF-QD conjugates could not be tested using this ELISA because of the considerable overlap in absorption spectra of QD and the assay end product. This assay documents the functionality of EGF but does not directly measure EGFR binding. To assess direct binding to EGFR, we did in vitro studies noted below.
In vitro binding studies. Approximately 3,000 HCT116, 10,000 DiFi cells, and 3,000 CHO K1 cells were seeded in each well of an eight-well chambered coverslip and incubated with the appropriate growth medium (400 μL). More DiFi cells than the others were seeded to compensate for the loss of the cells growing in suspension. After 24 to 36 h, the medium was removed and the cells were washed twice with PBS (1×) and then incubated overnight with serum-free medium for 10 to 12 h before the experiment. On the day of the experiment, background fluorescence from the cells was measured using confocal microscope. Immediately after the background measurements were obtained, serum-free medium was removed, and the cells were washed twice with PBS (1×), pulsed with 100 μL of EGF-QD nanoprobe (1 pmol equivalent of QD), and observed again under the confocal microscope for QD fluorescence. To confirm the selective binding of EGF-QD nanoprobe to cell surface receptors, two control experiments were done. In the first control experiment, cells were treated with unconjugated raw QD (1 pmol equivalent of QD). In the second control experiment, HCT116 and DiFi cells were pretreated for 1 h with the anti-EGFR antibody cetuximab (1 μg/mL; 200 μL) diluted in serum-free medium, washed once with PBS (1×), pulsed with 100 μL of EGF-QD (1 pmol equivalent of QD), and observed under the confocal microscope.
In vivo optical imaging.In vivo optical imaging was done using the IVIS imaging system 200 series (Xenogen Corporation). The imaging system consists of a dark chamber with heated stage, gas anesthesia inlet and outlet ports, and scanning laser for alignment. The heated stage can be moved vertically to fix the desired field of view. A 150-W quartz halogen lamp is used as the excitation source, and the emitted signal is collected using a cryogenically cooled (−105°C), back-thinned, back-illuminated grade-1 CCD camera (26 × 26 mm) that is capable of imaging 2,048 × 2,048 pixels. In this study, we set the emission filter at 840 ± 30 nm to get the NIR emission signal from QD nanoparticles and EGF-QD nanoprobes. Although QD has maximum absorption cross-section in the UV-visible region, the excitation filter was set at 640 ± 25 nm to minimize the effective absorption due to the native tissue chromophores below this wavelength range. All the images were acquired with the following variables optimized to improve the signal-to-noise ratio: acquisition time, 20 s; field of view, 6.5 × 6.5 cm; aperture stop, f/8; CCD binning, 4. The signal from the CCD is coupled to a high-performance data acquisition computer containing Xenogen Living Image software with image acquisition controls and image analysis options. The collected fluorescence emission signal was stored in radiance units that refer to photons per second per centimeter squared per steridian (ph/s/cm2/sr). The acquired fluorescence images were pseudocolored.
When the animals were ready for imaging, baseline images were acquired after carefully wiping the animals with alcohol to remove any fluorescing contaminants (from bedding, stool pellets, and food particles). During the image acquisition process, the animals were kept anesthetized with 2% isoflurane, and the heated stage was maintained at 37°C. After the background measurements were made, with the animals still anesthetized, 10 pmol (equivalent of QD) of EGF-QD nanoprobe was injected through the tail vein, and the animals (n = 8) were again imaged at 3, 15, and 30 min and 1, 1.5, 2, 3, 4, 6, 18, and 24 h postinjection. No signs of discomfort were observed during the injection or the entire experiment.
To confirm the specificity and selectivity of the binding of EGF-QD nanoprobe with cellular EGFR, the following two control experiments were done: (a) 10 pmol of unconjugated QD (raw QDs) was injected through the tail vein of tumor-bearing mice (n = 7), and fluorescence images were acquired at the same times as for the experimental animals; and (b) tumor-bearing mice were pretreated with cetuximab (∼1.05 nmol via the tail vein) to block EGFR 24 h after the injection, 10 pmol (equivalent of QD) of EGF-QD nanoprobe was administered (via tail vein), and images were acquired to detect the QD emission signals. Two identically sized circular regions of interest were selected: one overlying the tumor served as the target signal, and the other, in the corresponding shoulder, served as the background signal. The average radiance in the selected region of interest was measured using Living Image software. From the measured average radiance, the tumor-to-background ratio was calculated for each image at each time after EGF-QD nanoprobe and QD nanoparticle injection.
Postacquisition image processing. To extract QD fluorescence from the overlapping autofluorescence and background, the images were processed using Image J image processing software from the NIH3
Euthanasia and tissue collection. Animals were euthanized after the injection of QD (n = 4 at 4 h; n = 3 at 24 h) and EGF-QD (n = 4 at 4 h; n = 4 at 24 h) nanoprobes by overdosing them with CO2. Immediately after euthanasia, the organs were harvested, rinsed with PBS, and imaged using the IVIS imaging system. After image acquisition, the tissues were cut into two pieces. One piece of tissue was transferred to a vial containing 1 mL PBS and frozen until tissue homogenization. The other piece was embedded in a plastic cassette containing optimal cutting temperature medium and slowly cooled over dry ice and methanol. These embedded tissues were stored at −80°C until they were sectioned into 5- to 7-μm-thick slices on microscope slides for observation under the confocal microscope and for immunohistochemical analysis.
Tissue homogenate. The tissues frozen in 1 mL of PBS were thawed and weighed before preparing the homogenates. The tissues were cut into small pieces, and the tissue homogenate was prepared manually using a tissue homogenizer. One milliliter of 10 N NaOH was added to the homogenate to completely digest the cellular components. Of the resulting tissue homogenate from each organ, 100 μL were transferred to a flat-bottomed 96-well plate and imaged with the IVIS imaging system. The average radiance over the selected region of interest was then measured using Living Image software. The QD signal in each organ was estimated on the basis of the weight of each organ.
Immunofluorescence. Frozen tissue slices were fixed in ice-cold acetone and blocked with protein blocking solution (100-400 μL) for 30 min at room temperature. Staining for EGFR and CD31 were done separately on different slides. Slides were incubated with rabbit monoclonal anti-EGFR antibody (1:100) and rat monoclonal anti-CD31 antibody (1:100), washed with PBS (5 min × 3), and incubated for 30 min at room temperature with fluorescence-labeled secondary antibody. FITC-conjugated Affinipure donkey anti-rabbit IgG and Alexafluor 594 goat anti-rat IgG were used as secondary antibodies for visualizing EGFR and CD31, respectively. After incubation with secondary antibody, the sections were covered with a coverslip using an anti-fade fluorescence mounting medium and processed immediately for confocal microscopic imaging.
Confocal microscopy. Laser-scanning confocal microscopy was done using a Fluoview FV1000 confocal microscope (Olympus America, Inc.) with a 60×/1.4 numerical aperture oil immersion objective (confocal aperture, 75 μm; aspect ratio, 1:1; image acquisition size, 512 × 512 pixels; image acquisition speed, 10 μs/pixel). Laser lines from a diode laser (FV5-LD405; Olympus America) at 405 nm and from an argon laser (FV10-COMB; Olympus America) at wavelengths of 458, 488, and 515 nm were used for excitation. The excitation laser beams were passed through a dichroic mirror (DM405/488/543), and the fluorescence emission was collected using a photomultiplier tube (PMT-R7862; Hamamatsu) through a 650-nm barrier filter. Differential interference contrast images were acquired using a second photomultiplier tube (PMT-R7400; Hamamatsu) with the same excitation laser beams. Both photomultiplier tubes were operated at a minimum gain level of 1 (to minimize the electronic noise), with operating voltage levels set at 610 and 115 V, respectively. The laser unit, confocal microscope, and detection units were connected to the computer and controlled using Fluoview software version 1.4 (FV10-ASW1.4, Olympus America), which was also used to perform the postacquisition data processing.
Statistical analysis. Statistical analysis was done using paired and unpaired t tests for comparisons between and within groups, respectively. Statistical significance was established at P < 0.05. Data are presented as the means ± SE.
Potency and binding affinity of EGF-QD nanoprobes. Our scheme for conjugation chemistry outlined in Materials and Methods is shown in Fig. 1A. The potency and binding affinity of the synthesized EGF-QD nanoprobes were shown using an ELISA. The ELISA-based binding assay was done using unmodified EGF (before reduction) and reduced EGF (after step 2 in the conjugation scheme) in the 0.8 to 200 pg/mL concentration range (Fig. 1B). Relative to unmodified EGF, reduced EGF displayed comparable but slightly lower binding affinity for anti-EGF antibody across the dynamic range of this assay, suggesting that the reduction of EGF during conjugation does not interfere substantially with EGF functionality. Western blot analysis of EGFR phosphorylation was done to compare the functional activity of unmodified EGF and reduced EGF (step 2 of nanoprobe synthesis process). Although significant EGFR phosphorylation was induced in cells treated with EGF, cells treated with r-EGF showed minimal phosphorylation of EGFR (see Supplementary Fig. S1). These results suggest that the EGF-QD nanoprobe might bind EGFR with good affinity without activating downstream signaling pathways induced by EGFR phosphorylation, making this a promising imaging agent that does not initiate proliferative responses.
To further validate the specificity of EGF-QD nanoprobes to cellular EGFR, in vitro confocal imaging studies were done on cell lines with varying levels of EGFR expression. Human colorectal cancer HCT116 (EGFR++) and DiFi (EGFR+++) cells expressing EGFR showed cellular fluorescence signals as early as 5 min after incubation with EGF-QD nanoprobes, whereas in the EGFR-negative CHO K1 cells no cellular fluorescence was observed. Furthermore, more intense cellular fluorescence was observed in DiFi cells than in HCT116 cells, which is consistent with the higher levels of EGFR expression in DiFi cells (Fig. 1C). Over a period of 90 min, the fluorescence signal was observed in the cytoplasmic region, which may be attributed to the internalization of EGF-QD nanoprobes (data not shown). The fluorescence intensity peaked at 15 min and remained stable for an additional 90 min. In the negative control study, incubation of HCT116 and DiFi cells with QD nanoparticles alone resulted in no cellular fluorescence. Pretreatment of HCT116 and DiFi cells with cetuximab for 1 h before incubation with EGF-QD nanoprobes abrogated the development of cellular fluorescence. Taken together, these in vitro studies showed the affinity and specificity of EGF-QD nanoprobes for cellular EGFR.
Imaging in vivo pharmacokinetics and tumor targeting with EGF-QD nanoprobes. Upon reaching a tumor size of ∼0.8 to 1 cm in diameter, the animals were administered an i.v. injection of 10 pmol QD nanoparticles (n = 7) and EGF-QD nanoprobes (n = 8) and imaged at various time points. The liver and lymph nodes showed the most prominent fluorescence signal in all animals. Intense QD and EGF-QD fluorescence was observed at the 3-min time point, which decreased rapidly by 1 h. Thereafter, at the 4-h time point, EGF-QD nanoprobes showed more prominent fluorescence when compared with QD nanoparticles. At 24 h, both QD and EGF-QD fluorescence decreased to near-baseline levels. The spectrally unmixed and remixed background and QD fluorescence of a representative animal at different time points after injection with QD and EGF-QD nanoprobes are shown in Fig. 2A to E and F to J, respectively. The spectrally unmixed background and QD images from a representative animal at 4 h post–EGF-QD injection along with the raw fluorescence image and remixed images are shown in Supplementary Fig. S2.
The tumor-to-background signal ratio estimated from regions of interest in tumor and normal tissue at different times after QD and EGF-QD injections are shown in Fig. 2K and L. Three distinct phases of tumor influx, clearance, and equilibration/accumulation of QD nanoparticles and EGF-QD nanoprobes were observed (Fig. 2M). During the first phase (influx), rapid nonspecific enhancement in QD and EGF-QD fluorescence was observed within ∼3 min after injection. The QD nanoparticles exhibited an ∼50% higher peak fluorescence than the EGF-QD nanoprobes. During the second phase (clearance), from 3 to 60 min, an exponential decrease in QD and EGF-QD nanoprobe fluorescence intensity was observed with clearance rate constants (k2) of 1.26 ± 0.38 h−1 and 0.57 ± 0.12 h−1, respectively. The corresponding clearance half lives for QD and EGF-QD nanoprobes were estimated as 0.64 ± 0.15 and 1.40 ± 0.32 h, respectively. After initial rapid influx and clearance from tumor attributable to increased vascular permeability of tumors, at ∼1 h postinjection, fluorescence from both QD and EGF-QD nanoprobes attained an apparent dynamic equilibrium between the vascular and extracellular space/perivascular compartments of the tumor. Subsequently, a steady accumulation of EGF-QD nanoprobes, measured as tumor-to-background ratio, was observed at the rate of 0.24 ± 0.10 h−1. In contrast, QD nanoparticles showed a slow exponential decrease with no progressive accumulation after reaching dynamic equilibrium (Fig. 2M). During the late phase of imaging, the tumor-to-background ratio decreased to near-baseline levels at 24 h for both QD and EGF-QD nanoprobes (Fig. 2K and L). For further characterization of EGF-QD nanoprobes, 4 h postinjection was chosen as a convenient early time point that clearly discriminated EGFR-specific EGF-QD fluorescence from the nonspecific QD fluorescence in tumor tissues (tumor-to-background ratios of 1.93 ± 0.27 and 0.86 ± 0.10, respectively; P < 0.03).
Pretreatment with cetuximab 24 h before EGF-QD injection led to abrogation of tumor-specific fluorescence over the entire 18-h imaging period. The infusion time of 24 h was chosen based on the literature demonstrating 100% and <87% tumor occupancy with cetuximab concentrations of 0.25 and 0.04 mg/injection, respectively (48). As the concentration of cetuximab used in the present study was 0.16 mg/injection, we anticipate tumor occupancy of >90% at 24 h. Representative pretreatment and 4-h postinjection images appear in Supplementary Fig. S3. Taken together, these imaging results show the favorable pharmacokinetics and specificity of EGF-QD nanoprobes for imaging EGFR in tumors.
In situ validation of in vivo imaging results.Ex vivo macroscopic fluorescence images of the organs extracted 4 and 24 h after injection of QD and EGF-QD are shown in Fig. 3A to D, respectively. At both time points, the liver and spleen showed maximum fluorescence intensity from QD and EGF-QD nanoprobes. More intense fluorescence from QD and EGF-QD nanoprobes was observed from the lung tissues at 4 h than at 24 h after injection (Fig. 3A and B). In contrast, QD and EGF-QD fluorescence from the kidneys were higher at 24 h after injection than at 4 h, possibly attributable to delayed renal clearance. QD and EGF-QD fluorescence intensity from tumors was greater at 4 h than at 24 h, with EGF-QD nanoprobes demonstrating greater intensity than QD nanoparticles.
Because ex vivo fluorescence images provide only qualitative information, a semiquantitative analysis of fluorescence emission from tissue homogenates was done. The average fluorescence of QD nanoparticles and EGF-QD nanoprobes per organ at 4 and 24 h after injection are shown in Fig. 3F and G, respectively. The average EGF-QD nanoprobe fluorescence from the tumor tissues at 4 h was significantly higher than that from QD nanoparticles at the same time point (P = 0.0008). No significant differences in average EGF-QD and QD fluorescence from the tumor tissues were observed at 24 h (P = 0.4137). No substantial differences in the average EGF-QD and QD fluorescence from the homogenates of other organs were observed at the 4- and 24-h time points. The higher EGF-QD fluorescence when compared with QD fluorescence in the tumor tissues at 4 h corroborates the data obtained from the in vivo imaging studies, further demonstrating specific binding of EGF-QD nanoprobes to EGFR in tumor tissues.
The selective localization of EGF-QD within each organ at the microscopic level was examined using a confocal laser-scanning microscope. The confocal fluorescence images of frozen tissue sections of organs extracted at 4 and 24 h after injection of QD and EGF-QD nanoprobes are shown in Fig. 4. Liver and spleen showed maximum fluorescence signal among all organs evaluated. Homogeneously distributed bright fluorescence from EGF-QD nanoprobes was observed from tumor tissues at 4 h postinjection. In contrast, nonhomogeneous and relatively weak QD fluorescence was observed from the tumor tissues at 4 h postinjection. Negligible QD and EGF-QD fluorescence was observed from the tumor tissues at 24 h postinjection. These observations are consistent with the in vivo, ex vivo, and tissue homogenate fluorescence measurements.
Last, to further characterize the observed in vivo difference in QD and EGF-QD nanoprobe fluorescence from tumor tissues at 4 h postinjection, tumor specimens were analyzed for EGFR and CD31 (tumor vasculature) by immunofluorescence staining (Fig. 5). EGF-QD nanoprobes dispersed diffusely throughout the tumor and overlapped with widespread areas of EGFR expression as well as patchy areas of vasculature, indicating selective binding of EGF-QD nanoprobes to tumor (white arrows) and, possibly, vascular EGFR (Fig. 5A and B). In sharp contrast to this, QD nanoparticles were localized to the perivascular space without any appreciable overlap with areas of EGFR expression, suggesting that nonspecific extravasation accounted for any fluorescence observed with these nontargeted nanoparticles.
Imaging of tumors expressing a growth factor receptor that is also expressed on adjacent normal tissues, albeit at a lower level than on tumors, is a significant challenge that requires optimization and characterization of the imaging probe for better discrimination of tissues with varying levels of receptor expression. In this study, we report on the development and validation of EGF-QD conjugates as nanoprobes for noninvasive optical imaging of EGFR expression in human colorectal cancer xenografts in mice. To the best of our knowledge, this is the first study that clearly shows the ability of EGF-QD conjugates to image a ubiquitously present receptor that is overexpressed on tumors compared with adjacent normal tissues.
In the present study, the choice of EGF as the preferred molecule for docking to EGFR is based on the following features. First, the binding affinity of the natural ligand (i.e., EGF) of a receptor (i.e., EGFR) generally exceeds that of synthetic antibodies, with varying binding affinities, directed against a given epitope on the receptor surface. Second, in contrast to monoclonal antibodies, Fab antibody fragments, or single-chain Fv antibody fragments, the smaller size of peptides permits increased penetration in solid tumors and more rapid clearance via the kidneys, with less likelihood of sequestration in the liver and spleen (49, 50). Third, the polyvalence effect of multiple peptides conjugated to a probe permits stronger binding to the receptor (51). Last, in the case of EGF-QD nanoprobes, the combined size of the peptide, the linker, and the QD is sufficiently small to allow tumor penetration of the conjugates. Further, the tumor model chosen for testing these EGF-QD nanoprobes is also noteworthy. Human EGF binds to both human EGFR on xenograft tumors and mouse EGFR on normal tissues (52). Therefore, a tumor model using HCT116 tumors (as opposed to highly EGFR overexpressing DiFi tumors) growing on mouse hind limbs is similar to a clinical scenario where tumors with moderately high levels of EGFR expression arise within normal tissues that have lower levels of EGFR expression. Furthermore, given the similarity in biological activities of mouse and human EGF (53), the ability to clearly visualize EGFR overexpression within tumors using a low concentration of EGF-QD nanoprobes confirms that these nanoprobes could compete effectively with endogenous (mouse) EGF.
Two reports in the literature document the ability to image tumor-related receptors in vivo using QDs (33, 54). In the first such report, human prostate cancer cells growing in vivo in nude mice were imaged with an antihuman antibody that targets a receptor (prostate-specific membrane antigen) expressed on prostate cancer cells but not on other tissues (33). In a more recent report, a single QD bound to an antihuman antibody to Her-2 receptor was tracked spatially and temporally as it traversed human tumor xenografts in nude mice (54). In both instances, the choice of an antihuman antibody with little cross-reactivity with murine antigens minimized background noise from the surrounding normal tissues to establish the proof of concept for imaging. The current study extends these observations but accomplishes the more challenging task of imaging entire tumors, akin to clinical scenarios, based on targeting receptors present on both tumors and adjacent normal tissues, both of which are capable of binding the imaging probe used. Further, to illustrate the clinical relevance of this probe, the current study characterizes the pharmacokinetics of this imaging probe to identify an optimal early time point for specific imaging of receptor expression and late time point for repeat imaging.
The kinetics of EGF-QD nanoprobe fluorescence from the tumor tissues contrasted sharply with that of QD nanoparticles. Increased vascular volume and permeability of tumors led to a rapid initial nonspecific tumor influx and an apparent early washout (clearance) of QD-derived fluorescence for EGFR-targeted and nontargeted nanoparticles within the first hour postinjection. The higher peak fluorescence during influx and the faster apparent initial clearance of QD than the EGF-QD nanoprobes is consistent with the smaller size of the QD nanoparticles (∼21 nm) compared with the EGF-QD nanopobes (∼26 nm). Leaky vasculature containing wide interendothelial junctions, incomplete or absent basement membranes, dysfunctional lymphatics, and numerous transendothelial channels contribute to this passive and nonspecific mechanism of extravasation of macromolecules and nanoparticles within tumors (55). Importantly, after reaching apparent dynamic equilibrium at ∼1 h, the steady rate of increase in EGF-QD nanoprobe fluorescence in tumor tissues between 1 and 6 h reflects receptor-specific binding and internalization, which we interpret as the EGFR expression-activity product. In contrast, QD nanoparticles lacking an EGFR binding domain did not accumulate in tumor tissues during this period. Following this passive extravasation, both QD nanoparticles and EGF-QD nanoprobes reached an apparent dynamic equilibrium at 1 h. Thereafter, the steady rate of increase in EGF-QD nanoprobe fluorescence in tumors, between 1 and 6 h, contrasted sharply with the equilibration of QD fluorescence in tumors. These observations suggest that the increase in tumor-to-background ratio of EGF-QD nanoprobes during this time period is not due to increased vascular permeability of tumors, but due to EGFR-specific binding and internalization of these nanoprobes. We interpret this as an index of the ability to image EGFR expression using EGF-QD nanoprobes. The immunofluorescence images demonstrating a pattern of diffuse dispersal of EGF-QD fluorescence within EGFR-expressing tumor parenchyma versus patchy perivascular sequestration of QD that lacks an EGFR binding domain further corroborates the EGFR and tumor specificity of this nanoprobe. Importantly, following this early imaging of tumor EGFR, EGF-QD fluorescence decreases to near-baseline levels at 24 h that could be attributed to slow degradation of the surface ligands and coatings of QDs by lysosomal enzymes leading to surface defects and fluorescence quenching of QDs, which in turn results in substantial loss of fluorescence signals (29, 33, 36). The quantifiable accumulation of EGF-QD nanoprobes in EGFR-expressing tumors at an early time point (4 h) and the subsequent return of signal to baseline on delayed imaging at 24 h suggest that this nanoprobe could potentially facilitate repetitive imaging of EGFR expression during and after therapeutic interventions. The kinetics and accumulation of EGF-QD nanoprobes reported in this study differs from earlier reports on the use fluorescence-dye conjugates that showed a maximum contrast at 24 h after injection of the fluorescent dye probe. This difference in the kinetics may be attributed to the differences in the sizes of the QD and fluorescent dye conjugates. However, it is worth mentioning that further extensive studies are needed to clearly understand the influence of probe size, biopolymer coating, and surface charge/hydrophobicity of QD and fluorescent dye conjugates on the kinetics and distribution in tumor tissues.
One additional attribute of EGF-QD nanoprobe is the low concentration, equivalent to 10 pmol of QD, which is required for optimal imaging of EGFR in vivo. The 10 pmol (equivalent of QD) concentration used in the current study compares favorably to the 1 nmol concentration of Cy5.5 required for optimal imaging of EGFR-expressing breast cancer xenografts using EGF-Cy5.5 conjugates (20). The greater specificity of the EGF-QD nanoprobes could be due to (a) the increased ratio of EGF to QD achievable with the current conjugation scheme that links to sulfhydryl groups (six cysteine residues) versus those that link to free amine groups (three lysine residues) to form EGF-Cy5.5 conjugates; (b) the image processing used in the current study; and (c) the more favorable optical and physical properties of QDs in comparison with organic dyes. Furthermore, given that the dose and surface coating of the semiconductor core of QDs are key determinants of toxicity (37, 38, 44, 47, 56) and no long-term toxic effects have been shown with QD concentrations of up to 20 pmol/g, the low concentration of EGF-QD nanoprobes used in the current study could be relatively nontoxic. Nevertheless, there are persisting concerns regarding the biocompatibility/toxicity of QDs and the possible activation of downstream EGFR signaling that need to be addressed before using these probes for possible clinical applications in the future.
In summary, this EGF-QD nanoprobe with favorable pharmacokinetic properties and good binding affinity to EGFR permits quantifiable imaging of EGFR expression in human colorectal cancer xenografts in mice. Once the concerns about the toxicity of QDs and the activation of EGFR downstream signaling have been addressed and/or when biocompatible probes with similar physical dimensions and comparable biopolymer coatings are available, this class of peptide-conjugated nanoprobes could have widespread clinical applications. Potential clinical scenarios where this class of imaging nanoprobes could find applicability include early detection of cancers (primary, recurrent, or metastatic), image-guided receptor-targeted sentinel node biopsies, individualized treatment stratification, treatment monitoring that provides early indications of positive responses to treatment or the lack thereof to continued therapy, customized dose optimization of targeted therapy, and accelerated drug screening and development.
Grant support: Hitachi Corporation, Japan (S. Krishnan).
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Note: Supplementary data for this article are available at Clinical Cancer Research Online (http://clincancerres.aacrjournals.org/).
J.M. Orenstein-Cardona and N.E. Colón-Casasnovas contributed equally to this work.
We thank Dr. Beth Beadle for insightful discussions and Karen F. Phillips for carefully editing the manuscript.