A Target Cell–Specific Activatable Fluorescence Probe for In vivo Molecular Imaging of Cancer Based on a Self-Quenched Avidin-Rhodamine Conjugate

A major goal of targeted molecular imaging of cancer is to improve sensitivity and specificity for tumors so that even minimal clusters of aberrant cells can be detected in vivo, allowing earlier and more complete therapy. Because the target (in this case, a tumor) is only detectable in reference to its background, the most common strategies to improve targeted tumor imaging are to increase the signal from the target tissue while minimizing nonspecific background signal.

Fluorescence imaging is attractive because of its high sensitivity and specificity, its low cost and portability of the imaging equipment, and its absence of ionizing radiation. Although fluorescence has been used in in vitro diagnostic tests for many years, it has been more difficult to apply in vivo because its “always on” characteristic makes it difficult to distinguish tumor and background. Enzymatic fluorescence activation has been reported with tumor-related enzymes such as cathepsin D (1) and matrix metalloproteinases-2 (MMP-2; ref. 2). For example, Weissleder et al. have developed a protease-activatable fluorescence probe, ProSense 680 (VisEn Medical, Inc., Woburn, MA). The fluorescence of this probe is normally self-quenched but can generate a strong near-IR fluorescence signal from Cy5.5 fluorophores upon reaction with tumor-associated protease such as cathepsin D (3), which “dequenches” and thereby activates the probe. One limitation of this technique is that it will activate in any environment containing the appropriate protease, leading to nonspecific activation (3–5).

Herein, we propose an alternative approach to fluorescence activation whereby a self-quenching targeting moiety is first bound and then internalized within the target cell. Intracellular degradation in the lysosome leads to dequenching and hence activation of the fluorophore only within the targeted cells and not in the surrounding tissue (6). In the lysosome, large polymeric proteins are separated into monomers and cut into small peptides or amino acids by oxidation/reduction, proteases, and/or pH-mediated degradation (6). In this study, we synthesized a self-quenched avidin-rhodamine X conjugate (Av-ROX) as a target cell–specific, activatable imaging agent. Avidin is a noncovalently bound homotetrameric glycoprotein with a molecular weight of 68 kDa, consisting of four monomers with 17-kDa molecular weight that binds via glycosyl chains to a lectin binding protein commonly expressed on cancer cells such as ovarian cancer, colon cancer, gastric cancer, and pancreatic cancer, which have potential to metastasize to the peritoneum (7). Using the self-quenched Av-ROX, we established an in vivo, cancer cell–specific, activatable molecular imaging technique that produced high tumor to background signal ratios when the agent was internalized by cancer cells (Fig. 1; refs. 8–10). Cross-linking the tetrameric bonds of avidin to make them covalent, and thus more resistant to degradation, led to a reduced fluorescent amplification, suggesting it is the breakdown of the avidin tetramer that results in dequenching.

Synthesis of avidin-rhodamine X conjugate. Avidin was purchased from Pierce Biochemical, Inc. (Milwaukee, WI). Rhodamine X succinoimidyl ester was purchased from Molecular Probes, Inc. (Eugene, OR). Av-0.5ROX has 0.5 rhodamine X molecules per avidin, whereas Av-3ROX has three rhodamine X molecules per avidin. At room temperature, 400 μg (5.9 nmol) of avidin in 80 μL of Na₂HPO₄ was incubated with 6 μg (9 nmol; for Av-0.5ROX) or 65 μg (100 nmol; for Av-3ROX) of rhodamine X succinoimidyl ester in DMSO for 15 min. The mixture was purified with Sephadex G50 (PD-10; GE Healthcare, Milwaukee, WI). Av-3ROX samples were kept in the refrigerator for 3 days, and the precipitated fraction was separated by centrifugation; the supernatant was used for further study.

The protein concentration of each sample was determined with Coomassie Plus protein assay kit (Pierce Chem Co., Rockford, IL) by measuring the absorption at 595 nm with a UV-Vis system (8453 Value UV-Bis system, Agilent Technologies, Palo Alto, CA). Then, the rhodamine X concentration was measured by the absorption at 587 nm with a UV-Vis system (8453 Value UV-Bis system, Agilent Technologies) to confirm the number of rhodamine X molecules conjugated with each avidin molecule.

Cross-linking study. Disuccinimidyl suberate (DSS; Pierce, Rockford, IL) was used to cross-link the avidin tetramer. At room temperature, 300 μg of Av-0.5ROX or Av-3ROX in 1 mL PBS was incubated with 100 μg/20 μL DMSO of DSS in 0.1 mol/L Na₂HPO₄ at pH 8.5 for 1 h. The non–cross-linked control was prepared by exactly the same procedure, without DSS. Cross-linked and non–cross-linked samples were used for in vitro SDS-PAGE analysis and fluorescence spectrometry as well as in vivo imaging studies.

Fluorescence intensity analysis of Av-ROX with SDS-PAGE. The fluorescence intensities of avidin monomer and tetramer were determined using SDS-PAGE under reducing conditions (10% 2-mercoptoethanol/95°C/2 min) with a 4% to 20% gradient polyacrylamide gel (Invitrogen/Novex, San Diego, CA). Immediately after separating the protein in the dark room, the fluorescence intensity of wet gels was analyzed with a high-resolution fluorescence scanner system (FLA-5100, Fujifilm Medical Systems USA, Inc., Stanford, CT). An internal laser of 532 nm was used for excitation, and a long-pass filter over 575 nm was employed for light emission. The lateral and longitudinal spatial resolution (pixel size) was 100 μm. The fluorescence intensity of each band was analyzed with commercial software (Multigage, Fujifilm Medical Systems USA), and the ratio of fluorescence intensities was determined. The gels were stained with Coomassie blue using a Colloidal Coomassie gel staining kit (Invitrogen/Novex), dried, and digitally scanned (Epson 6300, Epson America, Inc., Long Beach, CA), and the protein concentration in each band was determined with ImageJ software.⁵

Determination of fluorescence after dequenching of Av-ROX with 5% SDS. To determine the fluorescence of each sample after dequenching, 780 ng/mL Av-0.5ROX and 780 ng/mL Av-3ROX samples were incubated in PBS with or without 5% SDS for 30 min at room temperature. The denaturation of avidin causes the rhodamine X molecules to dequench resulting in fluorescence, which was measured with Perkin-Elmer LS55 fluorescence spectrometer (Perkin-Elmer, Shelton, CT).

Cell culture. An established ovarian cancer cell line SHIN3 (11) was used for generating i.p. disseminated cancer microfoci. The cell lines were grown in RPMI 1640 (Life Technologies, Gaithersburg, MD) containing 10% fetal bovine serum (Life Technologies), 0.03% l-glutamine, 100 units/mL penicillin, and 100 μg/mL streptomycin in 5% CO₂ at 37°C.

Transfection of red fluorescence protein (DsRed2) to the SHIN3 cell. RFP (DsRed2)–expressing plasmid was purchased from Clontech Laboratories, Inc. (Mountain View, CA). The plasmid was transfected into the SHIN3 cells to validate the results with targeted fluorophores (see below). The transfection of RFP was done with an electroporation method using Gene Plus II (Bio-Rad Laboratories, Hercules, CA). Briefly, 3 μg of DsRed2-express plasmid was mixed with 2 million SHIN3 cells in 400 μL of the cell culture medium (RPMI 1640 with 10% FCS). The cell suspension was then placed in a pulse cuvette (Bio-Rad Laboratories), and 250 V pulses were delivered after 950 cycles.

Fluorescence microscopy. SHIN3 cells (5 × 10⁵) were plated on a cover glass–bottomed culture well and incubated for 16 h. Av-0.5ROX or Av-3ROX was added to the medium (20 μg/mL), and the cells were incubated for either 30 min or 6 h. Cells were washed once with PBS, and fluorescence microscopy was done using an Olympus BX51 microscope (Olympus America, Inc., Melville, NY) equipped with the following filters: excitation wavelength, 530 to 570 nm; emission wavelength, 590 nm long pass. Transmitted light differential interference contrast images were also acquired.

In vivo spectral fluorescence imaging study. Thirty micrograms of Av-0.5ROX or 30 μg Av-3ROX (equivalent total avidin dose), or 7 μg Av-0.5ROX or 22 μg Av-3ROX (equivalent fluorescence intensity), or 50 μg cross-linked Av-3ROX or 50 μg non–cross-linked Av-3ROX (in the cross linking study) were diluted in 300 μL PBS and injected into the peritoneal cavities of mice with peritoneally disseminated cancer implants. Mice were sacrificed with carbon dioxide immediately, 1 h, or 3 h after injection (n = 3 in each group). The abdominal cavity was exposed, and spectral fluorescence images were obtained using the Maestro In-Vivo Imaging System (CRi, Inc., Woburn, MA). Whole abdominal images and close-up peritoneal membrane images were obtained. A band-pass filter from 503 to 555 nm and a long-pass filter over 580 nm were used for emission and excitation light, respectively. The tunable filter was automatically stepped in 10-nm increments from 550 to 800 nm while the camera captured images at each wavelength interval with constant exposure. The spectral fluorescence images consisting of autofluorescence spectra and rhodamine X spectra were obtained and then unmixed, based on their spectral patterns using commercial software (Maestro software, CRi).

To compare the signal intensities from the tumors and the backgrounds between Av-0.5ROX and Av-3ROX immediately and 3 h after injection with 30 μg Av-0.5ROX or 30 μg Av-3ROX and 3 h after injection with 7 μg Av-0.5ROX or 22 μg Av-3ROX, a region of interest (ROI) as large as each abdominal cavity was drawn to make the histogram using ImageJ software.⁶

Semiquantitative comparison of fluorescence intensity between cross-linked and non–cross-linked Av-3ROX in vivo. To compare the fluorescence intensities of tumor foci semiquantitatively between the cross-linked Av-3ROX and non–cross-linked Av-3ROX, close-up peritoneal membrane images were obtained 3 h after i.p. injection of 50 μg cross-linked Av-3ROX or 50 μg non–cross-linked Av-3ROX. Using the unmixed fluorescence image of the two peritoneal membranes, ROI as large as the peritoneal membrane was drawn inside the bowel, and histograms (number of pixels at specific fluorescence intensity) were made using ImageJ software.⁶ Then, a threshold was set in the fluorescence intensity, above which a pixel is counted. The total number of pixels (N) within the threshold range was calculated at a threshold value of t, where i is the fluorescence intensity in arbitrary units, n is the number of pixels at the fluorescence intensity of i, t is the threshold value, and N is the total umber of pixels within the threshold range (i ≥ t; Eq. A; ref. 12):

The common logarithm (log) values of N were calculated and plotted as a function of t. The regression line and the correlation coefficient (r) were calculated from these data sets (t and log N) by the Microsoft Excel 2003 (Microsoft, Redmond, WA). Briefly, the dynamic range of signal intensity in the unmixed fluorescence image was set from 0 to 255 in arbitrary units, and the threshold value (t) was changed from 40 to 240 in increments of 10 because the background signals, such as the normal peritoneal membrane excluding tumors and the nonfluorescent plate, were mostly <40 arbitrary units. Then, the total number of pixels (N) within the threshold range was calculated as a function of threshold (t), and regression line was calculated in each ROI. For comparison of fluorescence intensity or “brightness,” the slope of regression line was compared between the two agents. If the absolute value of r is ≥0.9, a comparison study using the slope values was done.

Assessment of the sensitivity and specificity of Av-3ROX and Av-0.5ROX for the detection of peritoneal cancer foci. The sensitivity and specificity of spectral Av-0.5ROX and Av-3ROX imaging for the detection of peritoneal disseminated cancer foci were studied using four tumor-bearing mice for each probe. The i.p. tumor xenografts were established 14 days after i.p. injection of 2 × 10⁶ RFP-transfected SHIN3 cancer cells suspended in 200 μL of PBS, in female nude mice (National Cancer Institute Animal Production Facility, Frederick, MD). Three hours after i.p. injection of 50 μg Av-3ROX or 50 μg Av-0.5ROX diluted in 300 μL PBS, spectral fluorescence images of the peritoneal membranes were obtained by Maestro In-Vivo Imaging System (CRi). A band-pass filter from 445 to 490 nm and a long-pass filter over 515 nm were used for emission and excitation light, respectively. The tunable filter was automatically stepped up in 5-nm increments from 500 to 800 nm while the camera captured images at each wavelength interval with a constant exposure. Spectral unmixing algorithms were applied to create the unmixed images of RFP, ROX, and autofluorescence. For each mouse, two to three different parts of the peritoneal membranes were randomly selected and spread out on a nonfluorescent plate, before close-up spectral fluorescence imaging was done. ROIs were drawn within the nodules depicted by unmixed RFP images, unmixed Av-3ROX or Av-0.5ROX images, or both. Additional ROIs were drawn in the surrounding nontumorous areas on the unmixed RFP images (standard reference for noncancerous foci). The average fluorescence intensity of each ROI was calculated on the RFP and the Av-3ROX and Av-0.5ROX spectral unmixed images using commercial software (Maestro software version 2, CRi). The minimum possible diameter for the ROI was 0.8 mm; thus, all visible nodules with short-axis diameters ≥0.8 mm on either image were included for analysis. Av-ROX–positive nodules were defined as having an average fluorescence intensity ≥1 arbitrary units on the unmixed Av-3ROX images or Av-0.5ROX images, whereas Av-ROX–negative nodules were defined as average fluorescence intensity <1 arbitrary unit. The numbers of foci positive for both Av-ROX and RFP and the number positive only for Av-ROX or RFP were counted. Sensitivity of Av-3ROX or Av-0.5ROX for the detection of peritoneal cancer foci was defined as the number of peritoneal foci positive for both Av-ROX and RFP divided by the number of peritoneal foci positive for RFP. Specificity of Av-ROX was defined as the number of peritoneal foci negative for both RFP and Av-ROX divided by the number of peritoneal foci negative for RFP.

Av-3ROX can be dequenched by detergent in vitro. Av-0.5ROX has 0.5 rhodamine X molecules per avidin, whereas Av-3ROX has three rhodamine X molecules per avidin. Consequently, Av-0.5ROX is fluorescent, whereas Av-3ROX is self-quenched. Therefore, at the same concentration (780 ng/mL), Av-0.5ROX was brighter than Av-3ROX in PBS, although it has fewer rhodamine X molecules (Fig. 2A and B). After incubation with PBS with or without 5% SDS for 30 min in room temperature, fluorescence intensities of Av-0.5ROX and Av-3ROX increased 3-fold and 39-fold, respectively (Fig. 2A and B). SDS-PAGE under reducing conditions showed that the avidin monomers (17 kDa) of Av-3ROX emitted 5-fold greater fluorescence than that of Av-0.5ROX (Fig. 2C).

Av-3ROX shows more fluorescence than Av-0.5ROX once internalized by SHIN3 cells in vitro. Serial observation of SHIN3 cells incubated with Av-0.5ROX or Av-3ROX was done using fluorescence microscopy to compare the temporal changes of intracellular fluorescence production and distribution. Fluorescence microscopy showed fluorescent dots within the cytoplasm as early as 30 min after incubation with both Av-0.5ROX and Av-3ROX (Fig. 2D). However, the fluorescence intensity of Av-0.5ROX was initially higher than that of Av-3ROX. The number of fluorescent dots within the cytoplasm increased 6 h after incubation in both groups. However, fluorescent dots produced by Av-3ROX became larger and brighter than those of Av-0.5ROX (Fig. 2D). These results indicate that both Av-0.5ROX and Av-3ROX are internalized into the cells (8, 12–14); however, the self-quenched Av-3ROX was activated once it was internalized and increased markedly in fluorescence likely as a consequence of proteolysis in the lysosome (Fig. 1).

Av-3ROX produces higher target to background fluorescence within tumor nodules in vivo than Av-0.5ROX. To show that Av-3ROX could be activated within the target tumor and to achieve target-specific cancer imaging with high signal-to-background ratio, in vivo spectral fluorescence imaging with Av-3ROX and Av-0.5ROX was done in a mouse cancer model of i.p. metastases using SHIN3 ovarian cancer cells. Immediately after i.p. injection of 30 μg Av-0.5ROX or 30 μg Av-3ROX in 300 μL PBS in tumor-bearing mice with peritoneal implants, the fluorescence intensity of Av-0.5ROX was higher than that of Av-3ROX, but tumor nodules were hard to detect due to the high background signals produced by Av-0.5ROX (Fig. 3A). Av-3ROX produced minimal background and tumor fluorescence resulting in poor tumor detection. However, by 1 h after i.p. injection, the fluorescence intensity arising from tumor nodules was higher with Av-3ROX compared with Av-0.5ROX, and the background signal was lower (data not shown). By 3 h after injection, the fluorescence intensity of the tumor nodules was much higher with Av-3ROX than with Av-0.5ROX (Fig. 3B). The background signal was comparable between the two groups. Because the fluorescence intensity of the two probes at the same concentration of avidin is different for Av-0.5ROX and Av-3ROX, a comparison was done in which the two optical agents had the same initial fluorescence that was achieved by using 7 μg Av-0.5ROX and 22 μg Av-3ROX. The fluorescence intensity of Av-0.5ROX from the tumor nodules was slightly higher than that of Av-3ROX 1 h after injection (data not shown). By 3 h, however, the fluorescence signal arising from Av-3ROX was much higher than that of Av-0.5ROX, although the macroscopic tumor sizes were comparable between Av-0.5ROX and Av-3ROX (Fig. 3C). Immediately after injection with 30 μg Av-0.5ROX or 30 μg Av-3ROX, the histogram of Av-0.5ROX showed a larger proportion of counts for pixels in the higher fluorescence intensity range than Av-3ROX (Fig. 3A). However, 3 h after injection with 30 μg Av-0.5ROX or 30 μg Av-3ROX, the histogram of Av-3ROX had a larger proportion of counts for pixels in the higher fluorescence intensity range than Av-0.5ROX (Fig. 3B). At 3 h after injection with 7 μg Av-0.5ROX or 22 μg Av-3ROX, the histogram of Av-3ROX showed a larger proportion of counts for pixels in the higher fluorescence intensity range than Av-0.5ROX (Fig. 3C). These results indicate that a higher signal-to-background ratio could be obtained with Av-3ROX compared with Av-0.5ROX in the settings of the same avidin dose or the same fluorescence intensity.

Cross-linking limits the in vivo dissociation of Av-3ROX, thus reducing fluorescence. Av-3ROX was stabilized by cross-linking with DSS, which covalently bound the avidin tetramer using amide residues on lysine molecules. Cross-linked or non–cross-linked Av-3ROX was separated into monomer, dimer, trimer, and tetramer with reducing conditions of SDS-PAGE. The fluorescence intensity of Av-3ROX monomer was 8-fold stronger than that of cross-linked Av-3ROX tetramer at the same protein concentration (ImageJ software; Fig. 4A). To assess the influence of cross-linking on the fluorescence intensity, the fluorescence intensity of cross-linked Av-3ROX was compared with non–cross-linked Av-3ROX at the same concentration with or without 5% SDS. Cross-linking decreased the fluorescence intensity of Av-3ROX with 5% SDS, although cross-linking did not affect the fluorescence intensity without SDS (Fig. 4B). To verify that cross-linking decreases the fluorescence intensity of Av-3ROX in vivo, the fluorescence intensity of cross-linked Av-3ROX was compared with non–cross-linked Av-3ROX. Non–cross-linked Av-3ROX clearly visualized submillimeter cancer implants with minimal background, whereas the fluorescence signal of the cross-linked Av-3ROX arising from the tumor nodules was considerably lower (Fig. 4C). To make an objective comparison of fluorescence intensity of the peritoneal implants between cross-linked Av-3ROX and non–cross-linked Av-3ROX, ROI as large as the peritoneal membrane was drawn inside the intestine using the unmixed Av-3ROX image, and a histogram depicting the distribution of pixel intensities was created (Fig. 4D). The correlation coefficients of cross-linked Av-3ROX and non–cross-linked Av-3ROX were −0.9927 and −0.9997, respectively (Fig. 4D). The slopes of cross-linked Av-3ROX and non–cross-linked Av-3ROX were −0.0375 and −0.0110, respectively (Fig. 4D). These results indicate that the activation of Av-3ROX could be hampered by stabilizing the Av-3ROX tetramer with cross-linking.

Sensitivity and specificity of Av-3ROX and Av-0.5ROX in vivo spectral fluorescence imaging determined by RFP labeling. To investigate the effectiveness of Av-3ROX, the sensitivity and specificity of spectrally unmixed Av-3ROX imaging to detect peritoneal cancer foci were assessed and compared with Av-0.5ROX using RFP-transfected SHIN3 cancer cells. Av-3ROX or Av-0.5ROX was injected into RFP-transfected SHIN3 tumor-bearing mice, and spectral fluorescence imaging was done. Spectrally resolved RFP images, Av-ROX images, and composite images were made. Unmixed Av-3ROX images showed ring-like accumulation around the foci, which were depicted by the unmixed RFP images (Fig. 5A). A total of 514 peritoneal tumor foci in four mice were identified by the unmixed Av-3ROX images, the unmixed RFP images, or both. Additionally, 499 ROIs were created in the nontumorous areas (i.e., where no tumors were visible on the RFP images). RFP-positive foci (positive standard) were defined as those whose fluorescence intensities were ≥60 arbitrary units on unmixed RFP images, and Av-ROX–positive foci were defined as those whose fluorescence intensities were ≥1 arbitrary units on unmixed Av-3ROX images or Av-0.5ROX images. Four hundred sixty-five foci showed Av-3ROX fluorescence intensities ≥1 arbitrary units among the 507 RFP-positive foci (Fig. 5B). Four hundred ninety-seven foci showed Av-3ROX fluorescence intensities <1 arbitrary unit among the 506 RFP-negative foci (i.e., fluorescence intensities <60 arbitrary units on unmixed RFP images). Thus, the spectral unmixed Av-3ROX imaging had a sensitivity of 92% (465 of 507) and a specificity of 98% (497 of 506). In contrast, unmixed Av-0.5ROX images showed ring-like accumulation around the foci, which were depicted by the unmixed RFP images; however, the fluorescence intensity of tumor foci on unmixed Av-0.5ROX images were suboptimal (Fig. 5C). Sixty-six foci showed Av-0.5ROX fluorescence intensities ≥1 arbitrary units among the 190 RFP-positive foci (Fig. 5D). Four hundred sixty-two foci showed Av-0.5ROX fluorescence intensities <1 arbitrary unit among the 474 RFP-negative foci. Thus, the spectrally unmixed Av-0.5ROX imaging had a sensitivity of 35% (66 of 190) and a specificity of 97% (462 of 474). Without injection of Av-3ROX, no foci showed Av-3ROX fluorescence intensities ≥1 arbitrary units among the 312 RFP-positive foci.

As shown on Fig. 3, the low background signal found with the activatable Av-3ROX agent enabled us to achieve targeted imaging of cancer nodules within a shorter time frame after injection than the conventional “always on” imaging agent Av-0.5ROX. Tiny tumor nodules, invisible to the naked eye, were visualized as early as 1 h after injection with activatable Av-3ROX. In contrast, avidin conjugated with three FITCs (Av-3FITC), known to be brighter than Av-3ROX in vitro (13) while also able to show submillimeter tumor nodules, required up to 4 h of incubation in the peritoneum. An important advantage of this strategy is that Av-3ROX is activated only after it binds to the specific target and is endocytosed (6). Thus, most of fluorescence signal arises from activated Av-3ROX within cancer cells and not from background normal tissues. Based on the current data, this target-specific activatable agent is more efficient than conventional nonactivatable agents because of the high specificity conferred by lectin binding and the high sensitivity due to the increased amplification of fluorescence following activation.

Fluorescence activation by enzymes such as cathepsin D and MMP-2, which are expressed by some tumors has been reported by Weissleder et al. using self-quenched synthetic polymers containing enzyme-sensitive sequences (1–4). Because these enzymes will work mostly outside the cancer cells, the activated fluorescence reagents may leak out of the cancer tissue (3) and increase the surrounding background signal. To the extent that such enzymes are ubiquitous, there will be nonspecific activation as well. With Av-3ROX, the activation of the probe by the lysosomal catabolism is produced only within the target cells following receptor internalization. Additionally, as shown in the cross-linking experiment (Fig. 4), the amplification of fluorescence signal in this system after activation was by at least a factor of 8 due to the dissociation of avidin tetramers to monomers. The catabolism of Av-3ROX by the proteolytic activity of lysosomes added approximately thrice more activation of fluorescent signal.

The strategy of signal activation by selective binding, internalization, and catabolism leading to dequenching can be applied to a wide variety of specific targets on the surface of cancer cells, provided the conjugates then undergo endocytosis and degradation. Like avidin, antibodies, one of the major target-specific macromolecular agents, also consist of four protein chains bound with disulfide bonds, although the self-quenching magnitude of antibody (trastuzumab: Herceptin) conjugated with three rhodamine X molecules was smaller than that of Av-3ROX probably because of the larger size of the antibody (data not shown) activation of fluorescence signal could also be induced by SDS-induced dissociation. Increased fluorescence signal after exposure to SDS was also observed with Av-0.5ROX. The precise mechanism of the fluorescence enhancement and reduction in rhodamine dyes is not yet understood (15–19), and the reason why SDS can amplify the signal of Av-0.5ROX and Av-3ROX in general is still unclear. We suggest that the detergent function of SDS can change the conformation of the avidin chain, which allows the rhodamine molecules to escape from the highly charged peptide, which normally compromises the fluorescence signal of rhodamine X. It is unclear whether the activation mechanism induced by detergent function is the same as that of lysosomal dissociation.

In previously reported pharmacokinetic studies with radiolabeled avidin or antibody, avidin and avidin-antibody complexes were digested and catabolized into monomers or short peptides consisting of several amino acids in the lysosome (9). However, the exact fate of the avidin tetramer after endocytosis is still unclear. Because the avidin tetramer was easily dissociated by detergent, it is difficult to determine the exact intracellular location where the tetramer was dissociated. Using fluorescence microscopy, numerous fine spotted fluorescence signals were seen both 30 min and 6 h after incubation with Av-0.5ROX. In contrast, self-quenched Av-3ROX showed larger and brighter fluorescent dots within the cytoplasm 6 h after incubation compared with Av-0.5ROX (Fig. 2D). A similar pattern of fluorescence was seen in vivo, using fluorescence microscopy 3 h after injection with Av-3ROX (data not shown). The difference in fluorescence patterns suggests that most of the self-quenched Av-3ROX probes became activated within the lysosome. In addition, the inhibition of fluorescence activation by the covalent cross-linking both in vitro and in vivo also suggests that catabolism of the avidin may activate fluorescence in cancer cells. In vivo fluorescence is more difficult to quantify than in vitro fluorescence because the images are always adjusted to minimize autofluorescence, which varies with each animal. Therefore, quantitation of in vivo optical fluorescence is only approximate.

This methodology holds promise for human imaging. The high target to background ratio brought about by activation enables the detection of very tiny tumor deposits. This could be useful for peritoneal debulking procedures for ovarian cancer and would require only minimal additional equipment in the operating room. It should be noted, however, that the immunogenic nature of avidin makes it an unlikely candidate for clinical translation. However, in principle, the same methods could be applied to binding molecules with lower immunogenic potential. Because the Av-3ROX and related compounds penetrate only the surface cells of a tumor, it should be possible to optically detect tumors over a range of sizes.

We were able to confirm the sensitivity and specificity of Av-3ROX for detecting ovarian cancer implants (for lesions at least 0.8 mm in short diameter) using an RFP-transfected SHIN3 cell line. Forty-two of the 507 lesions seen by RFP were not confirmed by Av-3ROX, indicating a 92% sensitivity for lesions ≥0.8 mm. Nine of 503 foci identified by Av-3ROX were not confirmed by RFP and were considered false positives, resulting in a 98% specificity. In this experiment, to co-excite but separately detect emissions from RFP (peak emission, 579 nm) and Av-3ROX (peak emission, 607 nm), we used a blue excitation light filter set (excitation, 445–490 nm/emission, 515 long pass). This filter is less efficient for excitation of Av-3ROX than the green light filter set (excitation, 503–555 nm/emission, 580 long pass) used in all the imaging studies. A green excitation light can improve the sensitivity value because the green light is able to yield 12-fold stronger emission from Av-3ROX than the blue light. However, RFP counter staining to Av-3ROX technically did not allow the use of a suitable excitation filter for exciting Av-3ROX, and the lower photon efficiency of Av-3ROX with blue light likely compromised the sensitivity. A clear distinction can be made between Av-3ROX–positive and Av-3ROX–negative foci as shown in Fig. 5. Although the high sensitivity/specificity can be determined for 0.8 mm or larger tumors, even smaller tumors are readily detectable with this method. Lesions as small as 100 μm were detectable, but it is technically impossible with the current image acquisition methods using the Maestro In-Vivo Imaging System to determine the limits of size detection or the sensitivity and specificity for these nearly microscopic cancer clusters.

In conclusion, a target-specific activatable cancer imaging system is proposed, using the catabolism function of the lysosome to cause dissociation and dequenching of an avidin tetramer bound to multiple fluorescent molecules. The fluorescence signal of the self-quenched Av-3ROX was activated ∼40 fold in vitro, and this activation strategy was verified in vivo. This novel molecular imaging technique enabled us to visualize submillimeter tumor nodules with minimal background signal within a short time following Av-3ROX injection and holds promise as a method of optically enhancing surgical or endoscopic procedures in humans.

Grant support: Intramural Research Program of the NIH, National Cancer Institute, Center for Cancer Research.

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.

We thank Paul Albert Ph.D. for kindly performing a receiver operating characteristic curve analysis.