In Vivo Imaging of Tissue-Remodeling Activity Involving Infiltration of Macrophages by a Systemically Administered Protease-Activatable Probe in Colon Cancer Tissues^,

Abstract

This study evaluated the detection of tumors using in vivo imaging with a commercially available and systemically administered protease-activatable fluorescent probe, ProSense. To this end, we analyzed the delivery and uptake of ProSense as well as the target protease and its cellular source in a mouse xenograft tumor model. In vivo and ex vivo multi wavelength imaging revealed that ProSense signals accumulated within tumors, with preferential distribution in the vascular leakage area that correlates with vasculature development at the tumor periphery. Immunohistochemically, cathepsin B, which is targeted by ProSense, was specifically localized in macrophages. The codistribution of tenascin C immunoreactivity and gelatinase activity provided evidence of tissue-remodeling at the tumor periphery. Furthermore, in situ zymography revealed extracellular ProSense cleavage in such areas. Colocalization of cathepsin B expression and ProSense signals showing reduction by addition of cathepsin B inhibitor was confirmed in cultured macrophage-derived RAW264.7 cells. These results suggest that increased tissue-remodeling activity involving infiltration of macrophages is a mechanism that may be responsible for the tumor accumulation of ProSense signals in our xenograft model. We further confirmed ProSense signals at the tumor margin showing cathepsin B⁺ macrophage infiltration in a rat colon carcinogenesis model. Together, these data demonstrate that systemically administered protease-activatable probes can effectively detect cancer invasive fronts, where tissue-remodeling activity is high to facilitate neoplastic cell invasion.

Introduction

Molecular imaging using exogenous tumor-specific fluorescent probes is becoming an essential tool for the study of diseases, therapeutic effects, and clinical applicability [1–3]. Human clinical studies for diagnostic and image-guided surgery are also emerging [4,5]. Tumor imaging with fluorescent probes has the potential to promote a paradigm shift in surgical inspection by enabling the localization of lesions that are difficult or impossible to detect by visual observation or palpation. This improved tumor detection can then improve patient outcomes by identifying much smaller lesions and/or excising tumor tissue more completely. However, the factors affecting tumor imaging performance, such as target molecule activity, tumor growth environment, vasculature, and stromal components, can vary on the basis of the fluorescent probe design and type of cancer. Therefore, to assess their clinical applicability, it is important to understand how the biologic characteristics of these tumors affect the in vivo imaging of these fluorescent probes.

Proteases have pivotal roles in altering local microenvironments during embryonic development and growth as well as in both physiological and pathologic tissue-remodeling processes. Cysteine cathepsins, simply referred to as cathepsins in this study, consist of 11 family members in humans (B, C, H, F, K, L, O, S, L2/V, W, and X/Z). Cathepsins collectively have important roles in lysosomal protein degradation and individually contribute to pathologic conditions [6]. Specifically, cathepsin B is one of the most prominent proteases in the tumor proteolytic network and is upregulated in many different tumor microenvironments, from early neoplastic to advanced metastatic lesions [7,8]. Cathepsin B is frequently translocated from lysosomes to the cell surface and/or secreted into the extracellular matrix of tumor cells to help their invasive growth by local activation of tissue proteolysis [9,10]. The central role of cathepsin B in carcinogenesis suggests that it is not only a promising target for chemotherapy and chemoprevention but also for tumor detection.

Previous reports have shown that in vivo fluorescent imaging of cathepsin B activity is capable of detecting tumors with high sensitivity [11]. The cathepsin probe is an activatable graft copolymer coupled with quenched fluorophores that are released after cathepsin B cleavage, which allows for the detection of small or early lesions. The probe is also designed for prolonged circulation by attaching several polyethylene-glycol side chains to the polymer. A commercially available probe, ProSense, is preferentially hydrolyzed by cathepsin B but can also be activated through proteolysis by other cathepsins, such as cathepsin L and cathepsin S. ProSense has been shown to detect a variety of experimentally induced tumors of the colon, ovary, mammary gland, pancreas, and esophagus [12–16]. However, the mechanisms underlying the tumor specificity of ProSense signals are still unclear because this probe is exposed to numerous complex in vivo processes. Systemically administered ProSense is delivered to tumor tissue during prolonged circulation, cleaved by target cathepsins in the local tumor microenvironment, and is then retained within the tumor tissue. Therefore, each of these processes needs to be investigated, as well as the relationship between them within the same tumor, to clearly understand the tumor specificity of ProSense.

In this study, we aimed to clarify how the biological characteristics of tumors affect the in vivo imaging of systemically administered protease activatable fluorescent probes. For this purpose, we performed a time course analysis of delivery and uptake of ProSense in a mouse xenograft model at both the macroscopic and microscopic levels using in vivo and ex vivo multiwavelength imaging techniques. We then determined the target molecules and cellular sources that generate the ProSense-cleaving activity within the tumor. We further confirmed the specificity of ProSense signals with protease expression in cultured cells. To investigate the influence of inflammation and immunity as well as carcinogenesis on the tumor microenvironment, we analyzed ProSense signals in an immunocompetent rat colon carcinogenesis model induced by azoxymethane (AOM).

Materials and Methods

Cancer Cell Lines

Human colon tumor cell lines HT29 and HCT116 and a murine macrophage-derived cell line RAW264.7 were purchased from DS Pharma Biomedical (Osaka, Japan). HT29 or HCT116 cells were cultured in McCoy's 5A medium (Life Technologies, Carlsbad, CA) supplemented with 10% FBS, 100 U/ml penicillin, and 0.1 mg/ml streptomycin. RAW264.7 cells were cultured in Dulbecco's modified Eagle's (DME) medium (Life Technologies) supplemented with 10% FBS, 100 U/ml penicillin, and 0.1 mg/ml streptomycin. Cells were routinely passaged at subconfluence. Only low-passage cells (passage < 10 times) were used.

Animal Model

To develop a xenograft tumor model, a total of 12 male BALB/c nu/nu mice (4 weeks old) were purchased from Japan SLC Inc (Hamamatsu, Japan). They received CA-1 (CLEA Japan Inc, Tokyo, Japan) as a basal diet and tap water ad libitum period. At 5 to 6 weeks of age, the mice were inoculated subcutaneously with 5 x 10⁶ HT29 or HCT116 cells in 0.1 ml of medium.

To develop an immunocompetent colon carcinogenesis model, 11 male F344 rats (4 weeks old) were purchased from Japan SLC Inc. The rats received CE-2 (CLEA Japan Inc) as a basal diet and water ad libitum throughout the experimental period. Starting at 6 weeks of age, rats were given subcutaneous injections of AOM (Sigma-Aldrich, St Louis, MO; 15 mg/kg body weight) once a week for 3 weeks as described previously [17].

All animal experiments were performed in accordance with the guidelines for animal experimentation of the Facility of Agriculture, Tokyo University of Agriculture and Technology.

In Vivo and Ex Vivo Imaging

To detect cathepsin activity, we used ProSense 750/750 EX (Perkin Elmer, Waltham, MA). To visualize the architecture of the blood vessels, we used fluorescein-conjugated Lycopersicon esculentum lectin (fluorescein-tomato lectin; Vector Laboratories Inc, Burlingame, CA) or DyLight 594-conjugated Lycopersicon esculentum lectin (DyLight 594-tomato lectin; Vector Laboratories Inc) [18]. We used a blood pool imaging agent, AngioSense 680/680 EX (PerkinElmer), to detect nonspecific probe delivery and uptake [19].

To image whole animals noninvasively, we used an epifluorescence imaging system (OV100; Olympus, Tokyo, Japan). We used a multiwavelength laser scanning microscope (IV100; Olympus) to acquire high-magnification and high-resolution images.

In the mouse xenograft tumor model, in vivo imaging was performed 1 to 3 weeks after inoculation of tumor cells. To reduce autofluorescence derived from the animals' intestinal contents, mice received an alfalfa-free CA-1 diet (CLEA Japan Inc) 2 to 3 days before in vivo imaging [20]. A solution mixture of 2 nmol each of ProSense 750 and AngioSense 680 in a total volume of 165 µl was injected intravenously into tumor-bearing mice. To capture the distribution kinetics of these reagents, fluorescent images were taken 10 minutes and 24 to 30 hours after injection. Fluorescein-tomato lectin (200 µg/100 µl) was injected intravenously and allowed to circulate for 5 minutes, followed by imaging. Mice were euthanized immediately after in vivo imaging, and tumors were excised and cut in half longitudinally. Ex vivo imaging of the tumor cut surface was performed to analyze the distribution of reagents within the tumor tissue. After ex vivo imaging, tissues were fixed in phosphate-buffered 10% formalin (pH 7.4) for at least 24 hours. The other half of the tissue sample was snap frozen in liquid nitrogen. To measure the probe fluorescent intensity, whole-animal and high-resolution images were first obtained by the OV100 and IV100 microscopes, respectively, and then analyzed using their respective software.

In the rat colon carcinogenesis model, a solution mixture of 10 nmol of ProSense 750/750 EX and 5 nmol of AngioSense 680/680 EX in a total volume of 450 µl was injected intravenously 23 to 27 hours before euthanasia at 28 to 37 weeks post-initiation. DyLight 594-tomato lectin (250 µg/250 µl) was injected intravenously and allowed to circulate for 20 minutes before euthanasia. Tumors were excised and processed for histologic analysis or immunofluorescence. Ex vivo imaging of the tumor cut surface was performed as described above.

Histology, Immunohistochemistry, and Immunofluorescence

To identify ProSense⁺-tissue component(s) within the tumor tissue, unfixed cryosections from ProSense 750-injected tissue sample were scanned with IV100, fixed with 4% paraformaldehyde, and then immediately stained with hematoxylin and eosin (H&E) for microscopic observation. AOM-induced rat colon tumors were diagnosed according to previously published criteria [21].

Immunohistochemistry (IHC) was performed in paraffin-embedded tissue sections by incubation with primary antibodies referred to in Table W1. Immunodetection was carried out with an HRP-avidin-biotin complex using a VECTASTAIN Elite ABC kit (Vector Laboratories Inc), with DAB/H₂O₂ as the chromogen. Sections were counterstained with hematoxylin.

Immunofluorescence was performed using methanol-fixed cryosections for mouse tissues or paraffin-embedded tissue sections for rat tissues and the antibodies referred to in Table W1. The sections were incubated with primary antibodies and then with species-appropriate Alexa Fluor-conjugated antibodies (1:200; Life Technologies) and counterstained with 4′,6-diamidino-2-phenylindole, dilactate (DAPI; Life Technologies).

Microscopic observation was performed with a fluorescent virtual microscopy system (VS120 Fluorescence; Olympus) or a fluorescent microscope (BX53; Olympus). Fluorescent images were analyzed using IV100 software.

Cultured Cell Experiments

For in vitro detection of ProSense signals and cathepsin B immunoexpression, HT29, HCT116, or RAW264.7 cells (4 x 10⁴) were seeded into a noncoated chamber slide (81 mm² per chamber; Iwaki, Tokyo, Japan). After 2 days, the medium was replaced with Hank's balanced salt solution (HBSS) containing 5 nM ProSense 680 (PerkinElmer). After incubation at 37°C for 3 hours, nuclear labeling was performed with Hoechyst 33342 (Life Technologies). The unfixed cultured cells were scanned with VS120 Fluorescence, fixed with methanol, and then stained for cathepsin B by immunofluorescence. For coculture experiment, HT29 or HCT116 cells (4 x 10⁴) were seeded into a noncoated chamber slide. After 1 day, RAW264.7 cells (0, 4, 10, or 20 x 10⁴) were added into the chamber and allowed to adhere and grow in culture medium containing 10% FBS/DME's medium for 1 day. ProSense signals were detected as described above, followed by immunofluorescence of cathepsin B with CD45. For protease inhibition assay, RAW264.7 cells were cultured in the presence of cell-permeable pan-cathepsin inhibitor (E64d; Sigma-Aldrich) or cell-permeable cathepsin B inhibitor (CA074Me; EMD Millipore, Billerica, MA). RAW264.7 cells (4 x 10⁴) were seeded into a noncoated chamber slide. After 2 days, the medium was replaced with HBSS containing vehicle (DMSO), E64d (5, 25, or 50 nM), or CA074Me (5, 25, or 50 nM). After incubating at 37°C for 1 hour, 5 nM of ProSense in the presence or absence of inhibitor was added, and the cells were incubated at 37°C for 3 hours. ProSense signals were detected as described above. All experiments were performed in duplicate.

In Situ Zymography

In situ zymography was performed on the fresh-frozen sections according to the method described previously [22], with some modifications. Gelatinase activity or ProSense-cleaving activity was detected in an unfixed cryosection using fluorescein-quenched gelatin (DQ gelatin; Life Technologies) or ProSense 680, respectively. Cryosections were overlaid with a solution of 50 mg/ml DQ gelatin or 13 nmol/ml ProSense 680, mixed with 1% (wt/vol) low-melting temperature agarose in phosphate-buffered saline (PBS; pH 7.4) containing propidium iodide (BD Biosciences, San Jose, CA) or DAPI to counterstain nuclei. Fluorescent images were acquired using either the IV100 or BX53 microscope.

Statistical Analysis

Fluorescent intensity data from the whole-animal images between the two groups were compared using a Student's t test when the variance was proven to be homogeneous among the groups by the equal variance test. When a significant difference in variance was observed, Aspin-Welch t test was performed. P values <.05 were considered significant.

Results

ProSense Signals in Xenograft Tumor Model

Although histopathologic patterns of xenograft tumors were different between HT29 and HCT116 in terms of the production of stromal connective tissue (Figure W1), both xenograft tumors showed tumor-specific ProSense signals with a significant increase in fluorescence relative to the skin (Figure 1, A and B). Compared with tumors, the liver and kidney revealed strong signals, whereas the colon and muscle resulted in weak signals (Figure W2). High-magnification (x20) images showed that ProSense signals in the liver and kidney were localized in Kupffer cells and renal tubular epithelial cells, respectively (Figures W2 and W3).

Figure 1.

In vivo imaging of ProSense in a mouse xenograft tumor model using HCT116 and HT29 human colon tumor cell lines. (A) Whole-body imaging of ProSense. Left: Bright field. Middle: Autofluorescence before injection of ProSense. Right: ProSense at 1 day post-injection. Asterisks indicate the kidneys. Scale bar, 10 mm. (B) Tumor-to-skin ratio of average ProSense fluorescent intensity in each tumor. Error bars represent means ± SD (n = 6). **P < .01 and *P < .05 (Aspin-Welch t test). (C) Ex vivo cut surface image of the HT29 tumor. Left: ProSense. Right: H&E-stained view of the tumor corresponding to the fluorescent image. Scale bars, 1 mm. (D) High-magnification (x20) view of ProSense image at the tumor periphery. Left: ProSense (red) with tissue autofluorescence represents connective tissue (green). Right: H&E-stained view of the tumor corresponding to the fluorescent image. Scale bars, 100 µm. TP, tumor parenchyma; S, stroma. (A, C, and D) Six animals examined showed similar results. All images are from the same animal.

The whole-tumor view at the cut surface showed that ProSense signals were primarily in stromal tissue at the tumor periphery and between the tumor parenchyma but rarely within the tumor parenchyma (Figure 1C). At high magnification (x20), strong ProSense signals were observed in the stromal area in both of the two xenograft tumors (Figure 1D and data not shown).

In Vivo Delivery and Uptake of ProSense within Tumor Xenografts

In vivo imaging revealed that AngioSense signals increased slowly within tumor portions after injection of the probes, reflecting vascular leakage of AngioSense specific to tumors (Figure 2A). Similar to AngioSense, ProSense also slowly increased within tumors, although the kidney showed a rapid increase in ProSense signals within 10 minutes after injection, reflecting renal clearance of uncleaved ProSense probes. Ear tissue rarely showed any increase in either AngioSense or ProSense signals (Figure 2B).

Figure 2.

In vivo imaging of ProSense with vascular imaging probes, tomato lectin, and AngioSense in a mouse xenograft tumor model using HCT116 and HT29 human colon tumor cell lines. (A) Whole-body time-lapse imaging of AngioSense (upper panels) and ProSense (lower panels). Left: Autofluorescence before probe injection. Middle: AngioSense or ProSense within 10 minutes after injection. Right: AngioSense or ProSense at 1 day post-injection. Asterisks indicate ProSense signals in the kidney. The arrows indicate large blood vessels around the tumor. The arrowheads indicate autofluorescence derived from animals' intestinal contents. Scale bar, 10 mm. (B) Noninvasive time-lapse imaging of AngioSense (blue) and ProSense (red) in the HT29 tumor (upper panels) and ear (lower panels). Left: Merged images of AngioSense and ProSense within 10 minutes after injection. Note the lack of ProSense signals and the autofluorescent signals of the hair follicles (green). Right: Merged images of AngioSense, ProSense, and tomato lectin 1 day after injection of AngioSense and ProSense and 5 minutes after injection of tomato lectin (green). Arrowheads indicate identical portions of vasculature. Scale bar, 500 µm. (C) Ex vivo imaging of the HT29 tumor surface with tomato lectin (green), AngioSense (blue), and ProSense (red). Upper panel: Low-magnification (x4) view. Lower panel: High-magnification (x20) view of boxed area as shown in the upper panel. Scale bar for the low-magnification view, 500 µm; for the high-magnification view, 100 µm. (D) The whole-tumor view of the cut surface imaging of the HT29 tumor. Left: Tomato lectin image. Middle left: AngioSense. Middle right: ProSense. Right: Merged image. Scale bar, 1 mm. (E) High-magnification (x20) view of the cut surface imaging of the HT29 tumor at the tumor periphery (upper panels) and tumor center (lower panels) as shown in the boxed areas of D. Scale bar, 100 µm. (F) Histogram of the fluorescent intensity with tomato lectin (green), AngioSense (blue), and ProSense (red). The fluorescent intensity profile was measured on the merged image (arrow) and plotted. (A–F) Five animals examined showed similar results. All images are from the same animal.

With ex vivo tumor imaging, we observed that ProSense signals were localized within cell cytoplasm, whereas AngioSense signals were broadly distributed in interstitial tissue (Figure 2C). Whole-tumor images at the cut surface showed tomato lectin signals in conjunction with the vasculature at the tumor periphery, but the tumor center showed reduced signals. This reflects the low density of the vasculature in the center compared with the tumor periphery (Figure 2D). Similarly, AngioSense signals were also primarily observed at the tumor periphery. Merged images of tomato lectin signals, AngioSense signals, and ProSense signals demonstrated that ProSense signals were distributed in areas showing high vascular density. At high magnification (x20), ProSense⁺ cells were preferentially observed in areas with increased vascular leakage at the tumor periphery, in contrast to the weak signals observed in areas with low vascular leakage at the tumor center (Figure 2E). Intensity profile analysis for tomato lectin or AngioSense with ProSense corroborated the imaging results of these signal distributions (Figure 2F).

Immunoreactivity for Cathepsins and Their Cellular Identity in Tumor Xenografts

Immunolocalization of cathepsin B, cathepsin L, and cathepsin S was similarly observed in the stroma at the tumor periphery (Figure 3, A–C). Cathepsin B localized strongly in the cytoplasm of stromal cells and faintly in the cytoplasm of tumor cells. These stromal cathepsin B⁺ cells were prominently observed at the tumor periphery. Although fewer stromal cells were distributed between tumor lobules than in the tumor periphery, those distributed between tumor lobules also showed strong cathepsin B immunoreactivity. With regard to cathepsin L, immunoreactive cells also showed weak cytoplasmic expression in the stroma at the tumor periphery, rarely in the tumor parenchyma, and no expression in tumor cells. Cathepsin S localized in the cytoplasm of stromal cells at the tumor periphery and at the interlobular connective tissue and also diffusely but weakly in the cytoplasm of tumor cells. The distribution of ProSense signals was found to be more closely related to that of cathepsin B than that of other cathepsins (Figures 2D and 3, A–C). ProSense signals were largely colocalized with cathepsin B fluorescence in the stromal areas at the tumor periphery (Figure 3D). Colocalization of ProSense signals and cathepsin B was also confirmed in Kupffer cells in mouse livers (Figure W3).

Figure 3.

Immunoreactivity for cathepsins targeted by ProSense in the HT29 tumor. (A–C) Cathepsins B, L, and S, respectively, in the tumor shown in Figure 2. Left: Whole-tumor view. Right: High-magnification (x20) view at the tumor periphery. Scale bar for the whole-tumor view, 1 mm; for the high-magnification view, 20 µm. (D) Cathepsin B immunoreactivity and ProSense signals at the tumor periphery in the tumor shown in Figure 1. Left: H&E. Middle: ProSense (red) at the area of the H&E-stained view with tissue autofluorescence representing connective tissue (green). Right: Cathepsin B (green) with nuclear DAPI staining (blue). The dotted lines indicate identical portions of the tumor stroma. Scale bar, 100 µm. (A–D) Tumors from at least five animals examined showed similar results.

In conjunction with cathepsin B immunoreactivity, CD68⁺ cells distributed mostly at the tumor periphery (Figure 4A). There was a good correlation between CD68 with cathepsin B in a scattergram. However, scattergrams of α-smooth muscle actin (SMA)⁺ or CD31⁺ cells with cathepsin B showed a “two-tailed” split [23], reflecting a dissociation of their distribution with cathepsin B (Figure 4, A–C). Imaging of double-stained cathepsin B and CD68 tissue under high magnification (x20) also revealed identical cellular distribution of these immunoreactivities. Furthermore, immunofluorescence of cathepsin B with other leukocyte markers, CD45, CD11b, Gr-1, or F4/80, in serial sections revealed that cathepsin B-expressing cells mostly coexpress all of these markers similarly with CD68 (Figure W4). Again, CD31 and α-SMA did not colocalize with cathepsin B, except for the colocalization of α-SMA with cathepsin B in blood vessel walls (Figure 4, A–C). Intensity profiling data showed similar distribution patterns between cathepsin B and CD68 or CD31, corresponding to the profile of ProSense signals and vascular probe signals within the tumor tissue (Figure 2F and Figure 4, B and C). α-SMA immunoreactivity, however, did not show a similar intensity profile (Figure 2F and Figure 4, A and C). Colocalization of ProSense signals was observed with immunofluorescence of cathepsin B or CD68 in identical tissue section imaging (Figure 4D).

Figure 4.

Colocalization analysis of immunofluorescence images for cathepsin B and markers for macrophages, myofibroblasts, and endothelial cells in the HT29 tumor. (A–C) Immunoreactivities for cathepsin B (green) with CD68, α-SMA, or CD31 (red) in the tumor shown in Figure 3D. Left: Whole-tumor view with nuclear DAPI staining (blue). Middle left: High-magnification (x20) view of the immunofluorescent signals. The dotted lines indicate identical portions of the tumor stroma. Middle right: Scattergram data from the whole-tumor images. Right: Histogram of the fluorescent intensity from the whole-tumor images (arrow). Scale bar for the whole-tumor view, 1 mm; for the high-magnification view, 50 µm. (D) ProSense signals and immunoreactivities of cathepsin B and CD68 in an identical section. Left: ProSense signals. Middle left: Cathepsin B immunoreactivity. Middle right: CD68 immunoreactivity. Right: Merged image. Scale bar, 50 µm. T, tumor; M, muscle; BV, blood vessel. (A–D) Tumors from at least three animals examined showed similar results.

In vitro studies using cultured cells revealed that RAW264.7, a murine macrophage-derived cell line, showed high ProSense signals and strong cathepsin B immunoreactivity, whereas HT29 or HCT116 cells showed low ProSense signals and weak cathepsin B immunoreactivity (Figure 5A and data not shown). Coculture experiments of HT29 or HCT116 cells with RAW264.7 cells showed that ProSense signals were increased with the increase in the coculture ratio of CD45⁺ cathepsin B⁺ RAW264.7 cells (Figure 5, B and C, and data not shown). ProSense signal of RAW264.7 cells was inhibited by pan-cathepsin inhibitor, E64d, or cathepsin B inhibitor CA074Me in a dose-dependent manner (Figure 5, D and E).

Figure 5.

ProSense signals in cultured HT29 human colon tumor cells or RAW264.7 murine macrophage-derived cells. (A) ProSense signals (red; upper panels) and cathepsin B immunoreactivity (green; lower panels) with nuclear Hoechst staining (blue) in the single-cell culture. Left: RAW264.7 cell image. Middle: HT29 cell image with the same imaging parameters as RAW264.7 cell image. Right: HT29 cell image with enhanced fluorescent intensities. Scale bar, 10 µm. (B) Localization of ProSense signals and immunoreactivities of cathepsin B and CD45 in the coculture of HT29 and RAW264.7 cells. Left: ProSense signals (red) with nuclear Hoechst staining (blue). Right: Immunoreactivities of cathepsin B (green) and CD45 (red) with nuclear Hoechst staining (blue). To avoid detecting cocultured RAW264.7 cells as tumor cells, cells were stained with anti-CD45 antibody by immunofluorescence. Scale bar, 10 µm. (C) ProSense signals (red) in the coculture of HT29 and RAW264.7 cells with different cell number ratio for cocultivation. Nucleus was counterstained with Hoechst (blue). Left: Single-cell culture of HT29 cells. Middle left: Coculture with 1:1 of HT29 cell-to-RAW264.7 cell ratio. Middle right: Coculture with 1:2.5 of HT29 cell-to-RAW264.7 cell ratio. Right: Coculture with 1:5 of HT29 cell-to-RAW264.7 cell ratio. Scale bar, 200 µm. (D) ProSense signals of RAW264.7 cells in the presence of pan-cathepsin inhibitor E64d. Left: Vehicle. Middle left: 5 nM. Middle right: 25 nM. Right: 50 nM. Scale bar, 10 µm. (E) ProSense signals of RAW264.7 cells in the presence of cathepsin B inhibitor CA074Me. Left: Vehicle. Middle left: 5 nM. Middle right: 25 nM. Right: 50 nM. Scale bar, 10 µm.

Tumor Proliferation and Tissue-Remodeling Activity in a Xenograft Tumor Model

IHC for proliferating cell nuclear antigen (PCNA) revealed higher proliferative activity of tumor cells at the tumor periphery than at the center (Figure 6A). Collagen IV immunoreactivity was rarely observed at the tumor margin, in contrast to the scattered distribution in stromal cells within tumor parenchyma. Unlike collagen IV, tenascin C immunoreactivity was observed at the tumor margin. Gelatinase activity also showed a similar pattern of immunoreactivity as tenascin C by in situ zymography of DQ gelatin (Figure 6B). In situ zymography of ProSense revealed that ProSense cleavage was observed at the collagenous tissue components at the tumor periphery (Figure 6C).

Figure 6.

Proliferation and tissue-remodeling activity in the HT29 tumor. (A–C) Immunoreactivity of PCNA and extracellular matrix proteins and in situ zymography in the tumor shown in Figure 3D. (A) PCNA. Left: Whole-tumor view. Middle: High-magnification (x20) view at the tumor periphery. Right: High-magnification (x20) view at the tumor center. Scale bar for the whole-tumor view, 1 mm; for the high-magnification view, 50 µm. (B) Immunoreactivity for collagen IV and tenascin C and in situ zymography of DQ gelatin at the tumor periphery with nuclear DAPI staining (blue). Left: Cathepsin B (green) and collagen IV (red). Middle: Tenascin C (red). Middle right: DQ gelatin-cleaving activity (green). Scale bar, 50 µm. (C) In situ zymography of ProSense. Left: ProSense-cleaving activity (red) with nuclear propidium iodide staining (blue). Middle: High-magnification (x20) view of ProSense-cleaving activity. Right: High magnification (x20) of the H&E-stained view of the tumor corresponding to the fluorescent image. Scale bar, 50 µm. The dotted lines indicate identical portions of the tumor stroma. The arrowheads indicate the tumor margin at the tumor periphery. T, tumor; M, muscle. (A–D) Tumors from at least three animals examined showed similar results.

ProSense Signals in a Rat Colon Carcinogenesis Model

Ex vivo imaging detected ProSense signals in rat colon tumors induced by AOM (Figure 7A). The whole-tumor view at the cut surface showed ProSense signals distributed at the tumor margin showing high vascular density and increased vascular leakage (Figure 7B). These ProSense signals were localized within the cell cytoplasm (data not shown). Double staining of cathepsin B and CD68 showed cathepsin B⁺ macrophages had infiltrated the tumor margin, corresponding to ProSense signals within the tumor (Figure 7, B and C). α-SMA and CD31, however, did not show colocalization with cathepsin B, except for the colocalization of α-SMA with cathepsin B in blood vessel walls. In some tumors, tumor cells appeared to express cathepsin B, in contrast to the lack of cathepsin B expression in the adjacent normal colon mucosa.

Figure 7.

ProSense imaging in AOM-induced rat colon carcinogenesis model. (A) Ex vivo imaging of ProSense in a colon tumor. Left: Bright field. Middle: AngioSense. Right: ProSense. Scale bar, 2 mm. (B) Cut surface imaging of the colon tumor. Left: Tomato lectin. Middle: AngioSense. Right: ProSense. Scale bar, 1 mm. (C) Immunoreactivities for cathepsin B (green) with CD68, α-SMA, or CD31 (red) and nuclear DAPI staining (blue) at the whole-tumor view (upper panels) or the high-magnification (x20) view (lower panels). Left: H&E-stained view of the tumor corresponding to the fluorescent image as shown in B. Middle left: Cathepsin B and CD68. Middle right: Cathepsin B and α-SMA. Right: Cathepsin B and CD31. Scale bar for the whole-tumor view, 500 µm; for the high-magnification view, 50 µm. The asterisk indicates a necrotic area. (A–C) Tumors from at least seven animals examined showed similar results. All images are from the same tumor.

Discussion

In this study, we observed a time-dependent accumulation of Angio-Sense in xenograft tumors by selective leakage from the microvasculature identified by tomato lectin. This is suggestive of an enhanced permeability and retention (EPR) effect of macro molecules [24]. We also observed a slow accumulation of ProSense signals in the areas showing EPR effects within tumors. These data suggest that Pro- Sense probes are delivered to tumors in this manner. Alternatively, it is possible that ProSense probes could be effectively delivered to, and accumulate in, areas with increased vascular leakage similar to AngioSense probes. Furthermore, we observed this EPR effect in the rat colon carcinogenesis model, which partially recapitulates clinical carcinogenesis. Our results suggest that EPR effect-based imaging agents like ProSense can be applicable in clinical settings for tumor detection.

We also observed that the cytoplasmic localization patterns of Pro- Sense signals showed a similar distribution with cathepsin B⁺ macrophages in tumor xenografts. The accumulation of ProSense signals has also been reported in conjunction with intestinal polyp accumulating macrophages in a genetically engineered mouse model [25]. ProSense⁺ macrophages were also found in mouse xenograft gliomas [26]. In the present study, we also found that cultured cells of RAW264.7, a macrophage-derived cell line, revealed cathepsin B expression and high ProSense signals. These results suggest that ProSense⁺ cells in our tumor xenografts were cathepsin B⁺ macrophages. We further found extracellular ProSense-cleaving activity in areas with accumulating cathepsin B⁺ macrophages. Considering the cytoplasmic localization of ProSense signals, we suggest that ProSense probes are incorporated into macrophages after cleavage by cathepsin B in the extracellular space. ProSense signals were also observed in the liver Kupffer cells and cultured RAW264.7 cells, suggesting that phagocytosed uncleaved ProSense could also be cleaved within lysosomes.

It has been reported that tumor cells could contribute to the accumulation of ProSense probes [15]. However, the xenograft tumor models used in the present study showed only low ProSense signals, consistent with the faint expression of cathepsin B we observed in colon tumor cells. In the AOM-induced colon tumors, although we could not detect ProSense signals in tumor cells, cathepsin B was expressed in some tumor cells. Together, these data suggest that macrophages are the major source of ProSense signals in tumor tissues, consistent with previous reports [27]. Therefore, ProSense signal intensity is likely determined by the degree of macrophage infiltration.When using these probes to detect tumors, it may be important to address not only the level of cathepsin protease activity but also the endocytic activity of the cells [28].

ProSense signals and cathepsin B⁺ macrophages in the tumor xenografts were mainly distributed at the tumor periphery, where we observed a higher density of vasculature and PCNA⁺ tumor cells than in the center. Although we also observed the localization of tenascin C and gelatinase activity at the tumor margin, collagen IV did not accumulate in the same way as reported by others [29]. Tenascin C is an extracellular matrix glycoprotein that has a role in tissue-remodeling [30]. In the present study, we further found increased extracellular ProSense cleavage at the tumor periphery. Although tumor xenografts did not show apparent invasive growth into the surrounding tissues, these results suggest that tissue-remodeling occurred at the tumor periphery in response to tumor proliferation. We also found that cathepsin B⁺ macrophages were mainly distributed around blood vessels at the tumor periphery. Macrophages that aggregate around the tumor vasculature function in tissue-remodeling [31,32]. Therefore, our results suggest that, in addition to the EPR effect, the increase in tissue-remodeling involving cathepsin B⁺ macrophage infiltration is a potential mechanism for the accumulation of ProSense signals within tumors.

We also evaluated how inflammation in the tumor microenvironment may affect ProSense accumulation within tumors using an immunocompetent AOM-induced rat colon carcinogenesis model. This immunocompetent model renders a natural tumor-host interaction. Notably, we could detect ProSense signals in conjunction with increased vascular leakage at the tumor stroma, especially at the tumor margin showing macrophage infiltration. Recent reports using immunocompetent animal models have also detected tumors by enhancing Pro- Sense signals at the tumor margin [14,33]. In human colorectal cancers, invasive fronts express cathepsin B, showing high activity to facilitate tissue-remodeling [34,35]. Although macrophage infiltration is positively correlated with a favorable outcome in some human colorectal cancers, macrophages could be protumorigenic or antitumorigenic depending on the degree of cellular contact with neoplastic cells [36–38]. These data suggest that ProSense can effectively detect cancer invasive fronts where tissue-remodeling activity is high to facilitate neoplastic cell invasion.

In conclusion, the higher ProSense delivery and uptake at the tumor periphery are likely the result of the infiltration of cathepsin B⁺ macrophages in the stroma and the high-density vasculature. We also demonstrated that ProSense signal activation occurs in the extracellular space, and then the activated molecules accumulate in the macrophages. This suggests that ProSense is a clinically valuable tool for the detection of malignant tumors that require tissue-remodeling. Therefore, the clinical introduction of intraoperative protease imaging may have the potential to improve the accuracy of tumor resections.