Microdistribution of Targeted, Fluorescently Labeled Anti–Carcinoembryonic Antigen Antibody in Metastatic Colorectal Cancer: Implications for Radioimmunotherapy

Radioimmunotherapy involves the use of antibodies against tumor-associated antigens to target therapeutic radionuclides selectively to the tumor site. This reduces the toxicity to normal tissues and also allows access to metastatic tumor deposits that might otherwise remain occult and inaccessible to external beam radiotherapy. Radioimmunotherapy has been implemented effectively for the treatment of non-Hodgkin's lymphomas (1–3), but except for some promising results using pretargeting (4, 5), its success in the therapy of solid cancers has generally been limited (6–11). However, there is now evidence that radioimmunotherapy might be highly beneficial in metastatic colorectal cancer. Many earlier clinical studies described radioimmunotherapy in patients with advanced or progressive disease, when large liver deposits were present (8, 10, 11). However, recent studies are showing encouraging results for treatment of small metastatic lesions and as an adjuvant therapy in both disseminated liver and peritoneal disease (12–16). This suggests the potential for greater use of radioimmunotherapy in these settings, but there is a need to understand and optimize the system for improved efficacy before it can become a standard treatment.

To develop optimization criteria and improve the efficacy of radioimmunotherapy for metastatic disease, it is imperative that antibody distribution within tumor deposits, as well as the factors that influence it, is more clearly understood. This can only be fully researched by investigating the microdistribution of antibody over time in relation to tumor pathophysiology. Unfavorable conditions within the tumor environment, mainly defined by an abnormal tumor microcirculation, clearly affect the localization of antibody. The vasculature in solid tumors is characterized by immature, structurally inadequate microvessels with a number of severe functional and structural abnormalities, which are responsible for the chaotic and heterogeneous blood flow and a considerably nonuniform distribution of oxygen and other factors (e.g., lactate, glucose, etc.). Such heterogeneous tumor pathophysiology significantly affects the delivery of, and response to, systemically administered treatments including radioimmunotherapy.

Most previous radioimmunotherapy studies that have investigated the tumor distribution of radiolabeled antibody are based on autoradiographic and radioluminographic data, which is at low resolution and therefore lacks detailed information (17–19). Using fluorescently labeled antibodies for the same purpose has the advantage of generating high-resolution digital microscopy images, in which antibody distribution can be analyzed quantitatively over time in relation to tumor parameters known to be influential, such as blood vessel and antigen distribution and vascular perfusion (20, 21).

The current study was designed to investigate the hypothesis that efficient antibody localization in a model of colorectal cancer metastatic disease is likely to be related to the size of tumor metastases. Confirmation that there is a localization advantage for smaller tumor deposits would make a strong case for the utilization of radioimmunotherapy in an adjuvant setting for the treatment of metastatic colorectal cancer. This study is the first to investigate and quantify the relationship of antibody microdistribution concomitantly with associated tumor factors, over time, in a clinically relevant orthotopic model of colorectal cancer metastases.

The carcinoembryonic antigen (CEA)–expressing human adenocarcinoma cell line LS174T was grown under standard laboratory conditions as an adhesion culture. Subconfluent cells in logarithmic growth were trypsinized (Trypsin-Versene solution, BioWhittaker), counted, and resuspended in serum-free medium at the required concentration (20 × 10⁶/mL).

Mice were anesthetized with halothane; the abdomen was swabbed with chlorhexidine and s.c. Temgesic analgesic (buprenorphine) was locally administered. The spleen was exteriorized through a vertical 1- to 1.5-cm subcostal incision in the left abdominal wall and LS174T cells (1 × 10⁶ in 0.05 mL of serum-free medium) were injected. After 2 to 3 min, the splenic vessels were ligated and a splenectomy was done. The inner wound was closed by Vicryl-40 stitches and the outer wound with metal clips, which were removed after 9 d (22).

All experiments were in compliance with the United Kingdom Coordinating Committee on Cancer Research Guidelines for the Welfare of Animals in Experimental Neoplasia.

Characterization of the metastatic model. Characterization of the metastatic model was carried out on 3-μm-thick sections from formalin-fixed paraffin-embedded tissues.

Morphologic Appearance. H&E staining was carried out to assess the morphologic appearance and distribution of metastatic tumor deposits.

CEA Distribution by Immunohistochemistry. CEA distribution was shown in sections by applying biotinylated anti-CEA antibody (A5B7) and then avidin-biotin complexes (ABC, Vector Laboratories Ltd.; ref. 17).

A5B7, a monoclonal anti-CEA antibody, was used for both biodistribution and therapy studies (23–25). This antibody and its fragments are in regular preclinical and clinical use and have shown efficacy in radioimmunotherapy and combined antivascular studies within our department (7, 26, 27).

Antibody labeling. A5B7 (2 mg/mL) was labeled with Alexa Fluor 546 Protein Labeling Kit (Invitrogen Ltd.) according to the manufacturer's instructions. Fluorescently labeled antibody was then separated from unincorporated dye by gel filtration on disposable purification columns supplied with the kit.

Antibody biodistribution. All mice were injected i.v. into the tail vein with 100 μg of fluorescently labeled A5B7 anti-CEA antibody. Groups of mice were then culled at the following time points: 10 min (two mice), 1 h (three mice), and 24 h (three mice). Livers were removed, snap-frozen in isopentane (cooled over liquid nitrogen), sectioned at 12 μm, and stored at −80°C.

CEA distribution by fluorescence microscopy. To confirm that injected antibody was binding to its antigen, cryostat sections were stained for CEA using biotinylated anti-CEA antibody (A5B7) at 1:50 dilution and secondary reagent Alexa Fluor 488 streptavidin (Invitrogen) at 1:200.

A total of 24 tumors were examined for each of the following time points: 10 min, 1 h, and 24 h after injection of antibody. Cryostat sections were used to study the distribution of fluorescently labeled A5B7 anti-CEA antibody in relation to selected tumor-related parameters using the following markers:

• (a) Perfusion. The in vivo DNA-binding dye Hoechst 33342 (Invitrogen) was injected i.v. at a concentration of 10 mg/kg, 1 min before sacrifice. The dye leaves perfused vessels and stains the nuclei of adjacent cells; it can be viewed directly under UV filter.

• (b) Tumor vasculature. An anti-CD31 antibody was used to visualize blood vessel distribution within and around the tumor deposits and the relevant staining procedures were done (19).

• (c) Antibody distribution. A5B7 labeled with Alexa Fluor 546 shows the localization of antibody within tumor deposits. It can be viewed directly under a rhodamine filter.

• (d) Morphology. H&E staining of the same or adjacent section was done to analyze the distribution of fluorescent markers in relation to tumor architecture.

Stained sections were viewed under a fluorescence microscope Axioscope2 (Zeiss). Images were captured with an AxioCam digital color camera using KS300 software (Zeiss). All three fluorescent parameters were investigated in the same section by switching to the appropriate wavelength filter for each of the markers. Perfusion was viewed with a UV filter (365 nm excitation), CD31 fluorescence with an FITC filter (450-490 nm excitation), and antibody distribution with a rhodamine filter (546 nm excitation). Whole liver sections were montaged with a 10× objective whereas individual/groups of tumor deposits were montaged with a 20× objective. High-power single images were taken with a 20× or 40× objective. The same, or an adjacent, tumor section was stained with H&E and scanned under a bright-field microscope. The fluorescent images were then coregistered using Adobe Photoshop software, resulting in a new multichannel image exhibiting overlapping fluorescently labeled structures for the different parameters.

Superimposed images were analyzed using AxioVision 4.6 software, taking a number of measurements for each single tumor deposit: total tumor area; perimeter; longitudinal and cross-sectional distances; the mean density of red (A5B7); green (blood vessels), and blue (perfusion) pixels within the deposit; and the depth of antibody penetration into the tumor deposit along eight different profiles (i.e., distance from the edge of the tumor deposit to the limit of its movement). The measurements of maximum distance of antibody penetration into the tumor were based on generating a spectrum profile.

The obtained data were analyzed with nonparametric statistical tests (Spearman rank correlation test and Mann-Whitney U test) using SPSS software.

Orthotopic liver metastases developed within 2 to 3 weeks after injection of LS174T cells. Tumors presented as deposits with a range of sizes, surrounded by normal liver tissue (Fig. 1A). Some deposits consisted of only viable tumor tissue. These tended to be the smaller tumor deposits, although some larger tumors with no obvious necrosis were also observed (Fig. 1A). The majority of deposits, however, showed viable tumor of variable width at the periphery, with a necrotic center.

In total, 72 selected tumor deposits were analyzed with areas ranging from 0.0042 to 3.011 mm². The number and size range of deposits analyzed at each time point are shown in Table 1.

CEA distribution. CEA was expressed only in tumor deposits, with no evidence of antigen presence in normal liver (Fig. 1B, C, and E). H&E images show the morphologic appearance of the same metastases, which were composed of viable tumor (Fig. 1A and F).

Pattern of blood vessel distribution and perfusion in tumors. Each tumor deposit was surrounded by a well-defined vascular capsule, which differentiated it from normal liver. In general, the metastases themselves were not well vascularized, although some possessed their own blood supply.

Hoechst staining showed that LS174T orthotopic tumors contained both well-perfused and poorly perfused areas. The vascular capsule separating normal liver and tumor was generally well perfused, whereas the tumor itself showed little or no perfusion. This correlated with the sparsity of vessels observed within the tumor mass and the fact that not all these vessels showed evidence of perfusion even when present.

Relationship between tumor size and tumor blood vessel presence. The relationship between the size of tumor deposits in the liver and the presence of intratumoral blood vessels was investigated using the nonparametric Mann-Whitney U test. Data from all analyzed tumor deposits were included (n = 72). The presence of intratumoral blood vessels was positively correlated (P < 0.001) with the size of tumor deposits. The area of the smallest tumor containing intratumoral blood vessels was 0.051 mm² (diameter, 216 × 170 μm). The range in tumor area for those deposits without and with intratumoral blood vessels is shown at Fig. 2. Approximately 90% of tumors without blood vessels had an area of <0.3 mm².

The distribution of A5B7 antibody within the tumor deposits was examined at 10 minutes, 1 hour, and 24 hours postinjection.

The antibody showed rapid and selective localization within tumors of all sizes. Even at the earliest time point of 10 minutes, the antibody localization was highly specific for the tumor, with no evidence of accumulation within normal liver.

In general, there was a lack of blood vessels within the tumors. In metastatic deposits without intratumoral blood vessels, antibody seemed to localize by diffusion from the vasculature in the surrounding capsule. However, in those tumor deposits where intratumoral blood vessels were obviously shown, extravasated antibody could clearly be seen in close relationship to the perfused vessels (Fig. 3A-a).

Antibody distribution at 10 minutes. By 10 minutes, A5B7 was clearly visualized in all tumor deposits examined. The antibody was localized at the periphery of the medium and large tumors, whereas in some of the small deposits it had already diffused fully across the tumor (Fig. 3A-a, b). H&E images show two viable tumors and a third with small area of central necrosis (Fig. 3A-c). Several small viable tumor deposits are shown in Fig. 3A-d.

Antibody distribution at 1 hour. By 1 hour, the distribution of antibody had not altered significantly from that seen at 10 minutes, although in some of the larger tumor deposits the antibody showed slightly greater penetration into the tumor while still remaining localized at the periphery (Fig. 3B-a). Figure 3B-b shows a small collection of tumor cells with antibody localization throughout. The equivalent morphology for Fig. 3B-a shows two medium-sized tumors surrounded by normal liver, with one containing an area of central necrosis (Fig. 3B-c), whereas the H&E of Fig. 3B-b shows a group of viable tumor cells before the development of a vascular capsule (Fig. 3B-d).

Antibody distribution at 24 hours. By 24 hours, the antibody had diffused further into the tumor, occupying a larger area of the deposit (Fig. 3C-a). As seen at 10 minutes and 1 hour, the antibody had diffused, and had been retained, throughout the small tumor deposits (Fig. 3C-b). Corresponding H&E images show large and medium tumor deposits with varying degrees of central necrosis (Fig. 3C-c) and a small viable tumor surrounded by normal liver (Fig. 3C-d). When fluorescence images were related to the morphology in H&E-stained sections, it was observed that the antibody was, in general, still retained in viable tumor at 24 hours, although diffusion into the necrotic areas was also present in some larger tumors (Fig. 3C).

Analysis of antibody distribution was done on 72 tumor deposits, as described in Materials and Methods and as illustrated in Fig. 4.

Relationship between tumor area and antibody localization. The dependence of antibody uptake on tumor size was analyzed for the selected time points. The mean density of red pixels (antibody) within the tumor deposit on combined fluorescence images was correlated with tumor area, using a nonparametric statistical test (Spearman rank correlation). It showed a correlation coefficient of −0.690 at 10 minutes, −0.576 at 1 hour, and 0.141 at 24 hours after antibody injection. The correlation was significant at the 0.01 level for time points 10 minutes and 1 hour, showing that the smaller tumors had significantly more antibody localization, as a percentage of their total area, than the larger tumors at these early time points. However, by 24 hours the difference was no longer significant.

Relationship between antibody uptake and the presence of tumor blood vessels. The relationship between antibody uptake and the presence of intratumoral blood vessels was analyzed with the nonparametric Mann-Whitney U test. It showed significantly (P < 0.05) higher uptake in tumors without intratumoral blood vessels.

Relationship between tumor size and antibody diffusion distance. Analysis of antibody travel distance for small (average diameter ≤125 μm) and larger (average diameter >125 μm) tumor deposits was done. Antibody seemed to travel into tumor deposits at the same rate in both small and large lesions. The median distance traveled was 50.02/52.62, 64.55/52.13, and 73.35/423.69 μm at 10 minutes, 1 hour, and 24 hours for small/large tumors, respectively (Fig. 5A). When antibody diffusion distance was analyzed as a percentage of tumor radius, it was shown that small deposits localized more rapidly with antibody than the larger ones (Fig. 5B).

This novel study investigates the kinetics of fluorescently labeled anti-CEA antibody in a clinically relevant orthotopic colorectal cancer model. It is the first time that the relationship between antibody localization and size of tumor deposit has been evaluated by using high-resolution fluorescence images to quantitatively map the rate and distance of antibody travel over a time course of 24 hours. Most previous studies addressing antibody pharmacokinetics have used s.c. models, which, unlike the metastatic model, do not adequately represent the clinical situation (23, 28–31). Where orthotopic models have been used (32), antibody pharmacokinetics was studied using autoradiograph of radiolabeled antibody, which is far less sensitive than high-resolution microscopy (resolution, ∼50 versus 12 μm respectively).

Antigen distribution and concentration are important factors in determining radioimmunotherapy efficacy because they influence antibody localization. CEA was expressed in all tumor deposits, and A5B7 localized efficiently and selectively to all these CEA-positive regions (Fig. 2C-E). In comparison, the fluorescently labeled, isotope-matched nonspecific antibody MOPC showed such low uptake in both metastatic and s.c. LS174T tumors that we were unable to image the distribution (21). Further studies to show specificity of antigen binding, using a non–CEA-expressing colorectal tumor, were not possible because the cells did not grow in the orthotopic setting. This may be due to the fact that CEA is thought to be involved in the metastatic process, although the exact mechanism is not yet clear (33).

Liver metastases showed a paucity of blood vessels compared with those seen in the s.c. model (20, 25). However, liver is an extremely well-perfused organ, and once tumors had grown beyond a few cells, they developed a highly vascular capsule around each deposit (Fig. 3B-a). This seems to originate from liver tissue, can provide a good supply of oxygen and nutrients, and helps to explain why even tumors lacking intratumor vessels showed excellent antibody localization. Even the microscopic clusters of cells, before capsule development, showed efficient antibody uptake (Fig. 3B-b). Some tumors did, however, contain intratumor vessels (Fig. 3 A-a, c and C-a, c), which seem to develop as branches from the surrounding vascular capsule. It is not yet clear what tumor factors initiate the growth of these vessels; preliminary immunofluorescence studies showed staining for the proangiogenic vascular endothelial growth factor on some vessels, but the majority showed no reactivity. Further investigations are ongoing. Where intratumor vessels were observed, antibody showed both peripheral localization at the tumor edge and perivascular central distribution at 10 minutes and 1 hour (Fig. 3A-a), but had diffused throughout the tumor mass by 24 hours. The presence of blood vessels was dependent on size, the larger tumors possessing a more developed vascular network. In spite of this, however, tumors with intratumor vessels did not show higher relative antibody uptake, suggesting that diffusion from the surrounding vascular capsule is a major contributor. A further factor is that larger tumors contained more central necrosis with lower, and later, antibody localization than viable tumor regions (Fig. 3C).

The antibody showed rapid and selective uptake into all tumor deposits. However, there was a strong negative correlation between total A5B7 uptake and relative tumor size at early time points (10 minutes and 1 hour), with smaller tumors having significantly higher uptake. This probably relates to the higher surface area to volume ratios for the smaller deposits. By 24 hours, the correlation was no longer significant because sufficient time had elapsed for equilibration of antibody uptake throughout the majority of larger tumors (Fig. 3C-a). These results are consistent with those of previous studies done by ourselves and others in s.c. colorectal xenografts, where smaller tumors had relatively higher antibody uptake (23, 30, 34, 35).

The speed and distance of antibody diffusion over time were also investigated. Smaller deposits (≤125 μm in diameter) showed diffusion throughout the whole tumor mass at a much earlier time point than the larger tumors (10 minutes versus 24 hour; Fig. 5B). This was because the antibody traveled at the same rate across both small and large deposits, and therefore penetrated to the center of the smaller tumors more rapidly (Fig. 3A-b). Furthermore, the antibody was still retained throughout small tumor deposits at 24 hours. These data suggest that adjuvant radioimmunotherapy of micrometastases might be highly beneficial: high antibody uptake and homogeneous localization at early time points after injection, followed by significant retention over time. In addition, these small tumors are less hypoxic than larger deposits (not shown) and thus are likely to be more radiosensitive. This is in agreement with recent preclinical results on i.p. adjuvant radioimmunotherapy (12–14). The current study also shows the movement of antibody across larger tumor deposits with time, combined with increased retention and uniformity by 24 hours (Fig. 3C-a), also observed in our s.c. colorectal models (18). Efficient tumor retention was also aided by A5B7 affinity (∼7 × 10¹⁰ mol/L), with two binding sites for CEA. The antibody remains at high levels in blood for >48 hours, allowing significant tumor extravasation over this period. Although antibody distribution is only presented here up to 24 hours, we have also investigated fluorescent and radiolabeled A5B7 and MOPC at 96 hours in this model. Whereas both forms of A5B7 showed highly selective tumor retention at this time point (10% injected dose/g of ¹²⁵I-A5B7), the fluorescent nonspecific antibody was barely discernable and ¹²⁵I-MOPC never reached positive tumor/normal tissue ratios (22). The current data agree with those from s.c. LS174T (24) and clinical studies (7), and suggest that adjuvant radioimmunotherapy of metastases would be beneficial for both large and small deposits.

Investigating patterns of antibody movement over time for metastases of different sizes allows us to rationally design novel molecules with potential for improved radioimmunotherapy by matching antibody distribution to the range over which radiation will be distributed. Monovalent antibodies have faster circulatory clearance and weaker antigen binding than IgGs, resulting in lower tumor uptake and therapeutic potential but reduced systemic toxicity. It has also been suggested that they have the ability to localize more rapidly into tumor deposits (37, 38). However, our results show that there was no physical barrier to localization and penetration of IgG (MW 150 kDa) into and through hepatic tumor deposits of all sizes. Indeed, in larger deposits, the rapid circulatory clearance of antibody fragments may prevent accrual throughout all viable tumor regions; we have previously shown that uniform ¹²⁵I-A5B7 localization in this model can take up to 96 hours for some of the larger metastases (22).

When considering radioimmunotherapy for treating minimal residual disease, as in an adjuvant setting, the efficiency of tumor targeting must be weighed against systemic toxicity (39, 40). Selection of suitable radionuclide depends on both tumor setting and antibody used. A therapy study in this metastatic model, using 5.5 MBq of ¹³¹I-A5B7, produced significant tumor growth inhibition compared with untreated controls or the same dose of the nonspecific antibody MOPC, with two of six mice showing disease-free survival until culled at 100 days posttreatment (22). This indicates that the antibody pharmacokinetic profile (circulatory clearance and tumor diffusion) and the half-life (8 days) and energy of the isotope (¹³¹I) are reasonably well suited for treating colorectal metastases of different sizes. It also supports results from clinical studies using the monoclonal antibody ¹³¹I-labetuzumab (15, 16), showing evidence of a promising survival advantage in colorectal liver metastases. However, the fact that radioimmunotherapy did not eradicate tumors in all our mice indicates that the current system could be improved. Indeed, some α particles and low-energy β particles may be more suitable for tackling small tumor deposits and adjuvant disease. We have previously used phosphor images of radioantibody localization to model dose distribution for a range of isotopes (18, 19, 37). However, the novel microscopy in the current study has provided a far more accurate picture of antibody distribution in relation to tumor microenvironment and will now inform models of energy deposition for antibody-bound radionuclides to determine optimal pairings of antibody and radionuclide for different situations. Combination therapies may be required to treat larger deposits (24, 25) and are currently being investigated.

In conclusion, the clinical use of radioimmunotherapy for treating colorectal metastases in the adjuvant setting is showing a promising survival advantage (15, 16). To increase our understanding and optimize the therapy, we have used novel techniques to investigate the kinetics of fluorescently labeled anti-CEA antibody in a clinically relevant liver metastasis model. The ability to characterize these metastases both morphologically and functionally has provided unique data to optimize radioimmunotherapy, including more detailed microdosimetry studies. We have shown extremely rapid and selective antibody uptake in all metastases regardless of size, followed by prolonged retention. The data suggest that it may be preferable to use antibody fragments to treat deposits of ≤125 μm, reducing toxicity and allowing redosing, but it may be necessary to use intact antibodies with longer circulating half-lives for treating larger deposits, which will allow effective tumor penetration but increase systemic toxicity. This in turn informs the selection of suitable antibody-radionuclide combinations and allows optimization of future radioimmunotherapy clinical trials as a single agent or in combined therapy approaches.

We thank Prof. A. Mantovani (Instituto di Ricerche Farmacologiche, Milan, Italy) for providing the anti-CD31 antibody, Dr. A. Green for helpful advice, and R. Boden for technical assistance.