Safety and Biodistribution of ^99mTechnetium-labeled Anti-CD44v6 Monoclonal Antibody BIWA 1 in Head and Neck Cancer Patients

HNSCCs²account for ∼5% of all malignant neoplasms in Europe and the United States. Worldwide, 500,000 new cases are projected annually, and this incidence is rising. For 1998, it was estimated that 41,400 Americans would develop HNSCC, and 12,300 would die from it (1). Despite improvements in locoregional treatment modalities, i.e., surgery and radiotherapy, for stages III/IV (70%)there is still a high failure rate, either locally or at distant sites. An effective adjuvant systemic treatment is therefore needed to improve survival rates in this patient group. In this respect, application of(neo)adjuvant chemotherapy, with its unfavorable therapeutic index, has mostly failed to accomplish any improvement in survival (2, 3, 4).

Considering the clinical need and the inherent radiosensitivity of HNSCC (5), our department has put effort on the development of MAbs capable of targeting radionuclides to HNSCC (6). In particular with regard to the MAb called U36, in vivo imaging/biodistribution trials revealed favorable biodistribution patterns with selective tumor targeting and high tumor uptake of 20.4 ± 12.4% of the ID/kg of tissue (7).

The antigen recognized by U36 appeared to be identical to the keratinocyte-specific CD44 splice variant epican, which contains the variant exons v3–v10. By screening overlapping synthetic peptides of the epican-specific region encoded by exons 7–11 (v3–v7) the corresponding epitope was mapped, revealing its localization in the v6 domain (8). Expression of v6-containing CD44 variants has been observed in several types of tumors including SCCs of the head and neck, lung, skin, esophagus and cervix, as well as adenocarcinoma of breast, colon, lung, and stomach (6, 9). Among normal tissues, expression was observed only in a subset of epithelial tissues, e.g., skin keratinocytes, breast and prostate myoepithelium, and bronchial epithelium (6, 9). Besides that, soluble v6-containing CD44 variants have been detected in the blood of control volunteers as well as of cancer patients (10). Possible molecular functions of CD44 isoforms are defined currently as adhesion molecules (11), signal transducers (12), regulators of cell migration (13), and as tumor metastasis-promoting proteins (14). Interestingly, v6-containing CD44 isoforms appeared to be capable of conferring metastatic potential on originally nonmetastatic tumor cells in a rat pancreatic carcinoma model (14). Moreover, MAbs against CD44v6 were able to prevent outgrowth of metastases in a syngeneic rat model (15). Overexpression of CD44v6 in tumors was shown to correlate with reduced survival of patients with breast and colon cancer and with non-Hodgkin’s lymphomas (16, 17, 18).

Alternative MAbs recognizing v6-containing CD44 isoforms have been developed, including MAb 17, MAb Var3.1, and MAb VFF18 (MAb BIWA 1;Refs. 9, 19, 20). BIWA 1 was selected from a panel of CD44v6-specific MAbs because of its high affinity to human tumor cells (9). Although BIWA 1 resembles U36, it binds to a different epitope and with a 35-fold higher affinity. According to the numbering of Kugelman et al. (21), the epitope recognized by U36 consists of amino acids 365–376, and the epitope recognized by BIWA 1 consists of amino acids 360–370. Because high-affinity MAbs may be better suited for tumor targeting (22, 23), we decided to evaluate ^99mTc-labeled BIWA 1 in a similar clinical RIS setting as performed previously with ^99mTc-labeled U36 (7). As a prelude to RIT, this report describes the results of a first Phase I clinical trial aiming to determine safety, kinetics, tissue distribution, tumor uptake, and diagnostic effectiveness of ^99mTc-labeled BIWA 1 in 12 patients with HNSCC. These objectives were evaluated at three different BIWA 1 dose levels.

Twelve patients (5 women and 7 men; age range, 41–76 years) with histologically proven advanced HNSCC (4 patients had stage III disease;8 had stage IV) were entered after informed consent was obtained (Table 1). The protocol was approved by the Institutional Review Board. Patients who participated in this study planned to undergo a resection of the primary tumor and a unilateral or bilateral neck dissection. Surgery was performed 40 h after administration of ^99mTc-labeled BIWA 1. For one patient, surgery was canceled because of disclosure of multiple metastases in the lungs and possibly in the liver. Prior to enrollment, a biopsy of the primary tumor had to show BIWA 1 binding to >50% of the tumor cells, as determined by immunohistochemistry, which consistently turned out to be>95% of cells.

A large number of hematological and biochemical parameters were measured from blood and urine samples obtained <3 weeks before injection, as described previously (7). To assess toxicity, these were compared with values obtained 21 and 40 h and 1 week after injection. Vital signs (i.e., blood pressure,pulse rate, breathing rate, and temperature) were recorded immediately before administration of ^99mTc-labeled BIWA 1 and each 20 min up to 2 h after injection.

For the assessment of tumor volumes, MRI data on optical disc were processed by in-house developed software run on a standard Sparc 10 workstation (SUN Microsystems, Palo Alto, CA). Tumor areas were calculated after displaying separate MRI images on a monitor on which visible primary tumor was manually enclosed into a region. Surfaces of each region were multiplied by slice thickness plus interslice gap,thus obtaining total tumor volume.

Characterization of the CD44v6-specific murine IgG1 MAb BIWA 1(Boehringer Ingelheim, Vienna, Austria) and generation of the BIWA 1 hybridoma cell line (also called VFF-18) have been described before (9). Clinical grade material was produced by Bio-Intermediair Europe BV, Groningen, the Netherlands. BIWA 1 (2 mg)was labeled with ^99mTc using the chelate S-benzoylmercaptoglycylglycylglycine(S-benzoyl-MAG3; Mallinckrodt Medical B.V., Petten, the Netherlands), and the immunoreactive fraction of each radiolabeled BIWA 1 preparation was tested as described previously (7,). As determined by a modified Lineweaver-Burk plot, the immunoreactive fraction of ^99mTc-labeled BIWA 1 was 88.3 ±5.5% at infinite antigen excess. TLC demonstrated a mean of 97.0 ± 1.6% of ^99mTc to be bound to BIWA 1. Unlabeled BIWA 1 was added to adjust the amount of MAb for the 2-, 12-,and 52-mg dose groups. BIWA 1 was subsequently injected i.v. over a 5-min period. The low-, middle-, and high-dose patient groups received 2.0 ± 0.2, 10.7 ± 1.0, and 54.5 ± 5.0 mg of BIWA 1,respectively, labeled with a mean activity of 768.5 ± 45.9,777.7 ± 15.2, and 728.9 ± 28.1 MBq of ^99mTc, respectively.

All patients were preoperatively examined by palpation, CT, and MRI of the neck, as reported previously (7). At 0–1 and 21 h after injection, planar whole body images (anterior and posterior projection) and 21 h postinjection planar spot images of the head and neck (anterior, posterior, and lateral projections) were acquired. An ADAC dual-headed gamma camera equipped with a low-energy collimator was used. SPECT images of the head and neck were obtained 21 h after injection. From the one patient whose distant metastases were disclosed shortly before the actual surgical procedure, additional SPECT images were obtained of chest and upper abdomen, 21 h after injection.

CT, MRI, and RIS images were each reviewed by the same experienced examiner, blinded to the results of other examinations and the pathological outcome, although notified of the primary site. Criteria for the optimal assessment of cervical lymph node metastases by CT or MRI were applied (25). Interpretation of RIS images was based upon asymmetry and retention.

Diagnostic effectiveness of RIS and conventional anatomical imaging modalities were compared. To this end, diagnostic results of palpation,CT, MRI, and RIS were evaluated per neck side and lymph node level, as described previously (7). Histopathological examination of the neck dissection specimens was used as the gold standard. Hereto,detectable lymph nodes were dissected from each surgical specimen and examined by a pathologist. All diagnostic modalities were correlated with the histopathological examination, and their diagnostic effectiveness was expressed in terms of sensitivity and specificity. Sensitivity is defined as TP/(TP + FN)and specificity as TN/(TN + FP), where TP is a true-positive, FN a false-negative, TN a true-negative, and FP a false-positive observation.

Biopsies were taken from tumor tissue (primary site and lymph node metastases) and from various normal tissues, present in the surgical specimen. Blood, bone, and bone marrow were obtained immediately before surgery under general anesthesia. From the patient not proceeding to surgery, only a sample of tumor and normal mucosa was obtained. Tissues and blood samples were weighed, the amount of ^99mTc was measured in a gamma-well counter (1282 Compugamma, Wallac, Turku, Finland), and data were converted to %ID/kg tissue, as described previously (7). Tumor:nontumor ratios were calculated using matched uptake values of one patient. If in a patient several biopsies of one kind of tissue were taken, the mean uptake in this tissue was calculated and used for further analysis. After counting, all biopsies were evaluated histopathologically to determine the presence or absence of HNSCC.

Uptake of MAb in tissues not present in the surgical specimen was assessed by drawing ROIs on planar anterior and posterior whole-body images, obtained 0–1 and 21 h after injection. Regions included whole-body, liver, left kidney, spleen, two lumbar vertebrae, heart(left ventricle), and right lung. Geometric means were calculated. Standards were used to correct for decay of activity and camera efficiency. Geometric mean activity within the whole-body region 0–1 h after administration was designated the ID. Activity within each ROI was corrected for background and used to obtain the relative organ activity, expressed as %ID per 50 pixels. By conversion to the total pixels for these organs, the activity was also expressed as %ID/whole organ.

Microscopic distribution of BIWA 1 throughout the tumor was assessed by immunohistochemical analysis of tumor tissue obtained from the surgical specimen. Frozen, acetone-fixed serial sections were stained applying the biotin-avidin-peroxidase method after incubation with biotinylated secondary antibody [F(ab′)₂ rabbit antimouse IgG, DAKO, Denmark]. For assessment of maximal binding, sections were incubated with BIWA 1, followed by the second antibody. The staining intensity was categorized into: negative, weak, and moderate to strong. Moreover, the percentage of CD44v6-positive cells having bound administered BIWA 1 was estimated.

Serial blood samples were drawn at various time points after completion of infusion. Urine was collected in portions of 24 h until the moment of surgery to determine renal excretion of ^99mTc. The amount of radioactivity in blood,plasma, and urine was assessed in a gamma-well counter and expressed as%ID/kg as described previously (9). Noncompartmental pharmacokinetic parameters were calculated from concentration versus time data using the WinNonLin computer program,version 1.5 (Scientific Consulting, Inc). Size exclusion chromatography(silica-based gel filtration high-performance liquid chromatography) was applied to assess the percentages of radiolabeled free BIWA 1 (Rf 15.9 ± 0.2 min) and complexed BIWA 1 (Rf 11.1 ± 1.1 min), essentially as described formerly (26).

In addition, the concentration of total circulating BIWA 1 was measured by a murine IgG serum ELISA. This assay does not distinguish free from complexed BIWA 1. Furthermore, by means of an ELISA using plates coated with GST-CD44v3-v10 (Boehringer Ingelheim), the concentration of circulating immunoreactive BIWA 1 was assessed. Both ELISAs detect radiolabeled as well as unlabeled BIWA 1.

Levels of sCD44v6 in serum were assessed by use of a commercially available sandwich type ELISA (Bender MedSystems, Vienna, Austria)within 2.5 h before administration, 1 week after injection and 6 weeks after injection. A MAb specific for an epitope of the CD44 standard molecule served as the capturing MAb absorbed to microtiter plates. BIWA 1 coupled to horseradish peroxidase was used for signal generation by the peroxidase/tetramethyl-benzidine system. The ELISA test was performed according to the manufacturer’s instructions as described by Kittl et al. (27).

A baseline plasma HAMA titer was determined in all patients, with follow-up at 1 and 6 weeks after injection, using a HAMA titer assay (28). HAMA titers ≥500 were arbitrarily considered to be positive.

No adverse reactions to the radiolabeled BIWA 1 were observed in any patient. No clinically significant changes were noted in blood and urine analyses or vital signs.

At 21 h after injection, planar whole-body images did not demonstrate unexpected uptake in normal tissues. Besides activity uptake in tumor tissue, only minimal accumulation of activity was observed in mouth, lung, spleen, kidney, bone marrow, and scrotal area. Of 12 primary tumors, 11 were depicted by RIS at 21 h after injection, 8 on planar images, and another 3 with SPECT. Tumors measured 2.5–30.2 cm³ (mean, 19.4 ± 9.1 cm³; Table 1). The only tumor not visualized had a size of 26.4 cm³.

In 12 patients, 19 neck dissected sides contained 89 neck levels. According to histopathological examination, 10 sides and 18 levels contained metastases of HNSCC. Diagnostic results of the four modalities were analyzed per neck side and lymph node level, with histopathology as the gold standard (Table 2). RIS correctly detected lymph node metastases in 6 of 10 sides(sensitivity, 60%) and 9 of 18 levels (sensitivity 50%). Detected lymph nodes measured 0.8 × 2.5 up to 4.5 × 5.5 cm. Four affected sides containing 12 tumor-positive lymph nodes distributed over seven levels were not disclosed by RIS. Eleven of these lymph nodes, sizes 0.7 × 1.5 to 2.4 × 3.4 cm, were reexamined histopathologically and appeared to contain either substantial necrosis, keratinization, and/or fibrosis (n = 8) or micrometastasis (n = 3). Besides these localizations,another two tumor-involved lymph node levels (containing three metastases with minimal diameters <1 cm) were not detected by RIS.

Three sides in two patients were designated tumor-positive by RIS,although meticulous histopathological examination did not reveal any lymph node metastases (specificity, 67%). One of these sides was falsely scored positive by MRI and CT as well; the other 2 only by RIS. Histopathological re-examination demonstrated that these neck sides contained reactively enlarged lymph nodes. Of 71 tumor-negative lymph node levels, 5 were falsely scored positive by RIS (specificity, 93%).

In one patient, multiple metastases in both lungs (∼10; diameters <1 cm) and possibly in the liver were disclosed by CT just prior to the arranged surgical procedure. Additional SPECT imaging of chest and upper abdomen were performed. None of the lesions was visualized on SPECT images; however, these clinical metastases were never histopathologically confirmed.

Uptake of radioactivity in biopsied tumors and normal tissues at each dose level is shown in Table 3. From the one inoperable patient, only tumor, mucosa, and blood samples could be obtained. Corresponding tumor:nontumor ratios are shown in Table 4.

Analysis of tissue samples obtained 40 h after injection revealed high antibody uptake in tumors, with a mean value of 14.2 ±8.4%ID/kg and a mean tumor:blood ratio of 2.0 ± 1.4. Differences between the three dose groups were not statistically significant,although a trend toward lower uptake in the highest dose group was noted. Uptake in primary tumors was higher than in normal tissues,including mucosa and skin, which also strongly express the antigen. Uptake in affected lymph nodes was slightly higher than in normal lymph nodes but lower than in primary tumors.

Upon an increase of the MAb dose, activity within the whole body at 21.5 h after injection (range, 19.5 to 25.2 h) appeared similar: 91.6 ± 1.9, 93.8 ± 3.6, and 95.8 ± 5.6%ID. In Table 5, the relative organ activity per 50 pixels as well as per whole organ at 21 h after injection is illustrated for liver, lumbar vertebrae, kidney, spleen, heart, and lung. These appeared to be comparable for each dose group.

The in vivo uptake of ^99mTc-labeled BIWA 1 as determined by immunohistochemistry in biopsies taken 40 h after injection was fairly consistent within a dose group. Nevertheless, variations between different deposits from one patient and between different patients within a group were observed. The lowest uptake was observed in the low-dose group, revealing binding of the MAb to ± 20% of the tumor cells (Fig. 1, A and B). An increase in antibody uptake was found in the intermediate- and high-dose groups. In these two groups,more cells were stained (±30–40%, Fig. 1, C–F)than in the low-dose group, whereas staining was more intense. Upon dose increase from 12 to 52 mg, no obvious improvement of staining patterns was found. Therefore, although >95% of the tumor cells did express the antigen, maximally 40% of the cells had been targeted in vivo. In general, the MAb uptake appeared higher at the tumor-stroma interface and lower in the center of tumor islands.

In three cases, normal-appearing mucosa in the vicinity of the tumor was included in the biopsy specimens. Consistently, uptake of BIWA 1 into basal layers of the mucosa was found. Despite high expression of CD44v6 in suprabasal layers of normal mucosa, no staining of these layers with the antimouse IgG antibody was observed (Fig. 2).

Blood activity in time was best described by a noncompartment model(mean t_1/2, 34.5 ± 6.1 h). The clearance from the blood was not significantly influenced by the MAb dose; the t_1/2 values were 31.2 ±7.5, 35.6 ± 6.6, and 36.5 ± 3.7 for the 2-, 12-, and 52-mg dose groups. Within 24 h after injection, 12.5 ± 3.1%ID,13.8 ± 3.6%ID, and 11.1 ± 4.7%ID was excreted via the urine in the low, intermediate, and high dose, respectively. At the time of surgery, 40 h after injection, this was 17.4 ±5.0%ID, 18.0 ± 6.6%ID, and 15.3 ± 4.1%ID, respectively.

On HPLC analyses, each radioimmunoconjugate batch prepared for injection showed a monomeric IgG peak (Rf 16.0 ± 0.4 min). In plasma samples obtained immediately after administration, another peak was consistently detected at Rf 11.1 ± 1.1 min, representing a high molecular weight substance. Most likely, BIWA 1 had formed a complex with sCD44v6 present in the circulation. Concentrations of free and complexed BIWA 1 at different time intervals after MAb administration are shown for the three MAb dose groups in Fig. 3.

Samples drawn 5 min after administration of 2 mg ^99mTc-labeled BIWA 1 contained higher concentrations of complexed than of free IgG (Fig. 3,A). The percentage of complexed IgG increased from 63.1 ± 12.2% at 0.5 h after injection to 92.8 ± 2.0% at 43.4 h. After injection of 12 mg ^99mTc-labeled BIWA 1,relatively more BIWA 1 remained in free form (Fig. 3,B). For patients receiving 52 mg of ^99mTc-labeled BIWA 1,free IgG was even more prevalent (Fig. 3 C). In the 12-mg dose group, the mean percentages of complex formed increased from 13.0 ± 4.3% at 0.5 h after injection to 24.2 ± 10.5%at 43.3 h after injection and in the 52-mg dose group from 4.3 ± 0.6% at 0.5 h to 5.6 ± 1.8% at 43.2 h. For all MAb doses, a more rapid blood clearance of free compared with complexed BIWA 1 was observed. HPLC analysis did not reveal the presence of small radioactive metabolic products in any of the samples.

In addition to HPLC analyses, ELISA analyses were performed to assess concentrations of total as well as immunoreactive BIWA 1. Assessment of the total BIWA 1 levels with ELISA revealed, as expected, similar values as derived from HPLC analyses (Fig. 3). Also concentrations of immunoreactive BIWA 1 appeared to be similar. This latter observation seems to be remarkable because massive complex formation had been observed, especially in the 2-mg dose group.

To obtain further insight in the phenomenon of BIWA 1 complexation, the serum levels of soluble CD44v6 were assessed by ELISA. Prior to administration of BIWA 1, concentrations ranged from 111 to 365 ng/ml. In sera taken 1 week after injection, sCD44v6 levels seemed consistently reduced (range, 43–170 ng/ml). This reduction appeared to be related to the BIWA 1 dose administered: 39.0 ± 23.6%,43.8 ± 27.6%, and 58.4 ± 10.1% in the low-,intermediate-, and high-dose group, respectively. Six weeks after injection, sCD44v6 levels had approximately returned to their preinjection values (range, 120–226 ng/ml).

No HAMAs were detected prior to injection of BIWA 1. Eleven patients developed elevated HAMA titers (up to 2567), irrespective of the BIWA 1 dose administered. Mean titers 6 weeks after injection measured 1159 ± 943, 1005 ± 398, and 912 ± 834 for the low-,intermediate-, and high-dose group, respectively. One patient developed her highest titer 1 week after injection and demonstrated a slight decrease at 6 weeks after injection, whereas in 10 patients, HAMAs were detected at 6 weeks but not at 1 week after injection. In two patients,additional samples were obtained at 12 weeks after injection. In both,a slightly diminished HAMA level in comparison to 6 weeks after injection was found.

The present study demonstrates that BIWA 1, a high-affinity antibody directed to CD44v6, can be safely administered to HNSCC patients. No drug-related adverse events were observed, nor were any significant alterations in laboratory parameters noted. Human antimouse IgG antibodies were detected in 11 of 12 treated patients. Chimeric as well as humanized versions of BIWA 1 have meanwhile been developed that should minimize the problem of immunogenicity.

Preferential BIWA 1 uptake was observed in tumor tissue because RIS with radiolabeled BIWA 1 clearly depicted 11 of 12 primary tumors. Biopsies of the only primary lesion not depicted demonstrated an uptake level of 13.3% ID/kg, whereas 30% of the cells had been targeted by the MAb. This MAb uptake was similar to that of visualized tumors,whereas the size of this tumor was relatively large (26.4 cm³). An explanation for this diagnostic failure might be that in this particular patient, the MAb uptake in normal mucosa was relatively high, resulting in a low tumor:mucosa ratio of 1.2.

For the detection of lymph node metastases, RIS alone did not demonstrate any diagnostic gain in comparison with the conventional imaging techniques CT and MRI. Sensitivity of RIS, palpation, MRI, and CT per lymph node level appeared to be 50, 44, 61, and 67%,respectively. A similar sensitivity rate was established for RIS with U36 (50%; Ref. 7). However, a rather high rate of false-positive observations was found in this study for each diagnostic modality. Moreover, small distant metastases in one patient were not detected by SPECT.

Substantial improvements are needed to make RIS with BIWA 1 of clinical value. ^99mTc is not the ideal radionuclide to be used in combination with whole IgG because of the short half-life of ^99mTc (t_1/2, 6 h), which necessitates imaging to be performed within the first 24 h after administration, before optimal tumor:nontumor ratios are achieved. In a recently initiated RIT study with ¹⁸⁶Re-labeled MAb(t_1/2 ¹⁸⁶Re,91 h), images are acquired up to 14 days after injection. In time,background activity diminishes, resulting in improved tumor detection.³

Measurement of activity in biopsies from the surgical specimen showed higher uptake levels in primary tumor tissue than in any other tissue evaluated. Uptake in primary tumor tissue compared with normal mucosa proved to be significantly higher (P < 0.025). This is thought to be attributable to a better accessibility of antigenic sites in tumor tissue because the literature does not provide any evidence of higher CD44v6 expression in comparison with normal mucosa. Tumor-infiltrated lymph nodes contained only slightly more activity than tumor-negative lymph nodes (5.3 versus 3.5%ID/kg; P > 0.1). Underestimation of uptake in pathological lymph nodes may have occurred because of contamination of samples with lymphoid tissue and necrosis. Immunohistochemical comparison of pathological lymph nodes and primary tumors revealed similar levels of BIWA 1 accumulation.

Radioactivity levels in blood appeared to be similar to the levels in bone marrow. Activity mainly resided in plasma, as became apparent upon centrifugation. This is of paramount importance because bone marrow toxicity is expected to become dose-limiting in RIT with BIWA 1,especially because of the long half-life time of the MAb in blood. In other normal tissues, low radioactivity levels were found (Table 3).

Measurements of radioactivity in biopsies as well as within ROIs did not reveal selective MAb uptake in tissues other than normal oral mucosa. Tumor:normal mucosa ratios at 40 h after injection were∼2 for the 2- and 12-mg BIWA 1 dose groups. Moreover,immunohistochemical analyses revealed that BIWA 1 binding was restricted to the basal cell layers (Fig. 2). Despite this high uptake in the basal cell layers, it has to be seen whether mucosal toxicity will occur when using BIWA 1 for clinical RIT. Maraveyas et al. (29) reconstructed a theoretical phantom of the larynx and derived local dosimetric data for the selection ofβ-emitting radionuclides. In their dosimetric calculations, the authors take into account the fact that some of the disintegration energy dissipates outside the distribution volume of the tissue. For ¹⁸⁶Re, the absorbed fraction in tumors was calculated to be ∼1.6 times larger than in the normal mucosa, which leads to a greater tumor:mucosa dose advantage.

Although the number of patients in this study is small, tumor uptake levels were not significantly influenced by tumor volume nor by levels of circulating CD44v6. Data may suggest a tendency toward dose dependency with lower tumor uptake levels and tumor:normal mucosa ratios at 52 mg (Tables 3 and 4). Most likely, observed differences can be attributed to statistical variation in view of the small number of patients/group.

Immunohistochemical analyses showed heterogeneous distribution of the injected MAb throughout the tumor, irrespective of the dose administered. Although >95% of the tumor cells expressed the antigen,maximally 40% of the cells had been targeted by BIWA 1. No obvious improvement was observed upon dose increase from 12 to 52 mg. It has been suggested for high-affinity MAbs that antibody-antigen interaction at vascular entry sites of tumors may impose a binding site barrier that retards MAb percolation, thereby restricting uniform distribution (30, 31, 32, 33).

It might seem remarkable that the tumor uptake levels in the 2-and 12-mg group were similar, because a much larger proportion of MAb became complexed in the 2-mg group than in the 12-mg group. Apparently,complexed IgG is still capable of binding to antigen present in the tumor. Evidence for this possibility was also found in the ELISA assays performed. With one assay, the total BIWA 1 concentration in the serum was measured, whereas with the other assay, only immunoreactive BIWA 1 was measured. Remarkably, even serum samples containing >50%complexed IgG yielded similar results in both assays, indicating that complexed BIWA 1 in the serum is still able to bind to immobilized antigen. Possibly, one of the antigen binding sites of the bivalent IgG is still available for tumor binding after complexation. Moreover, BIWA 1 may have a higher avidity and affinity for membrane-bound,v6-containing CD44 variants than for soluble CD44v6 variants. The high density of CD44v6 on tumor cell membranes may allow bivalent binding,whereas bivalent binding to soluble CD44v6 variants may be restricted by the low concentration of these molecules in the circulation. Furthermore, v6-containing CD44 variants present on tumor cell membranes may have better interaction with BIWA 1 than the variants present in the circulation because of conformational differences. Reduced affinity for binding to circulating antigen in comparison to immobilized antigen has also been described for other antibodies,including MAbs directed against carcinoembryonic antigen (34, 35).

Levels of sCD44v6 apparently fell during the first week after BIWA 1 administration. This does not mean that sCD44v6 is removed from the circulation. It can also be attributable to complex formation between antigen and antibody. BIWA 1 bound to sCD44v6 may compete with the horseradish-peroxidase conjugated BIWA 1 used in the ELISA test. By such competition, sCD44v6 levels will become underestimated.

Because complex formation (a) does not result in an obviously diminished uptake of BIWA 1 in tumor tissue (Table 3) and(b) does not result in an increased uptake in specific normal tissues, e.g., the liver (Table 5) and (c)can be kept relatively low by using high MAb concentrations, the presence of soluble antigen in the circulation is not a limiting factor for the therapeutic applicability of BIWA 1.

As outlined in the “Introduction,” similar studies have been performed with another CD44v6-specific MAb, called U36. Although U36 resembles BIWA 1, it binds to a different epitope, with a 35-fold lower affinity. The tumor uptake of U36 was not significantly influenced by the MAb dose and certainly not lower than the tumor uptake of BIWA 1. Immune complexes were barely formed, whereas a homogeneous MAb distribution throughout the tumor was achieved when increasing the MAb dose to 52 mg. This indicates that high-affinity BIWA 1 does not result in improved tumor targeting when compared with U36.

In conclusion, CD44v6-specific BIWA 1 is a promising MAb because it can be safely administered to HNSCC patients, and it shows high and selective tumor uptake. RIS with ^99mTc-labeled BIWA 1 is as reliable as other imaging techniques, although it was performed under suboptimal conditions (21 h after injection). However,BIWA 1 is immunogenic, forms complexes, and shows a heterogeneous accumulation throughout the tumor, suggesting that BIWA 1 is not able to penetrate deeper cell layers. Because in RIT targeting of the whole tumor cell population is to be preferred and possibly repeated antibody infusions are required, several chimeric and humanized MAb BIWA versions with different affinities have been constructed for further clinical evaluation. Preclinical in vitro and in vivo studies revealed less complex formation and more efficient tumor targeting when the affinity of CD44v6-specific MAbs was diminished.⁴Chimeric or humanized BIWA with an optimal affinity will be labeled to an escalating dose of ¹⁸⁶Re in a subsequent Phase I RIT trial with HNSCC patients.

We thank Remco de Bree for clinical support, Willem den Hollander for biopsy measurements and high-performance liquid chromatography analyses, Gerard J. van Kamp for HAMA assaying, and Miranda Siegmund for labeling support.