In Vivo Targeting of Malignant Melanoma by ¹²⁵Iodine- and ^99mTechnetium-labeled Single-chain Fv Fragments against High Molecular Weight Melanoma-associated Antigen¹

MM³is a skin cancer that affects ∼40,000 new patients each year in the United States and an estimated 100,000 worldwide (1, 2). Every year, the incidence of MM increases (3); and for Caucasians, the lifetime risk of developing MM may rise to 1 in 75 by the year 2000 (4). Melanoma is an important cause of cancer among young patients (ages, 30–50 years) increasing the economic importance of the disease.

Survival of melanoma patients is strongly determined by stage at presentation because the risk of developing metastases increases with increasing thickness of the initial tumor (5). Once melanoma has metastasized, the most important determinants of survival are the site of metastasis (i.e., local, regional, distant),the number of sites involved, and whether a visceral site is affected (6). These data underline the need for accurate methods to detect metastases both to provide prognostic information and to select appropriate treatment(s). Unfortunately, existing treatments can improve long-term survival for only a minority of patients with metastatic melanoma (7). Experimental forms of treatment such as immunotherapy, gene therapy, or combination biochemotherapy have often shown promise in early clinical trials (8, 9, 10, 11)but often fail to produce consistent increases in survival on further testing. Many of these studies have been conducted on patients with late-stage melanoma, and it is possible that metastatic MM may be more amenable to treatment at an earlier stage, which adds emphasis to the need to develop methods to detect metastases before they become widespread.

A number of standard diagnostic methods are used to detect metastatic MM including plain chest X-rays (CXR), ultrasound scans (principally of abdominal organs), computerized tomography (CT) and magnetic resonance imaging (MRI). Nuclear imaging techniques such as positron emission tomography (PET) and ISG are also used but are only available in a few centers in the United Kingdom. A definitive clinical study comparing the diagnostic value of these techniques for metastatic MM has not been conducted, but a recent review (12) concluded that nuclear imaging techniques are superior to standard diagnostic methods in terms of sensitivity, specificity, and financial cost. Furthermore, whereas positron emission tomography relies on the nonspecific accumulation of ¹⁸fluoro-2-deoxy-d-glucose (FDG) in metabolically active MM lesions, the detection of metastatic MM by ISG is specific. The MAbs that are used for ISG may also have therapeutic potential (13).

ISG using antimelanoma antibodies as radioimmunopharmaceuticals for the detection of metastatic melanoma has been extensively investigated. A recent literature survey found 58 patient trials (excluding case studies) involving a total of 3638 patients (12). The majority (>80%) of these studies used MAbs against HMW-MAA proteoglycan. HMW-MAA is a well-characterized melanoma surface antigen with very limited expression by normal tissues (14, 15) and low heterogeneity of expression from patient to patient (16). It has also been used as a target for the treatment of MM (17, 18). The sensitivity of ISG using anti-HMW-MAA antibody preparations for the detection of clinically known metastatic MM deposits varies but is generally around 75%, which compares favorably with standard diagnostic methods (12). ISG is also able to survey the entire body for metastases in a single step and can detect a substantial number of otherwise clinically occult lesions,which makes it more useful compared with standard imaging in the diagnosis of metastatic MM (19).

The main limitation to the routine clinical use of ISG for metastatic MM is nonspecific accumulation of IgG in normal organs, which obscures the presence of metastases located in those tissues (20). In many tissues, this is attributable to the binding of IgG to Fc receptors on cells of the mononuclear phagocytic system(e.g., Kupffer cells in the liver, mesangial cells in the kidney, alveolar macrophages in the lung) as well as uptake by other cells involved in antibody metabolism and clearance (21, 22, 23). The relatively large size of MAbs also leads to slower clearance from the tissues (24), which further reduces the contrast between tumor deposits and the surrounding normal tissues. To avoid these problems, ISG has been performed with F(ab′)₂ and F(ab′) antibody fragment preparations(which lack the Fc part) resulting in lower levels of background accumulation and improvements in the ability of ISG to detect metastases (25, 26).

More recently, further reductions in levels of background accumulation and improved tumor targeting have been achieved in mice and patients with the use of scFv antibody fragments as radioimmunopharmaceuticals directed against nonmelanoma tumors (27, 28, 29). scFvs(M_r 27,000) consist of antibody variable domains connected by a synthetic linker (30). We and others have reported the construction of anti-HMW-MAA scFvs (31, 32, 33, 34, 35, 36) that may produce similar improvements in the quality of ISG images for metastatic MM compared with existing F(ab′)₂ or F(ab′) preparations.

For clinical studies, factors such as high yield, stability,solubility, and affinity are important. However, scFvs derived from MAbs or non-immune phage display libraries often have low affinities or yields. Furthermore, scFvs are relatively small molecules and have a limited number of amino acids available for chemical modification,thereby reducing the efficiency of radiolabeling (37). We have previously described an anti-HMW-MAA scFv (RAFT2; Ref. 31) derived from the mouse MAb LHM2. RAFT2 binds human melanoma in vitro with improved specificity when compared with the parental MAb (15, 31). However, melanoma binding is weak and RAFT2 is expressed at levels too low for clinical studies. We, therefore, used chain-shuffling and selection on melanoma cells to generate an anti-HMW-MAA scFv (RAFT3) with the same epitope specificity as RAFT2 but with improved melanoma binding and expression.⁴

The purpose of this study was to characterize RAFT3 scFv in an animal model and determine its suitability for use in a Phase I clinical trial of ISG in patients with metastatic melanoma. We demonstrate specific targeting of human melanoma xenografts by RAFT3 scFv in vivo. Moreover, RAFT3 scFv is easy to label with ^99mTc for ISG. The implications for diagnosis and therapy of metastatic melanoma will be discussed.

The human metastatic melanoma cell line A375M (38) was used for all experiments. The human scFv directed against human CD18 has been described previously (31). RAFT3 scFv was obtained by replacing the mouse V_κ of anti-HMW-MAA scFv RAFT2 (31) with a human V_κ.⁴Both scFvs were expressed in vector pUC119 His₆Xba c-myc (kind gift of G. Winter, Medical Research Council Centre for Protein Engineering, Cambridge, United Kingdom). The vector encodes an epitope bound by anti-c-myc MAb 9E10(American Type Culture Collection) and a COOH-terminal His₆ sequence for IMAC purification as described previously (37, 39). ScFvs were expressed in Escherichia coli strain TG1 (40).

Large scale inductions (3 liters) in Luria-Bertani medium (Life Technologies, Inc.) with ampicillin (50 μg/ml) and 100μ m IPTG(isopropyl-β-d-thiogalactopyranoside) were rotated at 200 rpm and 30°C for 17 h. Sodium azide (0.02%) was added to the culture supernatant prior to centrifugation at 12,000 × g for 2 h. The bacterial supernatant was filtered (0.45μm), concentrated to ∼100–200 ml (Flowgen Mini-Ultrasette; M_r 10,000 cutoff), and dialyzed against PBS overnight at 4°C.

A 50-ml column packed with 40 ml of Chelating Sepharose (Pharmacia) was loaded with 0.1 m CuSO₄ in distilled H₂O and pre-equilibrated with 2-column volumes of PBS. Prior to loading onto the IMAC column, sodium chloride was added to the concentrated supernatant at a final concentration of 1 m. The IMAC column was washed with PBS + 1 mNaCl. Elution was carried out using a continuous gradient of imidazole from 20 mm to 200 mm in PBS + 1 mNaCl. Fractions containing scFv (eluted between 80 and 87 mm imidazole) were concentrated to a final volume of 0.5–1.0 ml by dialysis against PEG-6000 followed by dialysis against PBS overnight at 4°C.

The RAFT3 scFv and anti-CD18 scFv required further purification by ion-exchange chromatography to remove a M_r 25,000 impurity recovered with the scFvs after IMAC purification. A 5-ml column was packed with 4 ml of Q-Sepharose Fast-flow gel (Pharmacia Biotech) according to the manufacturer’s instructions. The column was equilibrated with 5-column volumes of PBS. scFv in PBS was loaded onto the column. Elution was carried out with a rising gradient of NaCl (8.0 mm-1.0 m) in PBS. Pure scFv was recovered in fractions between 8 and 420 mmNaCl and was concentrated to 0.5–1.0 ml by dialysis against PEG-6000 before dialysis against PBS overnight at 4°C.

Expression of scFvs in the raw bacterial supernatant was estimated by Western blot (41). Yields of scFv after IMAC and ion-exchange purification were estimated by SDS-PAGE analysis using 1-,3-, and 5-μg ovalbumin standards (42).

LHM2 F(ab′)₂ fragments were produced by pepsin digestion of LHM2 IgG1 mouse MAb (gift of I. Leigh, Imperial Cancer Research Fund Centre for Cutaneous Research, London, United Kingdom) using standard techniques (43). Pure LHM2 IgG1(0.5 mg/ml) in PBS was dialyzed against sodium citrate (pH 3.4)overnight at 4°C. Pepsin (1 mg/ml) was prepared in the same buffer. Digestion was carried out at 37°C for 6 h. The reaction was terminated by the addition of one-tenth volume of 3 mTris-HCl (pH 8.8).

Gel filtration and Protein A immunoaffinity chromatography were used to remove undigested LHM2 IgG1, Fc fragments, and pepsin from the digestion mixture. Undigested IgG1 and Fc fragments were removed by passing the digestion mixture over a Protein A column allowing the LHM2 F(ab′)₂ to pass through. Microaggregates and pepsin were not removed by Protein A and were separated from the LHM2 F(ab′)₂ by gel filtration on a HiPrep S-300 prepacked gel filtration column (Pharmacia Biotech). All of the digestion mixtures were dialyzed against PBS overnight at 4°C before each purification step.

Yields of LHM2 F(ab′)₂ after Protein A and gel filtration purification were estimated by SDS-PAGE analysis using 1-,3- and 5-μg ovalbumin standards (42).

LHM2 IgG1, polyclonal mouse IgG1, LHM2 F(ab′)₂,and RAFT3/anti-CD18 scFvs used for biodistribution and pharmacokinetic studies were directly iodinated with ¹²⁵I(Amersham) using 1 μg of chloramine-T per μg of protein. Between 50 and 100 μg of protein (1 mg/ml in PBS) were labeled with 9.25–29.6 MBq of ¹²⁵I in each reaction. A reaction time of 1–2 min at room temperature was used for all proteins. Reactions were terminated by the addition of an excess of d-tyrosine (1μg per μg of chloramine T) in 50 mm phosphate buffer(pH 7.5) containing 2.4 mg/ml sodium metabisulfite. Unincorporated ¹²⁵I was separated from the protein on a prepacked PD10 column (Pharmacia).

RAFT3 scFv was labeled with ^99mTc as described previously (25). All of the reactions were carried out at room temperature. A vial of MDP (Amerscan Medronate II Technetium Agent, Amersham International) was reconstituted with 5 ml of normal saline. An aliquot of protein (50 μg) in 50–100 μl of PBS was placed in a 2-ml microcentrifuge tube. The MDP agent (40μl) was added to the scFv immediately before adding ^99mTc (maximum, 100 μl). The reaction was allowed to proceed for 10 min and was terminated by separating the labeled protein from the free ^99mTc on a PD10 column. The reaction mixture was washed through the column with 300-μl aliquots of normal saline. Fractions containing the highest activity (fractions 10–16) contained the radiolabeled protein and were pooled. Counts in these fractions were expressed as a percentage of the total counts in all of the fractions + counts in the PD10 column,yielding a value for percentage incorporation.

The immunoreactivity of the ¹²⁵I- and ^99mTc-labeled proteins was tested by ELISA using melanoma cells as described previously (31). Briefly,A375M cells were grown to confluency in 96-flat-well microtiter plates,the media were discarded, and plates were air-dried. Serial dilutions of samples (e.g., LHM2 IgG1, LHM2 F(ab′)₂, or scFv) were made in RPMI + 10% FCS +azide and incubated at room temperature for 1 h. The plates were washed 3 times with PBS. The LHM2 IgG1, LHM2 F(ab′)₂, and scFv were detected using an antimouse IgG horseradish peroxidase (DAKO Ltd, Cambridgeshire,United Kingdom). In the case of scFv, the supernatant from 9E10 hybridoma cells was added for 1 h prior to the peroxidase treatment. Signal intensity was measured using a Microplate Reader(Bio-Rad) coupled to a personal computer. All of the plates were read at 490 nm.

The radiochemical purity of ^99mTc-labeled RAFT3 scFv was determined using ITLC on ITLC-SG paper (ITLC-silica gel;Gelman) with distilled H₂O as the solvent. For ITLC, the chromatography strip was placed on a gamma camera after 10 min in the solvent, and the relative amounts of activity at the origin(^99mTc bound to protein) were compared with the radioactivity at the solvent front (free ^99mTc or ^99mTc bound to the MDP agent). The activity at the origin was expressed as a percentage of the total radioactivity in the ITLC strip, yielding a value for radiochemical purity. The formation of radiocolloid was <3% under the conditions used for labeling (44). The stability of the radiolabeled scFv was tested by trichloric acid precipitation after incubation in fresh human serum for 24 h at 37°C and after incubation in PBS for 24 h at 4°C.

The human melanoma cell line A375M was used for all of the animal experiments. A 100-μl volume of cell suspension in normal saline containing 1 × 10⁶ cells was injected s.c. into the right flank of BALB/c nu mice under inhalational anesthesia. Tumor xenografts reached a usable size (7 mm ± 1 mm, geometric mean diameter; average weight, 133 mg) after 4–6 weeks. Samples of the xenografts were then taken and subjected to immunohistochemical analysis to ascertain that the epitopes for the immunopharmaceuticals were expressed in vivo (data not shown). Groups of 3–5 mice were used for pharmacokinetic and biodistribution experiments.

Each mouse received 0.3 μg (0.037 MBq/μg) of ¹²⁵I-labeled LHM2 IgG1, polyclonal mouse IgG1,LHM2 F(ab′)₂, RAFT3 scFv, or anti-CD18 scFv, made up to 100 μl in normal saline. ^99mTc-labeled RAFT3 scFv was used as above at a specific activity of 4 MBq/μg. The preparation was filter sterilized and administered by tail vein injection. For biodistribution studies using IgG1, and F(ab′)₂, mice were killed at 6, 18, 24, and 48 h after injection. For biodistribution studies using scFvs,mice were killed at 1, 3, 6, and 18 h after injection. For blood clearance studies of IgG1 and F(ab′)₂, samples were obtained at 1, 15, 30, and 80 min and 3, 6, 18, 24, and 48 h after injection. For studies using scFvs, samples were obtained at 1,15, 30, and 80 min and 3, 6, and 18 h after injection.

Mice were killed by cervical dislocation, and blood samples were obtained by cardiac puncture. Six tissues were sampled, including whole tumor xenografts, lungs, kidneys, spleen, liver, and blood. Tissue samples were blotted dry on tissue paper, and any adherent blood clots were removed. The wet weight of all samples was recorded, and the radioactivity in each sample measured with a CompuGamma CS gamma counter (LKB Wallac) as cpm. Errors were calculated as SEs of the measurements obtained from groups of 3–5 mice, using the computer program JMP (version 3.1.5, SAS Institute).

For blood clearance studies, results were expressed as a percentage of the injected dose retained in the blood. The 100% value was taken as the level of activity found at 1 min after injection of the radioimmunopharmaceutical. All of the values were then expressed as a percentage of this level of activity obtained from groups of 5 mice. Curve-fitting was carried out using Origin software (version 4.0,Microcal) and errors were calculated as SEs of the measurements obtained from groups of 5 mice.

Results of biodistribution studies were expressed as %ID/g, as a RI or as T:N tissue ratios. The %ID/g was calculated as cpm in the tissue sample divided by total cpm administered to each mouse and divided by the wet weight of the sample in grams. The RI was calculated from the%ID/g of tumor-specific radioimmunopharmaceutical (i.e.,LHM2 IgG1 and RAFT3 scFv) in a tissue at a given time point divided by the %ID/g of a non-tumor-specific radioimmunopharmaceutical(polyclonal mouse IgG1 and anti-CD18 scFv) in the same tissue. The T:N ratio was calculated from the %ID/g of tumor-specific radioimmunopharmaceutical in the tumor at a given time point divided by the %ID/g of radioimmunopharmaceutical in the normal tissues (liver,lung, kidney, or spleen) at the same time point.

Incorporation of ¹²⁵I into LHM2 IgG1 after direct iodination with chloramine-T was 30%, yielding specific activities of 0.3 MBq/μg. For LHM2 F(ab′)₂ incorporation was 20%, yielding specific activities of 0.2 MBq/μg. Incorporation of ¹²⁵I into polyclonal mouse IgG1 was higher at 60%, yielding specific activities of 0.4 MBq/μg. The higher incorporation probably reflects the heterogeneity of different IgGs in the polyclonal mix. Incorporation of ¹²⁵I into RAFT3 scFv was only 6.9%, yielding specific activities of 0.037 MBq/μg. The low level of incorporation of ¹²⁵I into RAFT3 scFv appeared to be a particular property of this scFv. The immunoreactivity of ¹²⁵I-labeled RAFT3 scFv was also low at 28.6% (Table 1).

In preliminary experiments, we tested the radiolabeling properties of a number of high-affinity antimelanoma scFvs from our panel using the direct labeling technique described by Siccardi et al.(25). The direct labeling technique described in our paper is well known and has been extensively used in previous studies to label other immunopharmaceuticals. The particular technique that we used is described in detail by Siccardi et al.(25) and was used in the preparation of ^99mTc-labeled F(ab′)₂ for use in a clinical trial of ISG for metastatic melanoma in 493 patients. We did not use glucoheptonate or gluconate as transfer agents but carried out preliminary experiments using mercaptoethanol and DTT as reducing agents to improve labeling efficiency. Both of the reducing agents severely depleted the immunoreactivity of all of the scFvs in our series (data not shown), and, therefore, we did not use this technique again.

Two of the scFvs in our series (B3 and B4) have been described previously (15, 37). Incorporation rates for B3 and B4 were 55 and 57%, respectively, and radiochemical purity was estimated by ITLC as 80 and 88%, respectively, 45 min after labeling. The immunoreactivity of B3 and B4 was unaffected by direct labeling with ^99mTc. However, direct labeling of B3 and B4 was unstable in vitro when tested by serial ITLC. For example,only 34% of the ^99mTc remained bound to the B3 scFv 120 min after labeling. Similar results were obtained with three other scFvs that were tested (data not shown).

In contrast, incorporation of ^99mTc into RAFT3 scFv after direct labeling was 68%, yielding a specific activity of 4.0 MBq/μg (Table 1). Moreover, radiochemical purity of RAFT3 scFv after direct labeling with ^99mTc was estimated as 92% by ITLC, 45 min after labeling. ^99mTc-labeling did not affect the immunoreactivity of RAFT3 scFv (Table 1). Radiochemical stability was estimated in vitro by TCA precipitation. RAFT3 scFv retained 92% of the label after incubation in PBS for 24 h at 4°C. In the presence of fresh human serum, RAFT3 scFv retained 84.5% of the label after 24 h at 37°C.

The in vivo blood clearance of ¹²⁵I-labeled LHM2 IgG1, LHM2 F(ab′)₂, and RAFT3 scFv was determined in the same nude mouse model used for biodistribution studies. The data (Table 2; Fig. 1) show that blood clearance followed a biphasic pattern for IgG1,F(ab′)₂, and scFv with t_1/2α representing clearance from the circulation and t_1/2β representing slower clearance from the tissues. RAFT3 scFv cleared from the circulation more quickly compared with LHM2 F(ab′)₂, which in turn, cleared more quickly than LHM2 IgG1. Values for t_1/2α and t_1/2β were similar to those for scFvs (27, 45) in previous studies, which confirmed the rapid rate of blood clearance for the smaller fragment.

When analyzing the biodistribution data for LHM2 IgG1, LHM2 F(ab′)₂ and scFvs, it was essential to compare equivalent time points after injection. To clarify this in the data(Table 3), we have highlighted the equivalent time points in italics.

The highest overall %ID/g observed in the tumor xenografts was achieved with LHM2 IgG1 (Table 3A), but this was not reached until 48 h after injection. In contrast, the %ID/g of LHM2 F(ab′)₂ and RAFT3 scFv achieved in tumor xenografts was much lower, although peak levels were reached more rapidly compared with LHM2 IgG1 (Table 3). The lower %ID/g has been attributed to the more rapid pharmacokinetics of antibody fragments that gives them less time to bind to the tumor xenografts (46).

Initially, the %ID/g of RAFT3 scFv in the kidney was extremely high compared with that of LHM2 IgG1 and LHM2 F(ab′)₂(Table 3). For example, the %ID/g in the kidney for RAFT3 scFv was 3.9-fold higher compared with LHM2 IgG1 and 3.2-fold higher compared with LHM2 F(ab′)₂ at 6 h after injection. However, as with the other normal tissues, RAFT3 scFv cleared rapidly from the kidney and by 18 h after injection the %ID/g for RAFT3 scFv in the kidney was comparable with LHM2 IgG1 (Tables 3A and 3C).

RI for LHM2 IgG1 showed a large rise in tumor xenografts with values 7.1-fold higher at 24 h and 23.6-fold higher at 48 h compared with those of polyclonal mouse IgG1 (Tables 3A and 3E Table 3; Table 4A). This suggests that accumulation of LHM2 IgG1 was attributable to antigen-specific mechanisms. However, values for RI also showed a gradual increase with time in liver, lung, kidney, spleen and blood (Table 4A). The increase in RI in normal tissues was probably attributable to nonspecific mechanisms, which accounts for the smaller rise in RI compared with that in tumor. However, values of RI for blood were nearly as high as for tumor. This may have been caused by the presence of circulating tumor antigen forming large complexes with the LHM2 IgG1 that did not clear quickly from the circulation. This hypothesis was not formally tested, but there are several possible reasons why a similar effect was not seen with RAFT3 scFv, which recognizes the same antigen. These reasons will be discussed in greater detail.

RI for RAFT3 scFv in the tumor xenografts also increased but more rapidly, reaching 12.2 at 18 h compared with 5.1 for LHM2 IgG1 at the equivalent time point, which emphasized the more rapid penetration of the tumor grafts achieved by scFvs (Table 4). In contrast to LHM2 IgG1, values of RI for RAFT3 scFv in all other tissues (except kidney)fell rapidly with time. These data suggest that accumulation of RAFT3 scFv in the tumor xenografts was attributable to antigen-specific mechanisms and emphasize the specificity of the localization compared with LHM2 IgG1, for which RI values showed an increase in all normal tissues, especially the blood.

At equivalent time points after injection, T:N ratios for RAFT3 scFv were higher compared with those for LHM2 IgG1 and LHM2 F(ab′)₂ in all organs except kidney which indicated much better tumor specificity (Tables 5A-D). In particular, the T:N ratios for LHM2 IgG1 in blood were very poor, which suggested a potential problem with the use of LHM2 IgG1 for ISG in terms of the blood pool. In contrast to previously published data (25), the poorest T:N ratios were observed with LHM2 F(ab′)₂. These data suggest that for imaging, the best contrast between the tumor xenografts and normal organs (except kidney) would be obtained using RAFT3 scFv. However, T:N ratios for RAFT3 scFv also suggested that tumor deposits located near the kidney might remain obscured because of the levels of nonspecific accumulation in this organ.

Comparisons with ^99mTc-labeled LHM2 IgG1 and LHM2 F(ab′)₂ were not carried out because of differences in blood clearance when compared with scFv. However, a comparison of T:N ratios for ¹²⁵ I- and ^99mTc-labeled RAFT3 scFv (Tables 5C and 5D)suggests that labeling with technetium had no significant effect on in vivo targeting. Moreover, T:N ratios for ^99mTc-labeled RAFT3 scFv appeared to be superior overall compared with those for ¹²⁵I-labeled RAFT3 scFv, although this may have been attributable to differences in immunoreactivity of the ¹²⁵I- and ^99mTc-labeled preparations of RAFT3 scFv.

The main purpose of this study was to characterize the in vivo targeting properties of an antimelanoma scFv as a preliminary to a Phase I clinical trial of ISG in melanoma patients. ^99mTc is the ideal radioisotope for ISG because of the optimal energy of the γ radiation emitted (140 keV), which makes it easier to detect. It is also inexpensive (the nominal cost of ^99mTc used for each experiment is£1.00 United Kingdom sterling or US $1.45) and is easy to dispose of because of its short half-life(t_1/2, 6 h), which matches the serum t_1/2 of scFv more closely than that of whole antibodies. However, direct labeling of scFvs with ^99mTc can be difficult because either incorporation is low and/or the ^99mTc-scFv complex is unstable in vivo (47). As a result, ^99mTc-labeled scFvs are in great demand for imaging studies.

We used a technique described by Siccardi et al.(25) and a commercial kit for direct labeling of RAFT3 scFv with ^99mTc. The technique was simple and had no discernible effect on the immunoreactivity of the scFv. Our rate of incorporation (Table 1) compares favorably with previous studies (47, 48). However, in comparison with these studies, ^99mTc-labeled RAFT3 scFv remained remarkably stable in vitro for 24 h in the presence of human serum and PBS, whereas the biodistribution data (Tables 3D and 5D) Table 5 suggested that it remained stable in vivo for at least 18 h. In contrast, direct labeling of B3, B4, and three other scFvs by the same technique proved unstable in vitro. More recently, ^99mTc-labeling of the His₆tail of a scFv has also been reported (49). Although this method should be applicable to most scFvs, kidney accumulation of the scFv reported in that study was much higher compared with that of RAFT3 scFv (T:N ratio of 0.013 at 24 h compared with 1.1 for RAFT3 scFv at 18 h; see Table 5). Although it was not the principal aim of this study, it would be of interest to characterize the mechanism of ^99mTc labeling of RAFT3 scFv in the future because amino acid and structural comparisons did not reveal any obvious differences between RAFT3 scFv and the other scFvs tested (data not shown).

Our biodistribution data show that RAFT3 scFv can be used to target human melanoma in vivo in a nude mouse xenograft model. RAFT3 scFv exhibited greater specificity for tumor xenografts compared with LHM2 IgG1 in terms of RI values and higher T:N ratios at equivalent time points after injection. RAFT3 scFv also exhibited faster localization (earlier peak %ID/g and more rapid rise in RI) to the tumor xenografts compared with LHM2 IgG1 at equivalent time points after injection. The slower rate of accumulation of LHM2 IgG1 in the tumor xenografts is probably attributable to slower penetration of the tissues by IgG compared with scFv, which has been observed in previous studies (24, 37).

The principle determinants of the speed with which radiopharmaceuticals are cleared from the circulation are their size and the presence of an Fc portion because both factors determine whether they will be preferentially removed by the mononuclear phagocytic system or by filtration in the kidney (23, 50, 51, 52, 53, 54). In general, radiopharmaceuticals that are large(M_r >60,000) or that possess an Fc portion are cleared by phagocytosis. scFvs are small and lack the Fc portion and are, therefore, removed from the circulation by rapid filtration in the kidney (55, 56). Removal from the circulation by phagocytosis is a slower, energy-dependent process compared with filtration, which is a passive process dependent on high blood flow in the kidney. The enormous importance of the kidney in the excretion of scFvs has been demonstrated by Laroche et al.(57), who showed that the t_1/2α of K₁₂G₀ scFv was increased from 10 min to 110 min after nephrectomy. We did not specifically look for the presence of antigen-antibody complexes because this was not the main purpose of this investigation. Nevertheless, any antigen-antibody complexes would still be cleared by the same mechanisms. If present,LHM2 IgG1-antigen complexes would be expected to form an equilibrium between bound and unbound antibody in the circulation, with both being cleared relatively slowly by phagocytosis. This might account for the higher values of RI in blood observed with LHM2 IgG1 (Table 4). Similarly, scFv-antigen complexes would form an equilibrium with the unbound scFv, which would be cleared rapidly from the circulation by the kidney while the large scFv-antigen complexes(M_r ≈175,000) would be removed by phagocytosis. In this scenario, clearance of the scFv would be delayed, compared with a situation in which scFv-antigen complexes were absent. Nevertheless, clearance of the scFv by a combination of routes would still be faster than clearance by phagocytosis alone and might account for the lower values of RI in the blood with RAFT3 scFv (Table 4) compared with those for LHM2 IgG1.

To some extent, speculating about the possible effects of antibody-antigen complexes on the clearance of RAFT3 scFv might be largely academic. This has been demonstrated by the effectiveness of scFv (58, 59) and IgG (60) in the targeting of CEA-secreting tumor xenografts in animals and the metastases of CEA-secreting colonic adenocarcinoma in humans (29). In both animals and humans, CEA is known to be shed into the circulation at high levels. Indeed, CEA levels are used in patients to monitor tumor load during therapy and to detect tumor recurrence during routine follow-up. Thus, theoretically, the absence of antigen-antibody complexes might be beneficial in terms of improved clearance, but in practice, their presence does not significantly reduce the effectiveness of tumor targeting, even in situations where immune complex formation is high. Moreover, in patients, HMW-MAA is known to be shed into the circulation at much lower levels (61) compared with CEA, and, therefore,complex formation is unlikely to be a significant problem when RAFT3 scFv is used in humans.

Differences in the mechanisms for clearance of IgG and scFv from the circulation also accounted for high levels of nonspecific accumulation of ¹²⁵I- and ^99mTc-labeled RAFT3 scFv in the kidney. This is a common observation in mouse studies of scFvs, although levels of RAFT3 scFv appeared to be relatively high compared with those in previous studies. scFvs are filtered by the renal glomerulus because it is permeable to small proteins(M_r <60,000) with the appropriate charge characteristics (55, 62). The charge selectivity of the glomerular basement membrane may explain the considerable heterogeneity in the level of renal accumulation observed among different scFvs and other small antibody fragments. For example, levels of renal accumulation reported by Milenic et al.(45) were 0.2%ID/g for ¹³¹I-labeled CC49 scFv 6 h after injection compared with 7.8%ID/g for RAFT3 scFv at the same time point. In contrast, King et al. (63) reported renal levels of 3.5%ID/g for scFv(M_r 27,000), 2.5%ID/g for dimers of scFv (M_r 54,000), and 3.7%ID/g for trimers of scFv (M_r 80,000) at 4 h after injection. Another reason for the heterogeneity of renal accumulation observed could be differences in the animal models used or differences in the techniques and radioisotopes used for scFv labeling. For example, T:N ratios for ^99mTc-labeled RAFT3 scFv in the kidney were 8.2 times higher at 6 h compared with ¹²⁵ I-labeled RAFT3 scFv using the same animal model (Tables 5C and 5D).

In practical terms, high levels of nonspecific accumulation of RAFT3 scFv in the kidney would limit the ability of ISG to detect metastatic melanoma deposits close to the kidney because these would be obscured. Moreover, it would limit the direct use of RAFT3 scFv labeled with therapeutic radioisotopes or drugs such as doxorubicin at therapeutic doses because of the risk of nephrotoxicity. However, it has been shown that dramatic reductions in renal accumulation of radioimmunopharmaceuticals can be achieved by i.v. administration of amino acids (64).

Although absolute levels of ¹²⁵I- and ^99mTc-labeled RAFT3 scFv in the tumor xenografts were lower compared with LHM2 IgG1, the lower levels of background accumulation in all organs (except kidney) suggest that contrast would be greatly improved if RAFT3 scFv were to be used for ISG. In particular, tumor:blood ratios for RAFT3 scFv were far better than those for LHM2 IgG1 at equivalent time points after injection. Indeed,LHM2 IgG1 has now been used for ISG of metastatic melanoma in humans in a small pilot study that showed very high levels of background accumulation in the blood pool and bone.⁵Thus, our biodistribution data appear to confirm one previous human study that suggests that imaging with scFv should produce less background accumulation in most normal tissues (29). Moreover, by 6–18 h after injection (Table 5), both ¹²⁵I- and ^99mTc-labeled RAFT3 scFv achieved T:N ratios in most tissues that were similar to or better than those previously reported at time points of 24–48 h using ^99mTc-labeled MFE-23 scFv (59) and ^99mTc-labeled 4D5 scFv (49). However, both ¹²⁵I- and ^99mTc-labeled RAFT3 scFv did not perform as well as ^99mTc-labeled MET scFv, which achieved T:N ratios at 24 h of ≈20:1 for blood and ≈17:1 for kidney (47).

The better performance of ^99mTc-labeled RAFT3 scFv compared with MFE-23 scFv or 4D5 scFv might represent a significant improvement in tumor targeting or might have been attributable to differences in the experimental model used in the other studies. This would be equally true when considering the better performance of MET scFv compared with RAFT3 scFv. For example, tumor uptake (and, therefore, T:N ratio) is known to be influenced by nonspecific factors such as: the radioisotope used (65);the size of the tumor xenografts (66, 67); the dose of radiopharmaceutical administered (68, 69); and the presence of circulating tumor antigens (60). The type of radioisotope used (^99mTc) was the same in all of the studies and was unlikely to have contributed to the difference in performance, although the labeling technique might have made a difference through effects on biological activity. Biological activity after ^99mTc-labeling was 87% for 4D5 scFv, 70%for MET scFv, 55% MFE-23 scFv, and 100% for RAFT3 scFv. These data suggest that biological activity was not responsible for the observed differences in performance of the different scFvs. The size of the tumor xenografts used in our study might have played a role because larger tumors accumulate more radiopharmaceutical, and this would tend to result in a higher T:N ratio. The xenografts used in our study averaged 133 mg per mouse (≈0.7 cm diameter) compared with 40 mg(0.4–0.5-cm diameter) for 4D5 scFv and for MET-scFv. The size of xenograft used was not reported for MFE-23 scFv. These data suggest that the xenografts in our study were larger than those used for 4D5 scFv and MET-scFv, which might explain the better T:N ratios compared with 4D5 scFv but would not explain the superior performance of MET-scFv compared with RAFT3 scFv. Tumor uptake is also increased as the dose (quantity of protein and specific activity) administered is increased, although this is usually at the expense of higher background accumulation. Each mouse in our biodistribution study received 1.2 MBq of RAFT3 scFv (0.3 μg at 4 MBq/μg) compared with 1.3 MBq of MFE-23 scFv (11 μg at 0.12 MBq/μg), 0.21 MBq of 4D5 scFv (3 μg at 0.07 MBq/μg), and 0.02 MBq of MET scFv. The relatively high dose of RAFT3 scFv might be expected to result in a higher %ID/g in the tumor at similar time points after injection. In fact, the %ID/g of RAFT3 scFv was 0.49 at 18 h (Table 3D) compared with 1.4 at 24 h for 4D5 scFv, ≈4.0 at 24 h for MFE-23 scFv, and ≈0.5 for MET scFv. The better T:N ratios of RAFT3 scFv compared with MFE-23 scFv and 4D5 scFv were simply attributable to better clearance from the normal tissues and were not attributable to higher tumor uptake. This was also the reason for the better T:N ratios obtained with MET scFv compared with RAFT3 scFv. This comparison suggests that none of the methodological factors examined appears to have been responsible for the observed differences in the performance of RAFT3 scFv, MFE-23 scFv, 4D5 scFv,and MET scFv. Rather, the performance of RAFT3 scFv observed in our study appears to have been a particular property of this scFv.

Although there are problems in extrapolating the results of biodistribution and pharmacokinetic studies from animals to humans and vice versa, radiopharmaceuticals generally exhibit much slower clearance in humans (70). However, this general rule does not seem to apply for scFv. Thus, from our study, t_1/2β for RAFT3 scFv in mice (≈3 h) was not very different from t_1/2β for MFE-23 scFv in humans (≈5 h) in the study by Begent et al.(29). The work by Begent et al.(29) is currently the only human study of the use of scFvs for ISG in the literature. In their study, they were able to obtain high quality ISG images at 4 and 22 h after injection of ¹²³I-labeled MFE-23 scFv(t_1/2 of ¹²³I,13 h). Therefore, we might reasonably expect similar values for t_1/2β in humans using RAFT3 scFv(compared with MFE-23) with similar improvements in the quality of the images obtained for ISG.

In contrast, similar extrapolations to humans are more difficult to make for LHM2 IgG1 because of the difficulty of anticipating the effects of greater background binding to normal tissues as well as the effects of the type of MAb, the type of radiolabel, and the antibody dosage used. The typical t_1/2β of mouse MAb in humans is 20 h with values as long as 181 h (71, 72, 73, 74), this compares with t_1/2β of ≈6 h for LHM2 IgG1 in our mouse study. However, if we assume that 20 h is a typical t_1/2β for IgG1 in humans(i.e., ≈3 times greater than t_1/2β for LHM2 IgG1 in mice), then a linear extrapolation from our animal data would predict maximum T:N ratios for LHM2 IgG1 in humans at ≈6 days (i.e., T:N ratios at maximum in mice at 48 h; therefore, T:N at maximum in humans at 144 h) postinjection. Waiting 6 days for maximum T:N ratios to be achieved before ISG is not a practical option for what should ideally be a routine investigation. Moreover, the t_1/2β of LHM2 IgG1 in mice is already longer than the t_1/2 of ^99mTc. Alternative radionuclides with a longer t_1/2 to match the pharmacokinetic properties of LHM2 IgG1 are available, but these then lack the ideal characteristics of ^99mTc for ISG.

Surprisingly we found that LHM2 F(ab′)₂ showed much poorer targeting properties compared with IgG1 or scFv. In contrast Siccardi et al. (25) have shown in clinical trials that F(ab′)₂ against HMW-MAA is superior to IgG1 for imaging. The poorer performance of LHM2 F(ab′)₂ in our own experiments may have been caused by experimental error or might be a particular property of LHM2 F(ab′)₂. Apart from this, the differences in the pharmacokinetic and biodistribution data between ¹²⁵I-labeled LHM2 IgG1 and RAFT3 scFv are in keeping with previously published data (24, 27, 29).

The purpose of this study was to determine the suitability of RAFT3 scFv for therapeutic and imaging trials in patients. Our data show that this scFv can be labeled directly with ^99mTc in a single step while remaining stable in vitro and in vivo. Our data also suggest that RAFT3 scFv exhibits properties in vivo that may make it superior to LHM2 IgG1 and LHM2 F(ab′)₂ for imaging trials. On the basis of these data, we intend to proceed with clinical trials of ^99mTc-labeled RAFT3 scFv for ISG of patients with metastatic melanoma.

We thank G. Winter for his gift of pUC119 His₆ Xba c-myc, I. Leigh for LHM2 MAb,and S. Mather for helpful discussions and critical reading of the manuscript.