Efficient tumor targeting by single-domain antibody fragments of camels

Abstract

The variable domain of functional heavy chain antibodies (VHH) devoid of light chains, found in camels, constitute the smallest intact antigen-binding domain fragment. Two camel single-domain fragments, cAb-Lys2 and cAb-Lys3, recognizing an overlapping epitope of lysozyme with a dissociation constant of 2 nM and 65 nM, respectively, and a bivalent cAb-Lys3 were investigated for their ability to target transgenic tumors expressing lysozyme on their membrane. Biodistribution studies revealed that these non-immunogenic monomeric and bivalent camel single-domain antigen binders specifically target lysozyme-expressing tumors and metastatic lesions. The excess of antibody is rapidly eliminated from the blood circulation and no cAb retention was observed in normal organs. The tumor to organ cAb-ratios at 2 and 8 hr were in the (2.1–10.8):1 and (6.2–23.7):1 range, respectively. The degree and specificity of tumor retention is independent of the affinity of the recombinant camel single-domain fragments for their antigen and from their univalent monomeric (15 kDa) or bivalent format (33 kDa). This study demonstrates the successful and specific in vivo targeting of tumors by camel single-domain fragments. It may open perspectives for their future use as tumor-targeting vehicle, due to their small size, soluble behaviour and because they are non-immunogenic and interact with epitopes that are less antigenic for conventional antibodies. © 2002 Wiley-Liss, Inc.

cAb, camel single-domain antibody; CDR, complementarity determining region; CH1, first constant domain of immunoglobulin heavy chain; %ID/g, percentage of injected dose per g of tissue; scid, severe combined immunodeficient; T:O, tumor/organ; V_HH, variable domain of a heavy chain of a camelid heavy-chain antibody; VH, variable domain of immunoglobulin heavy chain; VL, variable domain of immunoglobulin light chain; scFv, single-chain variable domain fragment.

Clinical research in the area of antibody-based tumor-targeted therapy has been driven for many years by the prospect of identifying surface antigens with sufficient restrictive tissue expression patterns to allow for the selective and specific accumulation of antibody in tumor tissue. Anti-tumor antibodies could target cancer cells selectively and could be used to deliver therapeutic or imaging agents to the tumor.1 Clinical trials however demonstrated important limitations of murine antibodies due to high immunogenicity, distribution to normal organs and poor penetration of solid tumors. Along with the recent progress in genetic engineering techniques, there have been major efforts to construct or engineer antibodies to obtain smaller binding units that retained the specificity and affinity of classical antibodies and/or to reduce the immunogenicity of the murine molecules (“humanisation”).2 The variable fragment (Fv) composed of the paired variable domain of the immunoglobulin heavy chain (VH) and the variable domain of the immunoglobulin light chain (VL) is the smallest, intact antigen-binding fragment one can obtain from a conventional antibody. However, it is more convenient to produce Fv as recombinant single-chain Fv (scFv), i.e., an Fv where the VH and VL domains are tethered by a flexible oligopeptide linker.3 To broaden the immunotherapeutic potential, more complex constructs have been engineered, e.g., by linking 2 different scFvs to bridge tumor cells with either T or NK cells (bispecific antibodies) or a scFv attached to a toxin or an enzyme to act on a prodrug.4 However, several of these scFv-based constructs proved difficult to express and purify, and exhibited several serious shortcomings in functionality. Common hurdles were the tendency to form aggregates due to the presence of an oligopeptide linker, the susceptibility of the linker to proteolytic cleavage and subsequent unfolding of the antibody constructs.5

The naturally occurring heavy-chain antibodies devoid of light chain and of CH1 domain that were discovered in camelids6 may constitute a promising alternative in this respect. The observation that camelids possess large amounts of functional heavy-chain antibodies lacking light chains formed the basis for generating functional single-domain antibody fragments (referred to as cAb for camel single-domain antibody)7, 8 from their variable domains (VHH). These small-sized molecules are well expressed and were shown to overcome to a large extent the solubility, aggregation and degradation problems often encountered with scFvs. Furthermore, they show good specificity towards their corresponding antigens and can be obtained with affinities comparable to scFvs.9, 10 Owing to these characteristics, it was proposed that cAbs will constitute perfect building blocks to tailor “at will” manifold constructs such as bispecific molecules or immunotoxins.11–13 In addition, the cAbs seem to recognize epitopes that are nonimmunogenic for conventional antibodies.8

Here we present the results of biodistribution studies in tumor-bearing scid mice performed with 2 cAbs directed against an overlapping epitope of lysozyme and differing 32-fold in affinity. In addition, we investigated whether increasing functional affinity (avidity) by complexing the monomeric cAb fragments into a homodimer, influences tumor targeting in vivo.

MATERIAL AND METHODS

Media and culture conditions

All cell lines were cultured in RPMI 1640 supplemented with 2 mM L-glutamine, 100 units/ml penicillin, 100 μg/ml streptomycin and 10% heat-inactivated fetal calf serum, and incubated at 37°C in a humidified incubator with 5% CO₂, 95% air.

Cell lines

The BW-Li variant was derived from the original BW5147 T-cell lymphoma (AKR origin, Salk Institute, La Jolla, CA) as described previously.14 The 3LL-R variant was generated from the Lewis Lung carcinoma (C57Bl/6 origin) as described by Remels and De Baetselier.15

Tumor cells were transfected by electroporation with vector pDisplay (Invitrogen, La Jolla, CA) encoding lysozyme, using a Biorad Gene Pulser (300 V, 960 μF). Stable transfectants were selected in medium containing neomycin. After 5 to 10 days, neomycin-resistant individual transfected clones were picked up and expanded in medium without antibiotic selection.

FACS analysis

Cells (1 × 10⁶) were incubated for 30 min with lysozyme-specific antibodies cAb-Lys2, cAb-Lys3 and (cAb-Lys3)₂. Subsequently, an optimal dilution of mouse anti-His-tag antibody (Serotec, Bicester, UK) was added, followed by incubation with FITC-labeled anti-mouse IgG (ICN Biomedicals, Costa Mesa, CA) according to the manufacturer's recommendations. Stained cells were washed twice with Dulbecco PBS and fluorescence was measured by FACScan (Becton Dickinson, San Jose, CA).

Cloning, expression and purification of cAb molecules

The isolation of the cAb-Lys2 and cAb-Lys3 antibody fragment was previously described.7 The genes were recloned in an expression vector under control of the lac promoter, between the Pel B leader signal and a carboxyterminal hexahistidine tail.8 For the bivalent construct, an expression vector that allows in tandem cloning of 2 VHH gene fragments spaced by a 29-mer polypeptide linker, was developed.13 The linker used to fuse the 2 cAb-Lys3 fragments, EPKIPQPQPKPQPQPQPQPKPQPKPEPMA, is derived from the upper structural hinge of the dromedary heavy-chain antibody isotype IgG2a.16 A hexahistidine tail is fused at the carboxy-terminus to facilitate purification by immobilized metal affinity chromatography (IMAC).

The monomeric cAb-Lys2, -cAb-Lys3 and the bivalent construct (cAbLys3)₂ were induced by IPTG in WK6 Escherichia coli cells as soluble periplasmic proteins and purified to homogeneity by 2 chromatographic steps, IMAC on NI-NTA resin (Qiagen, Chatsworth, CA) followed by gel filtration on a Superdex75 HR 10/30 column (Pharmacia, Gaithersburg, MD). Finally, the cAb proteins were concentrated using VIVASPIN 15 concentrators (Vivascience, Ltd., Binbrook Lincoln, UK). The functionality of all constructs was tested by solid-phase binding ELISA as described previously.7 Binding constants were measured using an IAsys Biosensor (Affinity Sensors, Cambridge, UK), in which lysozyme was immobilized covalently (surface density, 10 ng/mm²) on carboxy-methyl dextran cuvettes with the EDC/NHS coupling chemistry, following the manufacturer's instructions.

Radiolabeling and assay for immunoreactivity

Radiolabeled cAbs were prepared using N-succinimidyl 4-¹²⁵I- iodobenzoate (¹²⁵I-PIB reagent, NEN). Samples containing 0.5 mg of the purified recombinant protein in 0.5 ml 0.2 M NaHCO₃ pH 9.0 were incubated with 0.5 mCi (25.2 mCi/ml, 6.4 mCi/μg) of PIB reagent. Non-incorporated reagent was removed on NAP-5 desalting columns (Pharmacia) pre-equilibrated with PBS. The incorporation of ¹²⁵I in the proteins was evaluated after SDS-PAGE by autoradiography. The immunoreactive fraction of the radio-iodinated antibody fragments was determined using lysozyme, immobilized on CNBr-activated Sepharose beads (5 mg/ml gel).

Mice

Eight-week-old C.B17/IcrHanHsd-scid mice (Harlan, Nederland) were subscutaneously inoculated with BW-Li or BW-Li/LYS (2 × 10⁶) cells. Biodistribution studies were performed when tumors weighted 300 to 400 mg. For the metastasis model, 2 × 10⁶ 3LL-R or 3LL-R/LYS cells were injected intravenously into scid mice and biodistributions were performed 17 days after tumor cell injection, when pulmonary metastases were clearly present.

Biodistribution studies

Initial biodistribution studies (up to 2 hr) were performed in urethra ligated BALB/c mice. Groups of 4 mice were injected i.v. with 20 μg ¹²⁵I-PIB-labeled cAb-Lys3. Mice were sacrificed at several time points post-injection and tissues were collected. Tissues were weighed and retention of radioactivity was counted in a gamma counter, which allowed us to determine the percentage of the injected dose localized per tissue (% ID).

For biodistribution studies in tumor-bearing mice, the radiolabeled cAb preparations were diluted in PBS to a concentration of 50 μg/ml. Each mouse was given 200 μl of radiopharmaceutical by tail vein injection. Cohorts of 3 tumor-bearing scid mice were sacrificed at 2 and 8 hr (or 24 hr) after injection: the cAb retention in tumor, organs and blood were determined as described above. The mean and SE for each group of data were calculated, and T:O ratios were determined. Significance levels were calculated using Student's t test.

Immunogenicity studies

Female BALB/c mice (5 mice/group) were injected intravenously with 10 μg cAb-Lys3 on 3 consecutive days. The treatment was repeated after 3 and 6 weeks, respectively. Pre-immune sera were taken 1 day before the treatment and immune sera were taken weekly. Three weeks after the last injection, mice were killed. Sera and spleens were taken to analyze the immune response.

T cell proliferation and cytokine assays

Spleen cell suspensions from individual mice were prepared in supplemented RPMI 1640 medium (fetal calf serum 10%, penicillin-streptomycin 100 U and 100 μg/ml, respectively, L-glutamine 2 mM, 2-mercaptoethanol 5×10⁻⁵ M, MEM amino acid solution 1× and sodium pyruvate 1 mM). For T cell proliferation assays, 2×10⁵ cells were separately stimulated with 10 μg/ml of cAb-Lys3, 2 μg/ml Con A (Sigma Chemical Co., St. Louis, MO) or without antigen at 37°C in 96-well flat bottom tissue culture plates (Becton Dickinson, Franklin Lakes, NJ). After 4 days in culture, 1 μCi of ³H-thymidine was added to each well and the plates were incubated for a further 18 hr. The cells were harvested onto glass fiber filters, and the incorporated radioactivity was estimated. Results were expressed as the ³H-thymidine incorporation from triplicate cultures, subtracting background values from nonstimulated cells.

Cytokines (IFN-γ and IL-10) were quantified in cell culture supernatants taken 24, 48, 72 and 96 hr after priming. The cytokine levels were analyzed by a sandwich enzyme-linked immunosorbent assay (ELISA) in accordance to the supplier's instructions (Pharmingen, San Diego, CA).

Measurement of antibody titers

Antibody against cAb-Lys3 in the sera was determined with an enzyme-linked immunosorbent assay (ELISA) using 96-well plates coated with His-tagged cAb-Lys3, or with bovine serum albumin. After blocking with 10% FCS in PBS, diluted preimmune or immune sera were added to the plate. After washes, HRP-labeled goat anti-mouse antibodies were added and the assay was developed by adding 3,3′,5,5′-tetramethylbenzidine (Sigma Chemical Co). For each sample, optical density (OD at 490 nm) determined on bovine serum albumin-coated plates was subtracted from the OD values obtained on the cAb-Lys3-coated plates.

RESULTS

Antibody expression, labeling and characterization

The univalent cAbs and the bivalent derivative used in this study were expressed as soluble periplasmic proteins and purified by 2 chromatographic steps. No degraded or aggregated products could be discerned since the 3 molecules eluted from the Superdex75 column as a single peak and migrated on SDS-PAGE as single bands under reducing and non-reducing conditions.13 The average yield for the univalent cAbs was 2 mg,7 and nearly 1 mg for the bivalent cAb-Lys3¹³ per liter culture.

A kinetic analysis of the interaction of purified antibodies with immobilized lysozyme on an IAsys Biosensor allowed the determination of the association and dissociation rate constants and the calculation of the dissociation constant for cAb-Lys2 (2 nM) and cAb-Lys3 (65 nM) as reported earlier.7 From these experiments we learned that the higher affinity of cAb-Lys2 was mainly the result of a much slower dissociation-rate constant (k_off 1.8 × 10⁻⁴ s⁻¹vs. 2.7 × 10⁻³ s⁻¹ for cAb-lys3). Using the same technique for the bivalent antibody fragment, we obtained an association rate comparable to the monovalent cAb-Lys3 (k_on ∼60,000 M⁻¹ s⁻¹). However, the dissociation rate was much lower for the bivalent construct (k_off ∼7 × 10⁻⁴ s⁻¹), indicating that both subdomains may bind simultaneously to the immobilized lysozyme.13 The 3 purified Ab fragments were labeled with ¹²⁵I using PIB reagent. We avoided intentionally the Iodogen method which labels tyrosines since it is well established that these residues have a solvent-exposed location in the CDRs.17 It was therefore anticipated that their modification would affect the immunoreactivity of the cAb formats. In contrast, lysines have a higher propensity to occur in the framework, away from the antigen-binding loop. The immunoreactivity for each of the 3 PIB-labeled Ab molecules as determined by the percentage of binding to lysozyme-Sepharose, exceeded 90%. The purity of the monomeric and bivalent molecules was illustrated by SDS-PAGE (Fig. 1). Only 1 single band for each sample could be discerned, and degradation products or major contaminants were clearly absent.

Figure 1. Autoradiograph of a SDS-PAGE of radioiodinated cAb-Lys2 (lane 1), cAb-Lys3 (lane 2) and bivalent cAb-Lys3 (lane 3) under reducing (left) and nonreducing (right) conditions. The numbers on the left indicate the apparent Mr and position of migration of standard proteins run in parallel.

Establishment of murine tumor models expressing lysozyme as TAA

At the time we planned to study the tumor targeting potential of single-domain antibody fragments, we did not possess cAbs that specifically recognize a tumor-associated antigen. However, we disposed of an abundant set of cAbs with specificity for hen egg-white lysozyme, differing from one another in their affinity (dissociation constant range from 65 to 2 nM). In addition, bivalent constructs of the cAbs against lysozyme were available. Hence, to evaluate the potential of cAbs to target solid tumors and metastatic lesions, we established 2 transgenic tumor model systems, totally adapted to the well-characterized anti-lysozyme VHH named cAb-Lys2 and cAb-Lys3.

The BW-Li variant,14 derived from the murine non-invasive, non-metastatic BW5147 T-cell lymphoma was genetically modified with hen egg-white lysozyme. The lysozyme molecules are surface-exposed and accessible for all 3 lysozyme-specific camel antibody fragments as evidenced by FACS analysis (Fig. 2a). Consequently, the lysozyme operationally functions as a tumor-associated antigen of the transgenic BW-Li/LYS cell line for cAb-Lys2, cAb-Lys3, and bivalent cAb-Lys3.

Figure 2. Expression of lysozyme in BW-Li T cell lymphoma and 3LL-R Lewis Lung Carcinoma cells. (A) FACS analysis of BW-Li (black surface) and BW-Li/LYS (white surface) cells. (B) FACS analysis of 3LL-R and 3LL-R/LYS cells. Lysozyme surface expression was measured with cAb-Lys2 (a), cAb-Lys3 (b) or bivalent cAb-Lys3 (c) as described under Material and Methods.

Likewise, the 3LL-R variant derived from the Lewis Lung carcinoma was transfected with the lysozyme-encoding vector and yielded transgenic cell clones (referred as 3LL-R/LYS, Fig. 2b). The tumorigenic properties of 3LL-R cells and their potential to selectively induce pulmonary metastases after intravenous inoculation into syngeneic C57Bl/6 mice have been well established.15 To ascertain that the genetically modified derivative 3LL-R/Lys maintained the properties of parental 3LL-R cells, we investigated their ability to form pulmonary metastases in mice. For this purpose, 2 × 10⁶ 3LL-R or 3LL-R/Lys cells were injected into the lateral tail vein of mice at 8 weeks of age. At 14, 17 and 21 days after the injection of the tumor cells, animals were sacrificed and the lungs were excised and fixed to visualize the disseminated tumor foci. After 14 days, ∼30 nodules with a diameter of about 1 mm were detected for both 3LL-R and 3LL-R/Lys. The number increased to approximately 50 after 17 days and to 180 after 21 days. Here again, no significant difference in nodule numbers was found between the two tumor cell lines (data not shown). We conclude that 3LL-R/Lys cells retain their potential for organ-specific metastasis formation. Based on the extent of metastasis formation, we decided to perform the biodistribution studies at day 17 when around 50 overt tumor nodules could be detected on the lung surface.

Pharmacokinetics

BALB/c mice were injected intravenously with ¹²⁵I-PIB-labeled cAb-Lys3 to study its pharmacokinetics. Mice were sacrificed at 15, 30 min, 1, 2, 4, 12 and 24 hr post-injection and several tissues and cumulative urine samples were collected over the first 2 hr. (Fig. 3). The VHH distribution in organs shows high initial levels in kidneys (64% ID/g or 23% ID at 15 min. post-injection), characteristic of a renal excreted protein of low Mr. Accumulation of radioactivity in the intestines peaks by 2 hr post-administration, suggesting rapid processing in the liver with hepatobiliary excretion of the lysine adduct of the ¹²⁵I-PIB-radiolabel after degradation of the protein. This is consistent with peak levels in the liver (3% ID/g at 30 min. post-injection), which are declining by 1 hr. As anticipated, this low Mr protein cleared rapidly from the blood, and an elimination half-life of 1.5 hr was calculated for this tissue. Cumulative fraction eliminated by urinary excretion appeared to plateau at about 35% ID by 2 hr post-injection (the animal model does not allow for urine collection at later timepoints).

Figure 3. Mean %ID at multiple time points post i.v. administration of ¹²⁵I-labeled cAb-Lys3 to BALB/c mice. Mean values are derived from n = 4 mice.

Biodistribution studies: the BW-Li tumor model

To evaluate the tumor targeting potential of cAbs, biodistribution studies were performed in scid mice bearing 300 to 400 mg of either subcutaneous BW-Li or BW-Li/LYS tumors using 10 μg of purified radioiodinated cAb-Lys2, cAb-Lys3 or bivalent cAb-Lys3. The biodistribution results (Table I) revealed a specific retention of all 3 cAbs on BW-Li/LYS tumors as compared to parental BW-Li tumors, at both 2 hr (ratio (3.3–6.4):1; cAb-Lys2: p = 0.0004; cAb-Lys3: p = 0.003; (cAb-Lys3)2: p = 0.01) and 8h (ratio (6.2–8.4):1; cAb-Lys2: p = 0.02; cAb-Lys3: p = 0.005; (cAb-Lys3)2: p = 0.01) after administration of the respective antibodies. (The values of the %ID/g 24 hr post injection dropped to ∼0.05 for BW-Li/LYS tumors. This was approximately 10 times the values for normal tissues, values that could be barely counted). The antibody fragment distribution in normal organs in scid mice bearing BW-Li/LYS or BW-Li tumors was similar. As expected, these small-sized molecules cleared rapidly from the blood and showed high initial levels in the kidneys (14.27 to 23.25 % ID/g at 2 hr post-injection), characteristic of a renal excreted protein of low Mr. Although cAb-Lys2 has about 32-fold higher affinity for the antigen compared to cAb-Lys3 as a result of a lower off-rate (Kd = 2 nM; k_off = 1.8×10⁻⁴ s⁻¹ and 65 nM; k_off = 2.7×10⁻³ s⁻¹, respectively), both were equally performing to target the BW-Li/LYS tumors after 2 hr (cAb-Lys2: 2.65 % ID/g; cAb-Lys3: 2.69 % ID/g). Moreover, dimerization of cAb-Lys3 did not result in a larger % ID/g retained in tumor tissue as compared with the corresponding cAb-Lys3 monomer.

Table I. Comparison of Biodistribution of ¹²⁵I-Labeled cAb-Lys2, cAb-Lys3 and Bivalent cAb-Lys3 in Tumor-Bearing scid Mice

Organ	2 hr						8 hr
	cAb-Lys2 (Kd = 2 nM)		cAb-Lys3 (Kd = 65nM)		(cAb-Lys3)2		cAb-Lys2 (Kd = 2 nM)		cAb-Lys3 (Kd = 65 nM)		(cAb-Lys3)2
	%ID/g	T:O	%ID/g	T:O	%ID/g	T:O	%ID/g	T:O	%ID/g	T:O	%ID/g	T:O
^a Iodinated antibody fragments were administered to scid mice bearing a s.c. BW-Li/LYS or parental BW-Li tumor (≈ 300–400 mg) by i.v. tail vein injection (10 μg per mouse). Cohorts of 3 mice were killed 2 and 8 h after injection, and organs were weighed and radioactively counted. Targeting results are expressed as %ID of ^125I-labeled cAb retained per g of tissue (or ml blood) from a representative experiment; 3 separate experiments were carried out with similar results. Unless otherwise indicated, SEs were <25% of the indicated value. T:O ratios are also reported. 2 SE = 44%. 3 SE = 33%. 4 SE = 36%. 5 SE = 34%. 6 SE = 39%. 7 SE = 33%. 8 SE = 39%.
BW-Li/LYS
Blood	0.72	3.70	1.06	3.22	0.963	2.64	0.026	15.05	0.029	10.44	0.035	8.55
Lung	0.47	5.62	0.54	5.59	0.74	2.76	0.044	9.24	0.028	11.25	0.047	6.20
Spleen	0.25	10.61	0.39	7.19	0.19	10.77	0.0415	7.67	0.209	1.45	0.012	23.72
Stomach	1.162	2.09	1.18	2.42	0.27	8.52	0.035	12.76	0.035	8.81	0.054	9.90
Liver	0.73	3.62	0.60	4.56	0.384	6.53	0.033	12.06	0.034	8.93	0.022	13.37
Kidney	19.23	0.13	15.26	0.15	23.25	0.09	1.697	0.27	1.064	0.29	1.760	0.18
Tumor	2.65	2.69	2.15	0.4106	0.30	0.2927
BW-Li
Blood	0.45	0.92	0.49	0.80	1.26	0.48	0.059	0.94	0.041	1.15	0.041	0.99
Lung	0.44	0.89	0.32	1.22	0.73	0.85	0.041	1.23	0.033	1.46	0.035	1.20
Spleen	0.20	2.09	0.37	1.07	0.25	2.50	0.051	0.96	0.163	0.28	0.012	4.37
Stomach	0.35	1.18	0.41	0.92	0.92	1.10	0.060	0.97	0.032	1.44	0.025	1.70
Liver	0.47	0.89	0.47	0.83	0.57	1.11	0.033	1.51	0.034	1.38	0.018	2.30
Kidney	15.53	0.02	14.27	0.02	20.18	0.04	0.793	0.07	0.863	0.06	1.897	0.05
Tumor	0.41	0.39	0.62	0.049	0.0488	0.045

When analyzing the biodistribution at the 8 hr time point, tumor is still targeted by all 3 cAb formats to a similar extent, thereby achieving tumor: blood ratios > 8.5. We did not observe a specific retention of the cAb formats in normal organs (tumor: normal organs ratios ranged from 6.20 to 23.72) except for the kidneys, where a significant Ab accumulation was still observed (cAb-Lys2: 1.70 % ID/g; cAb-Lys3: 1.06 % ID/g; (cAb-Lys3)₂ : 1.76 % ID/g).

The 3LL-R tumor metastasis model

To evaluate the potential of cAbs to target metastases, the biodistribution of the univalent cAb-Lys2 and cAb-Lys3 monomers and bivalent cAb-Lys3 was studied in scid mice bearing either parental 3LL-R or transgenic 3LL-R/LYS pulmonary metastases. The pulmonary localization of all three cAb fragments was significantly greater in scid mice bearing 3LL-R/LYS pulmonary metastases than that in animals bearing parental 3LL-R metastatic nodules in their lungs (cAb-Lys2: p = 0.005; cAb-Lys3: p = 0.0001; (cAb-Lys3)₂: p = 0.02). As for the BW-Li tumor model, a relatively high accumulation of the cAb-Lys2 and cAb-Lys3 monomers in the stomach of 3LL-R/LYS pulmonary metastases-bearing mice was observed (1.11–3.34 % ID/g tissue) 2 hr after their administration. In addition, a significant retention of the univalent cAbs in the blood of mice carrying 3LL-R/LYS metastases was observed as compared to the 3LL-R-bearing control mice. At 8 hr, a specific pulmonary targeting was still observed for cAb-Lys2 and the bivalent cAb-Lys3 but not for the monovalent cAb-Lys3 in 3LL-R/LYS tumor-bearing mice, whereas the Ab distribution in normal organs was similar in 3LL-R and 3LL-R/LYS tumor-bearing mice. Comparison of the results obtained with univalent and bivalent cAbs revealed a slightly higher lung (metastases): normal organ ratio for the bivalent cAb-Lys3 (Table II).

Table II. Comparison of Biodistribution of ¹²⁵I-Labeled cAb-Lys2, cAb-Lys3 And Bivalent cAb-Lys3 in Pulmonary Metastases-Bearing scid Mice1

Organ	2 h			8 h
	cAb-Lys2 Kd = 2 nM %ID/g	cAb-Lys3 Kd = 65nM %ID/g	(cAb-Lys3)2 %ID/g	cAb-Lys2 Kd = 2 nM %ID/g	cAb-Lys3 Kd = 65 nM %ID/g	(cAb-Lys3)2 %ID/g
1 Iodinated antibody fragments were administered to scid mice bearing pulmonary 3LL-R/LYS or parental 3LL-R metastases by i.v. tail vein injection (10 μg per mouse). Cohorts of 3 mice were killed 2 and 8 hr after injection, and organs were weighed and radioactively counted. Targeting results are expressed as %ID of ^125I-labeled cAb retained per g of tissue (or ml blood) from a representative experiment; 2 separate experiments were carried out with similar results. Unless otherwise indicated, SEs were < 30% of the indicated value. T:O ratios are also reported. 2 SE = 36%. 3 SE = 54%. 4 SE = 30%. 5 SE = 36%. 6 SE = 34%.
3LL-R/LYS
Blood	1.652	3.034	0.85	0.12	0.07	0.08
Lung	1.75	1.85	1.27	0.23	0.12	0.31
Spleen	0.323	0.65	0.14	0.05	0.15	0.02
Stomach	1.113	3.34	0.28	0.69	0.17	0.17
Liver	0.85	1.27	0.30	0.33	0.08	0.12
Kidney	24.33	36.71	11.3	3.11	3.11	2.68
3LL-R
Blood	0.53	0.75	1.50	0.03	0.02	0.05
Lung	0.24	0.31	0.426	0.03	0.10	0.11
Spleen	0.20	0.53	0.22	0.12	0.28	0.14
Stomach	0.555	0.36	2.79	0.04	0.09	0.04
Liver	0.60	0.45	0.83	0.04	0.04	0.02
Kidney	19.25	32.44	20.97	0.95	0.81	1.98

Immunogenicity of camel single-domain fragments

A drawback associated with antibody-based therapy is the development of an immune response by the host to treatment with the antibodies. Therefore, it was of interest to see whether repeated treatment with cAbs induces an immune response. To this end, mice received repeated intravenous infusions of 10 μg cAb-Lys3 based on other studies employing similar protocols.18–21 Blood samples were collected before, during and after the treatment for anti-cAb antibody response. No cAb-Lys3-specific antibodies could be detected in sera taken before, during or after treatment from all mice (Fig. 4a), indicating that no mouse anti-cAb antibodies are produced. To investigate possible cell-mediated immune responses induced by cAb-Lys3 therapy, spleen cells from treated mice were restimulated with antigen in vitro. No lymphocyte proliferation was observed in response to cAb-Lys3 (data not shown). In addition, negligible levels of IFN-γ (< 80 pg/ml) and no IL-10 were produced by these spleen cells upon restimulation with cAb-Lys3. whereas cytokine secretion could be observed from cells non-specifically stimulated with mitogen (Fig. 4b).

Figure 4. Humoral (a) and T cell (b) responses of BALB/c mice receiving multiple intranvenous infusions of cAb-Lys3. Production of IFN-γ and IL-10 was quantified by sandwich ELISA in culture supernatants from spleen cells taken 3 weeks after the last antibody infusion and stimulated with cAb-Lys 3 or concanavalinA. The presence of cAb-Lys3 specific antibodies was assessed by ELISA with a 1:100 dilution of sera from individual mice (IMS) taken 3 weeks after the last antibody administration. Negative and positive controls were sera from normal mice (NMS) and from mice immunised with cAb-Lys3 emulsified in CFA (PS).

DISCUSSION

To be successful as imaging or therapeutic agent for cancer, antibody-based molecules should be small enough to reach the tumor-associated antigen but, in the case of therapeutic antibodies, still large enough to remain for apt periods of time in the circulation. Once the mass of the antibody fragment becomes less than 60 kDa, their administration will require careful management to maintain the blood concentrations required to permit diffusion into the tumor. In this respect, increasing the functional affinity by decreasing the k_off rate, may help to “trap” the antibody fragment that diffuses into the tumor. This can be accomplished by manipulating the intrinsic affinity properties22 or through the creation of multivalent antibody fragments.

Hitherto, the performance of single-domain antibodies (VH from conventional antibodies) as tumor targeting reagent has not yet been evaluated. Cloning of VHs from an immunized mouse and subsequent screening of clones proved that it is feasible to isolate antigen binders.23 However, these single-domain antibodies exhibited serious shortcomings. The antigen-binding site is matured during the immunization as a combination of VH and VL. Upon removal of the VL domain, the affinity for the antigen drops, and a large hydrophobic surface becomes exposed to the aqueous environment rendering the VH sticky and difficult to handle.23, 24 In nature, single-domain antigen-binding units are naturally occurring in camels and llamas as part of their unique heavy-chain antibodies. The variable domains of these camel heavy-chain antibodies (VHHs) possess a high sequence homology to human VHs. Moreover, specific mutations in the framework 2 region of the variable domain of camel VHH render them more hydrophilic.25 As such, these VHHs are expected to overcome the problems of stickiness encountered with conventional VH. Due to their single-domain nature, no oligopeptide linkers are necessary to keep different antibody domains together, thereby avoiding the problem of aggregation and instability which are frequently encountered with scFv.5

We have investigated the tumor targeting properties of two camel single-domain antibodies, cAb-Lys2 and cAb-Lys3, both in a univalent format and the latter also in bivalent form. The epitope mapping and lysozyme-inhibition experiments indicated that cAb-Lys2 and cAb-Lys3 bind to an overlapping epitope on hen-egg white lysozyme (S. Muyldermans, unpublished results) with this in contrast to earlier published data indicating that cAb-Lys2 and cAb-Lys3 recognized different epitopes.7 Their dissociation constants are in the nanomolar range (2 nM and 65 nM, respectively), which is typical for monovalent antibody fragments isolated from immunized animals. These affinities are also in the range of the best scFv or Fab retrieved from extremely large phage libraries carrying naive or synthetic antibody fragments.26–28 By selecting camel single-domain antibodies with overlapping epitopes, we excluded the possibility that differences in targeting performance may be due to different degrees of accessibility of the epitopes within the tumor. This allowed us to determine the effect of the off-rate component of affinity or avidity on the amount of antibody associated with tumor at different times after intravenous injection.

We established 2 transgenic tumor models, totally adapted to the well-characterized cAbs. The introduction of the hen egg-white lysozyme gene into BW-Li and 3LL-R tumor cells resulted in transgenic cell lines that expressed lysozyme on their cell membrane and were recognized by the lysozyme-specific camel antibody fragments. In contrast, these antibody fragments did not bind to the parental BW-Li and 3LL-R tumor cells. Therefore, these tumor cell lines provided an appropriate control for the specificity of the cAbs to target their tumor-associated antigen in solid tumors and lung metastases, respectively.

All ¹²⁵I-labeled cAb formats examined in this study were rapidly cleared from the blood and antigen-negative organs at the time points tested, in accordance with their molecular weight (15 and 33 kDa for the mono-domain and dimerized format, respectively) below the renal threshold and their non-sticky behavior. Both in the solid tumor and lung metastases settings, targeting was strictly dependent on lysozyme-expression, corroborating the tumor cell specificity of the cAbs used.

However, mice bearing 3LL-R/LYS micrometastases in their lungs had a significant retention (0.85–1.65 %ID/ml) of the antibody molecules in their blood circulation 2 hr after i.v. administration. A lysozyme-specific RT-PCR performed on blood samples from 3LL-R/LYS tumor-bearing mice showed the presence of lysozyme-encoding transcripts. Moreover, in vitro culture of blood samples from 3LL-R/LYS tumor-bearing mice resulted in the outgrowth of tumor cells that were recognized by all 3 cAbs as evidenced by flow cytometry (data not shown). These data indicate that 3LL-R/LYS tumor cells are continuously circulating in the blood and act as a scavenger for the cAb. This finding is further supported by the absence of a specific retention of cAbs in the blood circulation of mice bearing 3LL-R lung metastases although tumor cells could also be retrieved upon in vitro culture of blood samples (data not shown).

In both tumor model systems, the monovalent cAbs as well as the bivalent cAb-Lys3 preferentially localized at the tumor site after injection and remained as such by 8 hr. There was no significant difference in T:B ratio for the higher affinity cAb, nor for the bivalent cAb, indicating that the affinity (avidity) does not always correlate with the tumor targeting potential of antibodies. These findings are contradictory to earlier published data, describing a significant impact of affinity on the in vivo tumor-specific retention potential of scFv molecules.29, 30 A possible explanation for this discrepancy may be that the number of folded lysozyme molecules expressed on the tumor cells may not reach the concentration at which the elevated affinity or avidity could have a measurable effect. The low FACS signal obtained upon binding of the cAb molecules to the transgenic tumor cells (Fig. 2) supported such low expression levels of folded lysozyme on the cell surface. Measurement of the number of lysozyme molecules on the surface of BW-Li/LYS and 3LL-R/LYS tumor cells using a fluorescence quantitation assay, calculated an average of 2.1×10³ (3LL-R/LYS) and 8.8×10³ (BW-Li/LYS) molecules on each cell. This is far beyond the antigen density calculated for human tumor cell lines such as the human ovarian carcinoma SK-OV-3 cells overexpressing HER2/neu (1.0×10⁶ HER2/neu per cell31) and the renal cell carcinoma cell line SN12P (2.9×10⁵ p97 melanotransferrin per cell32). Furthermore, the spatial distribution of these lysozyme molecules on the cell surface may further hamper the simultaneous binding of the 2 subdomains of the bivalent cAb, thereby reducing the favorable effect of avidity. To address this question, we screened more than 80 lysozyme transgenic cell clones but failed to identify clones with higher expression of target antigens than the ones used in this study. Alternatively, it may be possible that the cAb molecules used in this study display a threshold affinity value above which no further improvements in tumor targeting can be achieved. Adams et al.33 examined a series of affinity mutants of the scFv molecule C6.5 and demonstrated that quantitative tumor retention did not significantly increase with enhancements in affinity beyond 10⁻⁹ M. This finding is in agreement with the theory of Fujimori et al.34 who postulated that high affinity antibodies will not successfully penetrate deeply into tumors due to a binding site barrier effect, in which interaction with the first antigen encountered at the periphery of the tumor will block further penetration and diffusion of the antibody into the tumor. Finally, the rapid renal clearance of cAb molecules may blunt the impact of improved affinity on tumor targeting by decreasing opportunities for these molecules to diffuse to tumor sites.

In conclusion, our results demonstrate that camel single VHH domains are non-immunogenic molecules that exhibit a rapid pharmacokinetic clearance and specifically target solid tumors and metastatic lesions. This property, together with the ease to produce functional and stable cAb constructs at grams per liter,35 and especially the fact that they recognize epitopes that are less antigenic for conventional antibodies8 may offer particular advantages when targeting tumors for diagnostic or therapeutic purposes.

Acknowledgements

We thank Dr. J. Reno (NeoRx Corporation, Seattle, WA) for his interest in camel single domain antibodies and performing initial pharmacokinetic studies. All animal experiments were carried according to the guidelines for animal welfare.