Skip to main content
NCBI home page
Protein Engineering, Design and Selection logoLink to Protein Engineering, Design and Selection
. 2008 Oct 28;22(3):209–216. doi: 10.1093/protein/gzn055

Engineered humanized diabodies for microPET imaging of prostate stem cell antigen-expressing tumors

Jeffrey V Leyton 1, Tove Olafsen 1, Mark A Sherman 3, Karl B Bauer 1, Patrick Aghajanian 1, Robert E Reiter 2, Anna M Wu 1,4
PMCID: PMC2644405  PMID: 18957406

Abstract

We have previously demonstrated preclinical in vivo targeting of prostate stem cell antigen (PSCA) using a humanized anti-PSCA 2B3 monoclonal antibody (mAb). However, humanization resulted in 5-fold loss of apparent affinity relative to the parental mAb (1 nM). In this study, diabodies (scFv dimers of 55 kDa) were generated from 2B3 including variants with different linker lengths as well as back-mutations to original murine residues to improve affinity. Parental 2B3 (p2B3) and back-mutated 2B3 (bm2B3) diabodies (Dbs) with five- or eight-amino acid linkers (p2B3-Db5, p2B3-Db8, bm2B3-Db5 and bm2B3-Db8) were evaluated for binding to PSCA by flow cytometry and affinities were determined by surface plasmon resonance. Back-mutation restored the affinity from 5.4 to 1.9 nM. Stability, evaluated by size exclusion, revealed that diabodies with eight-residue linkers existed as a mixture of dimeric and monomeric species at low concentrations (≤1 mg/ml). Shortening the linker from eight to five residues improved dimer stability, notably in the bm2B3-Db8 compared with bm2B3-Db5. Both p2B3-Db8 and bm2B3-Db8 were radioiodinated with 124I and evaluated by serial micro-positron emission tomography imaging in mice bearing LAPC-9 human prostate cancer xenografts. Localization in LAPC-9 xenografts was seen at 4 h, whereas at 20 h most of the activity had cleared from the tumor. Highest tumor-to-background contrast ratios and best images were obtained at 12 h. Although the higher affinity bm2B3-Db8 demonstrated improved tumor retention at later time points (20 h), it did not improve tumor targeting or imaging compared with p2B3-Db8 at 12 h.

Keywords: affinity maturation, diabody, PET, prostate cancer, PSCA

Introduction

The humanized 2B3 monoclonal antibody (hu2B3 mAb) can localize to and image the GPI-linked prostate stem cell antigen (PSCA) in xenograft-bearing mice (Olafsen et al., 2007). PSCA has been shown to be a promising target for antibody-based imaging and therapy of prostate cancer (Saffran et al., 2001; Olafsen et al., 2007) due to its abundant expression in the majority of primary and androgen-independent prostate cancers specimens and its correlation with higher disease stage (Gu et al., 2000; Lam et al., 2005). An analysis of circulating tumor cells by RT–PCR suggested that PSCA was more promising as a molecular diagnostic than prostate-specific antigen (PSA) or prostate-specific membrane antigen (PSMA) in men with prostate cancer (Hara et al., 2002). Additionally, PSCA expression has been associated with a subset of prostate epithelial cells that were shown to represent a cancer-specific transitional population (Tran et al., 2002). A major goal for prostate cancer imaging is to achieve more accurate disease detection and characterization (Hricak et al., 2007). The mainstays for detection and staging of prostate cancer continue to be ultrasound, CT, MRI and MRS, although nuclear medicine approaches are expanding. Much work has been with metabolic-specific probe development, with [11C]-choline, [18F]-choline and [11C]-acetate appearing especially promising as tracers sensitive to the increased fatty acid metabolism exhibited by these malignancies (reviewed in Bouchelouche and Oehr, 2008a, 2008b). The development and approval of [111In]-capromab pendetide (ProstaScint™) represented one of the first examples of an antibody imaging agent targeting a tissue-specific marker in tumors, namely PSMA.

Prostate-associated antigens, such as PSCA, PSMA and Mindin/RG-1, have been targeted using radioiodinated antibodies for tumor imaging in preclinical studies (Smith-Jones et al., 2003; Parry et al., 2005; Olafsen et al., 2007). Despite the ability of antibodies to effectively deliver high tumor radioactive uptake, their prolonged blood residence results in high background signal making early imaging of cancer lesions difficult. The antibodies targeting PSMA, Mindin/RG-1 and PSCA have all required days (2–7 days) for the blood to clear enough in order to obtain high contrast tumor images. For PSMA, improvements that produced good tumor-to-blood ratios at early time points were achieved by enzymatically reducing the molecular weight of PSMA-targeting antibodies (Carter et al., 2004). Radioiodinated Fab and F(ab’)2 fragments displayed optimal targeting times at 24 and 48 h, respectively.

Diabodies (Dbs; 55 kDa) are bivalent antibody fragments composed of two non-covalently bound scFv fragments. Their bivalency has been shown to enhance tumor retention compared with monovalent fragments of equal molecular weight (Adams et al., 2006). Diabodies are engineered into dimeric species through manipulation of the linker peptide connecting the variable (V) domains in either the VL–VH or the VH–VL orientation. Generally, the use of linker from 3 to 12 amino acids has been shown to promote the association to a diabody (Holliger et al., 1993; Perisic et al., 1994; Wu et al., 1996; Dolezal et al., 2000). Reducing the linker length further results in different patterns of oligomerization (Dolezal et al., 2000). However, linker length rules are not rigid, and the transitions between dimer, trimer and tetramer forms can be highly dependent on the individual antibody under study. For example, scFvs constructed from a CD22-specific antibody by direct fusion into VL-0-VH constructs resulted in an almost pure dimer population (Arndt et al., 2004), yet the affinity of the anti-CD22 dimer was similar to its monovalent scFv counterpart, suggesting that only one Fv arm might be functionally active due to steric constraints. Poor tumor targeting due to possible steric constraints and monovalent binding was observed for an anti-murine laminin-1 scFv with a five amino acid linker (Huang et al., 2005). Of interest, other linker-less scFvs form multimers which can overcome poor binding activity, as shown in a recent study of a zero-linker anti-Lewis Y scFv that multimerized into trimer/tetramer species and demonstrated excellent in vivo targeting (Kelly et al., 2008). In addition to linker length, the thermal instability and dissociation equilibrium of the VL–VH interface varies from antibody to antibody and can add complexity to dimer stability.

Diabodies targeting carinoembryonic antigen (CEA), HER-2, TAG-72 and Lewis Y have demonstrated promising targeting in preclinical studies (Adams et al., 1998; Pavlinkova et al., 1999; Wu et al., 1999; Power et al., 2001). Positron emission tomography (PET) imaging of colorectal and breast cancer xenografts in mice with anti-CEA and anti-HER-2 diabodies have shown that high tumor to background ratios can be achieved at early time points with this fragment (Sundaresan et al., 2003; Robinson et al., 2005; Cai et al., 2007). A systematic analysis of the relationship between tumor imaging and antibody size [intact versus F(ab’)2 versus minibody versus diabody versus scFv] predicted that the diabody would be the optimal candidate for same day imaging, particularly when paired with short half-life radionuclides (Williams et al., 2001). However, functional parameters such as optimal imaging time, required affinity and stability of dimeric diabodies remain unspecified and largely unexplored for prostate cancer detection.

An issue in repetitive administration of mAbs in the clinic is the response to xenogeneic antibodies by the human immune system. Grafting the complementary determining regions (CDRs) onto human antibody frameworks decreases the human anti-mouse immune response. For this reason, humanization of the PSCA-specific murine mAb 1G8 was performed (Olafsen et al., 2007). However, the humanization resulted in a 5-fold decrease in apparent affinity. Nevertheless, the resulting humanized 2B3 mAb demonstrated comparable targeting and imaging with the 1G8 mAb in PSCA-expressing xenografts in vivo (Olafsen et al., 2007), suggesting that the change in affinity did not impair the biological performance of intact antibody. The assembly of an anti-PSCA 2B3 minibody (VL-18-residue linker-VH-IgG1 hinge-CH3 dimer; 80 kDa) has been described (Leyton et al., 2008). The 2B3 minibody localized well to PSCA-expressing xenografts in vivo when radioiodininated with the positron emitter 124I (t1/2 = 4.2 days) and evaluated in mice by PET. However, reformatting the hu2B3 mAb to a minibody resulted in a further 9-fold loss of apparent affinity.

Here, we have engineered four 2B3 diabody variants which differed in linker length and affinity. The goal was to generate a stable, rapid clearing antibody fragment with specificity for PSCA and evaluate its imaging properties for prostate cancer in xenograft-bearing mice. The parental 2B3 diabody (p2B3-Db) was subjected to molecular modeling and six amino acids in the V domains were identified to be potentially affecting its three-dimensional structure and binding to PSCA. These were mutated back to the original mouse residues to produce a high affinity back-mutated 2B3 diabody (bm2B3-Db). Use of either five- or eight-amino acid residue linkers yielded four diabodies, p2B3- and bm2B3-Db5 and p2B3- and bm2B3-Db8, which were evaluated for dimer formation and stability in vitro. The p2B3- and bmDb8 were labeled with 124I and evaluated for tumor targeting and microPET imaging of prostate cancer xenografts in order to assess the impact of affinity on in vivo delivery.

Materials and methods

Diabody gene assembly and affinity maturation strategy

An scFv fragment in the VL–VH orientation with an 18 amino acid linker and a six histidine tag was amplified from a plasmid encoding the 2B3 minibody (Leyton et al., 2008). In order to generate diabodies, V genes were initially amplified separately with primers designed to encode either a GGGSGGGG (8) or a GGGGS (5) amino acid linker between the VL and the VH domains. The full diabody genes were then assembled by overlap extension PCR. The resulting two constructs (Db5 and Db8) were either subcloned into the pEE12 mammalian expression vector as previously described (Leyton et al., 2008) or inserted into the pSYN1 bacterial expression vector (Bell-Pedersen et al., 1996).

A refined molecular model (bm2B3) of 2B3 Fv was built using a previously described model as a template for the framework (Olafsen et al., 2007) and several recently published crystal structures (PDB files 1OTS, 1DQD, 1E6J, 1PSK, 1AD0, 1MIM, 1IAI, 12E8, 1QOK, 1IG3, 1IIF and 1KXT) were selected as templates for the CDR loops due to their high sequence identity with 1G8. Individual framework residues that packed against CDR-L1 and CDR-H2 (the two most highly exposed antigen-binding loops) were identified and their roles in maintaining CDR loop structure were examined. On the basis of this evaluation, six framework residues, two in the VL chain (Kabat L70, L71) and four in the VH chain (Kabat H48, H49, H66 and H69), were selected for back-mutation to their murine counterparts. Subsequent DNA base substitutions were performed by overlap extension PCR.

Expression and purification of proteins

Transfection of the p2B3- and bm2B3-Db8 in the pEE12 mammalian expression vector was performed as described (Leyton et al., 2008). Screening for high expressing clones was performed by western blots; membranes were probed with an anti-penta-His antibody (1:10 000) (Qiagen, Valencia, CA, USA), washed and then detected using an alkaline phosphatase (AP)-conjugated goat anti-mouse (Fc specific; 1:5000) antibody (Jackson ImmunoResearch, West Grove, PA, USA).

The scFv fragment, with an 18 amino acid linker, and p2B3- and bm2B3-Db5 diabodies were expressed bacterially by transforming Escherichia coli TG1 cells with pSYN1 containing the specified gene. Cells were plated on 2TY/AMP-2% glucose and incubated overnight at 37°C. Single colonies were picked and grown at 37°C overnight with shaking at 250 rpm in 10 ml of 2TY/AMP-2% glucose. The overnight cultures were inoculated into 500 ml 2TY/AMP-glucose and grown to an OD600 of 0.6–0.8. Protein expression was induced by the addition of IPTG (1 mM final) at 30°C shaking for 4 h. Pelleted cells were suspended in 10 ml of ice-cold periplasmic extraction buffer [30 mM Tris–HCl (pH 8.0), 20% (w/v) sucrose, 1 mM EDTA], 200 units of lysozyme (Epicentre, Madison, WI, USA) were added and incubation was performed at room temperature for 10 min. Lysozyme-treated bacteria were pelleted by centrifugation and the supernatant was dialyzed overnight against PBS at 4°C.

Both mammalian and bacterially expressed proteins were purified using a 1 ml Ni-NTA-agarose column (BioRad Laboratories, Hercules, CA, USA). The column was washed with Ni-column buffer (PBS, 30 mM imidazole, 300 mM NaCl, 0.05% Tween-20, pH 8.0) before eluting the bound protein with 250 mM imidazole in Ni-column buffer. Eluted proteins were dialyzed against PBS using Slide-A-Lyzer Dialysis Cassettes (10 000 mwco) (Pierce, Rockford, IL, USA) or porous membrane tubing (10 000 mwco) (Spectrum Laboratories, Rancho Dominguez, CA, USA), and concentrated using a Vivaspin 20 concentrator (10 000 mwco) (Vivascience, Hannover, Germany).

Biochemical characterization

Purified proteins were analyzed for purity by SDS–PAGE (4–20% Tris–HCl) (BioRad Laboratories). Specific binding was demonstrated by flow cytometry using PSCA transfected SKW 6.4 cells (Leyton et al., 2008). Briefly, 1 µg of diabody or scFv was incubated with 5 × 105 cells on ice for 30 min. Cells were washed in PBS/1% FBS and incubated with anti-penta-His antibody (1:100) (Qiagen) on ice for 30 min. Cells were washed again and incubated with PE-conjugated F(ab’)2 (1:200) goat anti-mouse (Jackson Immunoresearch). Data were acquired and analyzed as previously described (Leyton et al., 2008). Apparent affinities of p2B3- and bm2B3-Db8 were determined by surface plasmon resonance (SPR) on a BiaCore X with immobilized PSCA as previously described (Olafsen et al., 2007).

Diabody stability properties

Diabody samples were prepared at 1 mg/ml and higher concentrations with a Vivaspin 20 concentrator (Vivascience). Approximately 250 µg of protein was loaded on a Superdex 75 chromatography column in PBS (10 ml bed volume) at a flow rate of 0.5 ml/min for 50 min. Elution was monitored by UV absorption at 280 nm.

For analysis of the concentration dependence of dimer formation over time, 250 µg of five- or eight-linker p2B3- and bm2B3-Db (loading concentrations >1 mg/ml) were injected separately on to a Superdex 75 column and a fraction containing both the monomer and the dimer species (at low concentration) was collected and stored at 4°C. After 24, 48 and 168 h, 0.1 ml aliquots were reinjected onto the Superdex 75 column and analyzed. Following the final time point (168 h), the remaining aliquots were concentrated and also analyzed by size exclusion. The percentage of dimer versus monomer at 24, 48 and 168 h when re-concentrated was estimated.

PSCA-tumor imaging and analysis

All animal studies were performed under an approved UCLA Chancellor's Animal Research Committee protocol. Establishment of LAPC-9 and PC-3 xenograft SCID mouse models, 124I radioiodination, and determination of labeling efficiency and immunoreactivity were performed as previously described (Olafsen et al., 2007). Initially, eight mice were injected with 122 ± 9 µCi 124I-labeled p2B3-Db8 [specific activity (SA) = 2.6 µCi/μg] and another eight mice were injected with 124 ± 19 µCi bm2B3-Db8 (SA = 2.5 µCi/μg) and imaged at 4 and 20 h. In a second study, mice were imaged at 8 and 12 h. In this study, four mice per group were injected with 129 ± 5 (6.1 µCi/µg) or 118 ± 2 (5.2 µCi/µg) µCi of the p2B3- or bm2B3-Db8, respectively. Mouse positioning, microPET acquisition procedures and biodistribution methods were performed as previously described (Olafsen et al., 2007).

Cylindrical regions of interest (ROI) were drawn from three-dimensional FBP reconstructed PET/CT co-registered images using AMIDE (Loening and Gambhir, 2003). From sagittal images, four ROIs were drawn per tumor and in areas adjacent but completely outside the mouse, and termed ‘scatter’. From the coronal orientation, four ROIs in the low activity areas adjacent to the tumors were drawn in the arm muscle region and termed ‘background’. Calculation and significance was performed as described (Leyton et al., 2008).

Results

Affinity maturation design

Careful inspection of a revised bm2B3 model suggested that the apparent loss of affinity, caused by humanization of 1G8 mAb, was possibly caused by two residues in the VL domain and four residues in the VH domain (Fig. 1A). These residues were selected for back-mutation to their murine counterparts, bringing the total number of retained murine framework residues to 12 in the diabody. Their locations in the revised bm2B3 model and the structural differences compared with the p2B3 model are shown in Fig. 1B. Residues L70 (asp) and L71 (phe) pack against the CDR-L1 loop and were therefore back-mutated (to ser and tyr, respectively) since in crystal structures with nearly identical L1 loops these two framework residues influence the conformation of this loop. Residues H48 (val) and H49 (ala), which precede the CDR-H2 loop (H50–H65), were likewise back-mutated (to ile and gly, respectively). H69 (ile) was back-mutated to methionine, and to make the post-CDR-H2 framework region essentially identical to murine 1G8 and humanized 2B3, H66 (arg) was back-mutated to lysine as well (residues H67 and H68 are already identical in the murine and humanized sequences). The glycine at H49 is particularly important since this residue packs against a framework residue (H69) that follows the CDR-H2 loop. Disruption of this pairing causes major shifts of the CDR-H2 loop, either toward or away from the CDR-L3 loop (as seen in the p2B3 model in which trastuzumab residues ala and ile are present) (Olafsen et al., 2007). A computer-generated image depicting the recognition surface differences between the p2B3 and the bm2B3 is shown in Fig. 1C.

Fig. 1.

Fig. 1

Open in a new tab

(A) Amino acid sequences of the back-mutated VL and VH domains of the humanized 2B3 scFv. (B) Ribbon representation (side and top view) of p2B3 (left) and bm2B3 (right) Fvs. Residues are labeled with Kabat numbers. (C) Solid renderings of the top view models shown in (B). Framework residues are colored cyan, CDRs red, initial back-mutations (p2B3 version) purple and additional back-mutations (bm2B3 version) underlined green or yellow.

Protein size, PSCA binding and affinity

Analysis of purified 2B3 scFvs or diabodies by SDS–PAGE (Fig. 2A) demonstrated that they all migrated with the expected monomeric MW (23–25 kDa). The scFvs (lanes 1 and 2) have a lower mobility than the diabodies (lanes 3–5), consistent with an 18 amino acid linker in the scFvs. Biacore analysis revealed that the p2B3- and bm2B3-Db8 had apparent KDs of 5.4 and 1.9 nM, respectively (data not shown). Binding to native PSCA by indirect flow cytometry on SKW 6.4-PSCA cells demonstrated the effect of affinity (Fig. 2B). The back-mutated scFv and the corresponding diabodies appear to exhibit superior binding to PSCA compared with their parental counterparts. Back-mutations in the scFv fragment produced a 1.9-fold stronger fluorescent shift than the parental scFv [mean fluorescence intensity (MFI) of 863 and 452, respectively]. As for the diabodies, the bm2B3-Db5 exhibited a 3-fold stronger fluorescence shift (MFI of 1562 and 492, respectively) compared with the p2B3-Db5, and the bm2B3-Db8 showed a 1.3-fold strong shift (MFI of 1361 and 1028, respectively) compared with the p2B3-Db8. These results suggest that the six back-mutated amino acids are involved in the cellular PSCA binding ability and that restoring these to the murine counterparts strengthens the interaction.

Fig. 2.

Fig. 2

Open in a new tab

(A) SDS–PAGE analysis. Lane 1, p2B3 scFv; lane 2, bm2B3 scFv; lane 3, p2B3-Db8; lane 4, bm2B3-Db8; lane 5, p2B3-Db5 and lane 6, bm2B3-Db5. (B) Indirect flow cytometry to study the effect of back-mutations on binding to PSCA-SKW 6.4 cells. For the negative control, only mouse anti-His6-tag antibody and PE-conjugated goat anti-mouse (Fc-specific) antibodies were used.

Stability of diabody species

It has been shown that the monomer/dimer equilibrium of scFv/diabodies can be strongly influenced by the protein concentration (Wu et al., 1996; Arndt et al., 1998). In order to assess the stability of the dimeric form of 2B3 diabodies, size exclusion chromatography was performed. Purified p2B3- and bm2B3-Db8 at a concentration of 1 mg/ml eluted as two species corresponding to scFv monomers and dimers (Fig. 3A). At higher concentrations of 5.7 and 3.6 mg/ml for p2B3- and bm2B3-Db8, respectively, both diabodies eluted as dimers (Fig. 3B). However, the bm2B3-Db8 still showed a slight shoulder corresponding to a monomer suggesting that factors other than the concentration may affect scFv dimerization. To further assess the extent of dimer instability at low concentrations over time, the effect of linker length was evaluated. All four diabodies were loaded at a high concentration (5.0 mg/ml of p2B3-Db5, 4.7 mg/ml of bm2B3-Db5, 4.4 mg/ml of p2B3-Db8 and 4.8 mg/ml of bm2B3-Db8) onto a 10 ml bed-volume Superdex 75 column, where it is approximated that a 10- to 20-fold dilution takes place (Wu et al., 1996). Fractions collected across the dimer and monomer mix were stored at 4°C and analyzed again at different time points. Integration over individual peaks at differing times demonstrated that both p2B3-Dbs were predominantly dimers regardless of the 10- to 20-fold dilution or time (Fig. 3C). The bm2B3-Db8 was the protein that displayed the greatest re-equilibrium between monomer and dimer species at all the times tested, and only when re-concentrated did the bm2B3-Db8 shift to a dimer. As for the bm2B3-Db5, it displayed a smaller shift at all time points suggesting that a shorter linker is necessary to shift the equilibrium to a diabody, but not sufficient to trap it as a dimer.

Fig. 3.

Fig. 3

Open in a new tab

Size exclusion chromatography of nickel purified anti-PSCA p2B3-Db8 and bm2B3-Db8. Elution profiles from a Superdex 75 gel filtration column of (A) p2B3-Db8 at 1 mg/ml (left) and bm2B3-Db8 at 1 mg/ml (right); (B) p2B3-Db8 at 5.7 mg/ml (left) and bm2B3-Db8 at 3.6 mg/ml (right). (C) Analysis of monomer/dimer equilibrium of purified five- and eight-amino acid linker 2B3 diabodies at various times (24, 48 and 168 h) at low concentrations and after being re-concentrated to high concentrations.

MicroPET imaging of LAPC-9 xenografts with p2B3- and bm2B3-Db8

To test whether the 2B3 diabody would function as an effective PET radiotracer for producing high contrast images at an early time point, 124I-labeled p2B3-Db8 was injected via the tail vein into SCID mice bearing PSCA-positive LAPC-9 xenografts (Fig. 4A). At 4 h, low activity was observed in the PSCA-positive xenograft. Higher contrast was achieved at 12 h, in agreement with the ROI tumor-to-background ratio analysis (Fig. 4B). Biodistribution studies, however, revealed that the highest tumor-to-blood ratio (0.88) was at 8 h (Fig. 4C). To confirm that the activity in the tumor was specific, a double tumor model of PSCA-negative (PC-3; 300 ± 25 mg) and PSCA-positive (LAPC-9; 770 ± 200 mg) xenografts was established and subsequent targeting with 124I-labeled p2B3-Db8 confirmed antigen-dependent recognition (Table I, Fig. 4A).

Fig. 4.

Fig. 4

Open in a new tab

(A) Coronal microPET images of LAPC-9 tumor-bearing SCID mice at 4, 12 and 20 h after administration of 124I-p2B3-Db8. MicroPET/microCT coregistration images are shown at 12 and 20 h. LAPC-9 tumor (+ arrow); PC-3 tumor (- arrow). (B) Tumor-to-blood ratios at different time points (4, 8, 12 and 20 h) of p2B3-Db8 (black) and bm2B3-Db8 (gray) determined by ROI analysis of the images. (C) Tumor-to-blood ratios at different time points (8, 12 and 20 h) for p2B3-Db8 (filled diamond) and bm2B3-Db8 (filled square) determined by ex vivo weighing and counting tumor and blood in γ-counter.

Table I.

Biodistribution of 124I-labeled anti-PSCA diabodies at 20 h in mice (n = 12)

Tissue p2B3-Db8 bm2B3-Db8
LAPC-9 (+) tumor (T) 0.57 ± 0.2 1.02 ± 0.2
PC-3 (−) tumor 0.34 ± 0.2 0.51 ± 0.1
Liver 0.27 ± 0.1 0.60 ± 0.1
Spleen 0.33 ± 0.2 0.55 ± 0.2
Kidney 0.67 ± 0.2 0.77 ± 0.3
Lung 0.41 ± 0.1 0.93 ± 0.3
Blood 0.86 ± 0.2 0.79 ± 0.1
Carcass 0.33 ± 0.2 ND
Ratios
 T/Blood 0.7 1.2
 T/negative tumor 1.7 2.0
Open in a new tab

Tumor and normal organ uptakes are expressed as percent injected dose per gram (%ID/g ± SD). ND, not determined.

The bm2B3-Db8 was also evaluated by microPET in LAPC-9 xenograft-bearing mice in order to evaluate the effect of higher affinity on tumor targeting. The tumor uptake of bm2B3-Db8 was higher than that of the p2B3-Db8 at all time points, and at 20 h it had a higher tumor-to-blood ratio (Fig. 4B). In addition, the target-to-background ratio was better at 20 h, however, targeting at 4, 8 and 12 h showed no beneficial effect. This suggests that higher affinity only improves tumor retention at 20 h, leading to higher tumor-to-blood and tumor-to-background ratios (Fig. 4B and C). In addition, the PSCA-positive-to-PSCA-negative tumor ratio was also improved with the higher affinity bm2B3-Db8. At 20 h, the uptake in the tumors was significantly different (P < 0.05) and the tumor-to-blood ratio was ∼2-fold higher compared with the p2B3-Db8. The comparison of biodistributions for both p2B3- and bm2B3-Db8 at 20 h is summarized in Table I.

Discussion

In the present work, we have generated two versions of PSCA-specific engineered humanized 2B3 diabodies that differ in linker length, with one version containing a five amino acid linker (p2B3-Db5) and the other an eight amino acid linker (p2B3-Db8). Each of these was affinity matured by back-mutation of six residues, identified by computer modeling analysis, resulting in the bm2B3-Db5 and bm2B3-Db8 diabodies. SPR analysis demonstrated that these six mutations conferred a ∼2.8-fold increase in affinity in bm2B3-Db8. Stability of the dimeric fractions of p2B3-Db8 and bm2B3-Db8 was concentration-dependent. In addition, the six residues seemed to affect dimer equilibrium independent of concentration. Finally, shortening the linker length from eight amino acids to five appeared to alleviate the equilibrium shift. Both p2B3-Db8 and bm2B3-Db8 were evaluated by microPET imaging following radioiodination with 124I. The p2B3-Db8 exhibited specific targeting of PSCA-positive human prostate cancer xenografts in mice as early as 4 h and provided an improved contrast tumor image at 12 h. Of interest, imaging using the higher affinity bm2B3-Db8 did not improve the contrast at the time points tested before 20 h.

In the initial humanization work of 1G8, two versions (A and B) of the humanized 2B3 mAb were made. When these were purified and assayed for antigen-binding activity, it became clear that residues H60–H65 (at the end of the CDR-H2) play a key role (Olafsen et al., 2007). For this reason, we decided to examine a molecular model of 2B3 to determine whether back-mutating any of the framework residues that pack against H60–H65 might restore the ∼5-fold loss of binding affinity. We also examined the residues that pack against CDR-L1 since this loop is highly exposed. The introduction of six back-mutations improved the apparent affinity of the bm2B3-Db8 (1.9 nM) approaching that of the murine 1G8 mAb (1.47 nM), but this had no effect on the ability to provide higher contrast images at 4, 8 or 12 h compared with p2B3-Db8. Although high affinity is a critical component of a good imaging antibody fragment, it has been shown that monoclonal antibodies with very high affinity may impair efficient tumor penetration and diminish effective therapeutic action (Adams et al., 2001). Conversely, with the lowest molecular weight fragment containing the complete antigen-binding site, the scFv has also demonstrated intrinsic affinity properties that regulate the quantitative delivery and penetration to solid tumors (Adams et al., 2001). Biodistribution studies with scFvs against HER2-expressing SK-OV-3 xenografts demonstrated that quantitative tumor retention did not increase with enhancements in affinity beyond 10−9 M.

The ability of a diabody to dimerize appears to be dependent on a multitude of factors, such as linker length, antibody sequence and external factors (Arndt et al., 1998). The notion that scFv dimers in the VH–VL orientation can be formed with linker lengths from 3 to 12 residues and that linkers below three residues can force scFv association into trimers was established with NC10 (anti-neuraminidase) scFv fragments (Atwell et al., 1999; Kortt et al., 2001). This transition was distinct, as opposed to scFv fragments in the VL–VH orientation, where reducing the linker from three to two residues produced a mixture of dimers and trimers with dimers being the predominant species. In the absence of a linker, a trimer/tetramer equilibrium mixture was the preferred oligomeric conformation. However, this is not absolute as stable dimers have been formed with VL–VH orientated scFv with zero-residue linker (Arndt et al., 2004). The precise transition from dimer to trimer of VH–VL orientated scFv is also not absolute. Dimer/trimer and trimer/tetramer to exclusive tetramer conformations have been observed for different antibody sequences (Le Gall et al., 1999, 2004; Arndt et al., 2004). It has been noted that linkers of the same length between the VL C-terminus and the VH N-terminus are more strained since the distance is larger than the one between the VH C-terminus and the VL N-terminus (Huston et al., 1991). Although the scFv fragments generated in this study were in the more strained VL–VH orientation, stable dimers were not formed using an eight-residue linker. Shortening the linker to five residues, however, improved dimer stability. This observation differs from the anti-CEA T84.66 and anti-CD20 2B8 VL–VH orientated diabodies, which form stable dimers with an eight-residue linker (Wu et al., 1999). In addition, there are reports of stable scFv dimers (VH–VL) with linkers longer than eight residues (Kortt et al., 1997; Asvadi et al., 2002). Thus, linker-dependent dimer/oligomerization behavior of scFv fragments appears highly dependent on the specific antibody selected and the orientation of the variable domains.

On the basis of our previous experience with the anti-CEA and anti-CD20 diabodies, p2B3- and bm2B3-Db8 were initially produced and purified from mammalian cells. When it was apparent that these did not produce stable dimer, it was decided to investigate the effect of a shorter linker. In order to be able to quickly analyze the proteins, both p2B3- and bm2B3-Db5 as well as the scFvs were expressed in E. coli as soluble proteins that were secreted into periplasmic space. However, a direct comparison of the full set of constructs in bacterial versus mammalian expression systems has not been undertaken.

Another possible issue with the 2B3 diabody may be that VL and VH have a high degree of thermal instability, i.e. low association constant of the VH–VL interface. Transient opening of the VL and VH interface will expose hydrophobic patches that favor aggregation with itself or other molecules present in the serum (Arndt et al., 1998; Willuda et al., 1999). However, since the 2B3 diabodies express well and do not form higher molecular weight aggregates (features expected if thermal instability was an issue), the dimer/monomer equilibrium phenomenon may also be attributed to other factors. In a study using the scFv McPC603 as a model system, it was found that the presence of antigen, high ionic strength and pH below 7.5 stabilized the VH–VL interface of the dimer (Arndt et al., 1998). As for the 2B3 diabody, it is possible that the back-mutated residues contributed to the observed instability. Indeed at 168 h, more than 20% of the bm2B3-Db5 is in the monomeric form, suggesting that the back-mutated residues may affect the stability of the dimer. Although the back-mutated residues are not directly located in the VH–VL interface, they may induce a conformational shift of the framework which affects dimerization. Thus, the longer linker combined with the back-mutations made may explain the overall lower stability of the bm2B3-Db8.

In order to maintain the desired size and valency of the diabody, it is important to produce stable scFv dimers since monomers will affect the pharmacokinetics and tumor targeting in vivo. An example of poor tumor targeting due to instability is the 15–9 diabody with a five-residue linker in the VH–VL orientation (Huang et al., 2005). The diabody showed no binding to antigen and did not accumulate in tumors as well as its scFv counterpart. The authors suggested that this lack of activity was probably due to the short linker which sterically prevented the formation of an active scFv monomer with one intact antigen-binding site.

Importantly, in this work, we show that the 2B3-Db8 diabody and its affinity matured counterpart, bm2B3-Db8, could be employed effectively as radiolabeled imaging agents. Suboptimal stability of the bm2B3 at long storage times (7 days) may not be as much a concern in the in vivo imaging setting, since the diabody format was selected for its rapid in vivo kinetics, allowing same day imaging. However, the accumulation of radiolabeled p2B3- and bm2B3-Db8 in the tumor was not very high; 0.6 and 1% ID/g at 20 h, respectively, resulting in low tumor:blood ratios of 0.7:1 and 1.3:1. In other diabodies, tumor:blood ratios of 30–80:1 and % ID/g levels of 4.7–13.7 in tumors have been reported (Adams et al., 1998; Pavlinkova et al., 1999; Wu et al., 1999; Power et al., 2001). The relatively low uptakes may be due to dissociation of the 2B3 scFv dimer in the serum, or perhaps due to internalization and metabolism of the radioiodinated protein, to a higher extent than seen for the 2B3 minibody (Leyton et al., 2008).

PSCA is a promising biomarker that can play a role in several aspects of prostate cancer management, and its utility in in vitro diagnostics has been demonstrated. The 2B3 Db extends the utility of PSCA-specific antibody reagents to non-invasive imaging in vivo, and may offer an approach for improving the detection of prostate cancer. However, current work suggests that the instability of this scFv dimer with an eight-residue linker may be a limiting factor for imaging applications. Since the Db5 constructs appear to be an improvement over Db8 constructs, i.e. more stable as uniform dimers and with equivalent antigen-binding affinity, they may localize better to tumors. Evaluation of their in vivo targeting would be useful to rule out dissociation and/or internalization and metabolism of this fragment. Additional approaches to stabilize this fragment such as reversing the orientation of the variable domains should be investigated.

Funding

This work was supported by the UCLA SPORE in Prostate Cancer (National Institutes of Health CA92131), the UCLA ICMIC (In vivo Centers for Molecular Imaging of Cancer) (National Institutes of Health CA86306) and Program Project Grant (National Institutes of Health CA43904). R.E.R. and A.M.W. are members of the UCLA Jonsson Comprehensive Cancer Center (National Institutes of Health CA16042).

Acknowledgements

We would like to thank Dr Vania E. Kenanova for inspiring scientific discussions. The authors would also like to thank Dr David Stout, Waldemar Ladno and Judy Edwards for assistance with microPET/CT scans and Felix B. Salazar for excellent technical assistance. We also acknowledge the assistance of Dr Michael Phillips at the UCLA-DOE Biochemistry Instrumentation Facility and Tammy Phung at the UCLA Jonsson Comprehensive Cancer Center Flow Cytometry Facility.

Footnotes

Edited by Peter Hudson

References

  1. Adams G.P., Schier R., McCall A.M., Crawford R.S., Wolf E.J., Weiner L.M., Marks J.D. Br. J. Cancer. 1998;77:1405–1412. doi: 10.1038/bjc.1998.233. [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Adams G.P., Schier R., McCall A.M., Simmons H.H., Horak E.M., Alpaugh R.K., Marks J.D., Weiner L.M. Cancer Res. 2001;61:4750–4755. [PubMed] [Google Scholar]
  3. Adams G.P., Tai M.S., McCartney J.E., Marks J.D., Stafford W.F., 3rd, Houston L.L., Huston J.S., Weiner L.M. Clin. Cancer Res. 2006;12:1599–1605. doi: 10.1158/1078-0432.CCR-05-2217. [DOI] [PubMed] [Google Scholar]
  4. Arndt K.M., Muller K.M., Pluckthun A. Biochemistry. 1998;37:12918–12926. doi: 10.1021/bi9810407. [DOI] [PubMed] [Google Scholar]
  5. Arndt M.A., Krauss J., Rybak S.M. FEBS Lett. 2004;578:257–261. doi: 10.1016/j.febslet.2004.11.011. [DOI] [PubMed] [Google Scholar]
  6. Asvadi P., Fletcher A., Raison R.L. J. Mol. Recognit. 2002;15:321–330. doi: 10.1002/jmr.594. [DOI] [PubMed] [Google Scholar]
  7. Atwell J.L., Breheney K.A., Lawrence L.J., McCoy A.J., Kortt A.A., Hudson P.J. Protein Eng. 1999;12:597–604. doi: 10.1093/protein/12.7.597. [DOI] [PubMed] [Google Scholar]
  8. Bell-Pedersen D., Dunlap J.C., Loros J.J. Mol. Cell Biol. 1996;16:513–521. doi: 10.1128/mcb.16.2.513. [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Bouchelouche K., Oehr P. Curr. Opin. Oncol. 2008;a 20:321–326. doi: 10.1097/CCO.0b013e3282f8b02b. [DOI] [PubMed] [Google Scholar]
  10. Bouchelouche K., Oehr P. J. Urol. 2008;b 179:34–45. doi: 10.1016/j.juro.2007.08.176. [DOI] [PubMed] [Google Scholar]
  11. Cai W., Olafsen T., Zhang X., Cao Q., Gambhir S.S., Williams L.E., Wu A.M., Chen X. J. Nucl. Med. 2007;48:304–310. [PubMed] [Google Scholar]
  12. Carter T., Sterling-Levis K., Ow K., Doughty L., Hattarki M., Shapira D., Hewish D., Kortt A.A., Russell P.J. Cancer Immunol. Immunother. 2004;53:533–542. doi: 10.1007/s00262-003-0460-1. [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Dolezal O., Pearce L.A., Lawrence L.J., McCoy A.J., Hudson P.J., Kortt A.A. Protein Eng. 2000;13:565–574. doi: 10.1093/protein/13.8.565. [DOI] [PubMed] [Google Scholar]
  14. Gu Z., Thomas G., Yamashiro J., Shintaku I.P., Dorey F., Raitano A., Witte O.N., Said J.W., Loda M., Reiter R.E. Oncogene. 2000;19:1288–1296. doi: 10.1038/sj.onc.1203426. [DOI] [PubMed] [Google Scholar]
  15. Hara N., Kasahara T., Kawasaki T., Bilim V., Obara K., Takahashi K., Tomita Y. Clin. Cancer Res. 2002;8:1794–1799. [PubMed] [Google Scholar]
  16. Holliger P., Prospero T., Winter G. Proc. Natl Acad. Sci. USA. 1993;90:6444–6448. doi: 10.1073/pnas.90.14.6444. [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Hricak H., Choyke P.L., Eberhardt S.C., Leibel S.A., Scardino P.T. Radiology. 2007;243:28–53. doi: 10.1148/radiol.2431030580. [DOI] [PubMed] [Google Scholar]
  18. Huang B.C., Foote L.J., Lankford T.K., Davern S.M., McKeown C.K., Kennel S.J. Biochem. Biophys. Res. Commun. 2005;327:999–1005. doi: 10.1016/j.bbrc.2004.12.114. [DOI] [PubMed] [Google Scholar]
  19. Huston J.S., Mudgett-Hunter M., Tai M.S., McCartney J., Warren F., Haber E., Oppermann H. Methods Enzymol. 1991;203:46–88. doi: 10.1016/0076-6879(91)03005-2. [DOI] [PubMed] [Google Scholar]
  20. Kelly M.P., Lee F.T., Tahtis K., Power B.E., Smyth F.E., Brechbiel M.W., Hudson P.J., Scott A.M. Cancer Biother. Radiopharm. 2008;23:411–423. doi: 10.1089/cbr.2007.0450. [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Kortt A.A., et al. Protein Eng. 1997;10:423–433. doi: 10.1093/protein/10.4.423. [DOI] [PubMed] [Google Scholar]
  22. Kortt A.A., Dolezal O., Power B.E., Hudson P.J. Biomol. Eng. 2001;18:95–108. doi: 10.1016/s1389-0344(01)00090-9. [DOI] [PubMed] [Google Scholar]
  23. Lam J.S., Yamashiro J., Shintaku I.P., Vessella R.L., Jenkins R.B., Horvath S., Said J.W., Reiter R.E. Clin. Cancer Res. 2005;11:2591–2596. doi: 10.1158/1078-0432.CCR-04-1842. [DOI] [PubMed] [Google Scholar]
  24. Le Gall F., Kipriyanov S.M., Moldenhauer G., Little M. FEBS Lett. 1999;453:164–168. doi: 10.1016/s0014-5793(99)00713-9. [DOI] [PubMed] [Google Scholar]
  25. Le Gall F., Reusch U., Moldenhauer G., Little M., Kipriyanov S.M. J. Immunol. Methods. 2004;285:111–127. doi: 10.1016/j.jim.2003.11.007. [DOI] [PubMed] [Google Scholar]
  26. Leyton J.V., Olafsen T., Lepin E.J., Hahm S., Bauer K.B., Reiter R.E., Wu A.M. Clin. Cancer Res. 2008 doi: 10.1158/1078-0432.CCR-07-5093. in press. [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Loening A.M., Gambhir S.S. Mol. Imaging. 2003;2:131–137. doi: 10.1162/15353500200303133. [DOI] [PubMed] [Google Scholar]
  28. Olafsen T., Gu Z., Sherman M.A., Leyton J.V., Witkosky M.E., Shively J.E., Raubitschek A.A., Morrison S.L., Wu A.M., Reiter R.E. J. Immunother. 2007;30:396–405. doi: 10.1097/CJI.0b013e318031b53b. [DOI] [PubMed] [Google Scholar]
  29. Parry R., et al. Cancer Res. 2005;65:8397–8405. doi: 10.1158/0008-5472.CAN-05-1203. [DOI] [PubMed] [Google Scholar]
  30. Pavlinkova G., Beresford G.W., Booth B.J., Batra S.K., Colcher D. J. Nucl. Med. 1999;40:1536–1546. [PubMed] [Google Scholar]
  31. Perisic O., Webb P.A., Holliger P., Winter G., Williams R.L. Structure. 1994;2:1217–1226. doi: 10.1016/s0969-2126(94)00123-5. [DOI] [PubMed] [Google Scholar]
  32. Power B.E., et al. Cancer Immunol. Immunother. 2001;50:241–250. doi: 10.1007/s002620100192. [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Robinson M.K., Doss M., Shaller C., Narayanan D., Marks J.D., Adler L.P., Gonzalez Trotter D.E., Adams G.P. Cancer Res. 2005;65:1471–1478. doi: 10.1158/0008-5472.CAN-04-2008. [DOI] [PubMed] [Google Scholar]
  34. Saffran D.C., Raitano A.B., Hubert R.S., Witte O.N., Reiter R.E., Jakobovits A. Proc. Natl Acad. Sci. USA. 2001;98:2658–2663. doi: 10.1073/pnas.051624698. [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Smith-Jones P.M., Vallabhajosula S., Navarro V., Bastidas D., Goldsmith S.J., Bander N.H. J. Nucl. Med. 2003;44:610–617. [PubMed] [Google Scholar]
  36. Sundaresan G., Yazaki P.J., Shively J.E., Finn R.D., Larson S.M., Raubitschek A.A., Williams L.E., Chatziioannou A.F., Gambhir S.S., Wu A.M. J. Nucl. Med. 2003;44:1962–1969. [PMC free article] [PubMed] [Google Scholar]
  37. Tran C.P., Lin C., Yamashiro J., Reiter R.E. Mol. Cancer Res. 2002;1:113–121. [PubMed] [Google Scholar]
  38. Williams L.E., Wu A.M., Yazaki P.J., Liu A., Raubitschek A.A., Shively J.E., Wong J.Y. Cancer Biother. Radiopharm. 2001;16:25–35. doi: 10.1089/108497801750095989. [DOI] [PubMed] [Google Scholar]
  39. Willuda J., Honegger A., Waibel R., Schubiger P.A., Stahel R., Zangemeister-Wittke U., Pluckthun A. Cancer Res. 1999;59:5758–5767. [PubMed] [Google Scholar]
  40. Wu A.M., Chen W., Raubitschek A., Williams L.E., Neumaier M., Fischer R., Hu S.Z., Odom-Maryon T., Wong J.Y., Shively J.E. Immunotechnology. 1996;2:21–36. doi: 10.1016/1380-2933(95)00027-5. [DOI] [PubMed] [Google Scholar]
  41. Wu A.M., Williams L.E., Zieran L., Padma A., Sherman M., Bebb G.G., Odom-Maryon T., Wong J.Y.C., Shively J.E., Raubitschek A.A. Tumor Targeting. 1999;4:47–58. [Google Scholar]

Articles from Protein Engineering, Design and Selection are provided here courtesy of Oxford University Press

RESOURCES