Purpose: More than two decades of research and clinical trials have shown radioimmunotherapy to be a promising approach for treating various forms of cancer. Lym-1 antibody, which binds selectively to HLA-DR10 on malignant B-cell lymphocytes, has proved to be effective in delivering radionuclides to non–Hodgkin's lymphoma and leukemia. Using a new approach to create small synthetic molecules that mimic the targeting properties of the Lym-1 antibody, a prototype, selective high-affinity ligand (SHAL), has been developed to bind to a unique region located within the Lym-1 epitope on HLA-DR10.

Experimental Design: Computer docking methods were used to predict two sets of small molecules that bind to neighboring cavities on the β subunit of HLA-DR10 surrounding a critical amino acid in the epitope, and the ligands were confirmed to bind to the protein by nuclear magnetic resonance spectroscopy. Pairs of these molecules were then chemically linked together to produce a series of bidentate and bisbidentate SHALs.

Results: These SHALs bind with nanomolar to picomolar Kd's only to cell lines expressing HLA-DR10. Analyses of biopsy sections obtained from patients also confirmed that SHAL bound to both small and large cell non–Hodgkin's lymphomas mimicking the selectivity of Lym-1.

Conclusions: These results show that synthetic molecules less than 1/50th the mass of an antibody can be designed to exhibit strong binding to subtle structural features on cell surface proteins similar to those recognized by antibodies. This approach offers great potential for developing small molecule therapeutics that target other types of cancer and disease.

Although surgery and conventional radiation therapy are effective for treating localized cancer, they fall short when faced with disease that is spread throughout the body. Conventional chemotherapy has been moderately successful in treating metastatic disease, but the differential between drug concentrations delivered to tumor and to normal tissues is small. High-specificity, high-affinity cancer-targeting agents that can deliver one or more tumoricidal agents directly to the cancer cells are needed.

There has been considerable optimism that a key component of the body's own disease-targeting system, antibodies, would provide such a vehicle. By tagging monoclonal antibodies with radioactive isotopes, clinicians hoped to bring radiation with its well-known cytocidal properties directly to the tumor. Over the years, significant progress has been made in identifying target molecules unique to cancer cells, raising monoclonal antibodies and determining the radiation dose distribution in the patient before and during treatment. These advances have led to substantial improvements in treatments for non–Hodgkin' s lymphoma (14). Phase III trials have shown a significant (80% vs. 56% overall response rate) increase in response rate with radioimmunotherapy over immunotherapy or conventional chemotherapy (5). In addition, measurable clinical responses have been achieved with almost every malignancy tested, including breast cancer, prostate cancer, brain malignancies, leukemia, and others (59). However, for radioimmunotherapy to cure patients without causing major complications, molecular-targeting agents that exhibit substantially higher tumor specificity and retention than has been achieved with existing antibodies and peptide chains are needed.

Relevant to this need, new methods have been developed that enable the design of small multidentate ligands that bind to the surfaces of proteins with both high affinity and specificity (1014). The mode of binding in these molecules mimics the process used by antibodies to establish specific high-affinity interactions with its target through the formation of multiple contacts between amino acid residues on the surface of the antibody and corresponding sites located on the surface of its target (15). Efforts pioneered by Fesik (10) and others (1114) have shown that two or more small molecules that normally bind to different sites on the surface of a protein with low-to-moderate affinities (substrates, cofactors, natural ligands) can be linked together to produce multidentate ligands with three to six orders higher affinity than the original components. Using this process, it is now possible to consider developing an entirely new class of inhibitors, targeting agents and diagnostics that bind to any two adjacent structurally unique sites on the surface of a protein.

Here, we describe a prototype selective high-affinity ligand (SHAL) that has been created by linking together a pair of small molecules that bind independently with low affinity to two structurally unique but nonfunctional cavities located on the surface of HLA-DR10. This HLA-DR was selected as the most suitable target for the first SHAL being developed using this approach because it is present on almost all malignant B cells (16), its density on malignant cells is at least 10-fold higher than on normal B cells, and the protein contains a unique epitope recognized by the antibody Lym-1 that has been well characterized (17, 18). Because Lym-1 has been shown to be effective in treating non–Hodgkin's lymphoma, the development of a small molecule SHAL targeting the same epitope recognized by Lym-1 will allow a direct comparison of the use of SHAL relative to a known therapeutic. Because all proteins have numerous cavities distributed across their surfaces that can be targeted in a similar manner for small molecule binding, this work provides an example of how SHALs can be designed to bind to subtle features on the surfaces of proteins with a much greater specificity than we previously thought possible. Results achieved by tethering pairs of these molecules together also show that affinities can be obtained that are similar to or better than those observed when an antibody binds bivalently to its target.

Homology modeling and in silico ligand screening. The homology model of HLA-DR10 was generated using the atomic coordinates obtained from different segments of four known HLA-DR crystal structures (1922). The approach used to generate the model was similar to that used earlier to successfully model both high and low homology target proteins (45). The particular segments of the HLA-DR crystal structures used in the model were selected based on similarities in their secondary structural elements. The solvent accessible surfaces of HLA-DR10 and HLA-DR1-4 were calculated, and two adjacent binding sites (concave pockets) on the modeled HLA-DR10 surface were selected as appropriate sites for ligand binding. These sites, identified as site 1 and site 2 (Fig. 1B), flank both sides of the most important amino acid in the Lym-1 epitope, R70.

Fig. 1.

Structure of the HLA-DR10 β subunit and ligand binding sites. A, homology model of the β subunit. The red stick-form residues are the amino acids (Arg70, Arg71, Ala74, Val85) that affect Lym-1 binding to the protein. The yellow stick-form cysteine residues form the disulfide bond in the subunit that must be present for Lym-1 recognition of HLA-DR10. B, a solvent accessible surface representation of HLA-DR10. The blue residue is Arg70, an important residue in the Lym-1 epitope. The magenta area is binding site 1, and the cyan area is binding site 2. C, the solvent accessible surface representation of HLA-DR3 (left) and the solvent accessible surface representation of HLA-DR10 (right). Residues are color coded by charge: blue, positively charged; red, negatively charge; green, polar and not charged; white, hydrophobic. Note that the site 1 binding cavity is filled in for HLA-DR3, whereas it is a well-defined cavity in HLA-DR10.

Fig. 1.

Structure of the HLA-DR10 β subunit and ligand binding sites. A, homology model of the β subunit. The red stick-form residues are the amino acids (Arg70, Arg71, Ala74, Val85) that affect Lym-1 binding to the protein. The yellow stick-form cysteine residues form the disulfide bond in the subunit that must be present for Lym-1 recognition of HLA-DR10. B, a solvent accessible surface representation of HLA-DR10. The blue residue is Arg70, an important residue in the Lym-1 epitope. The magenta area is binding site 1, and the cyan area is binding site 2. C, the solvent accessible surface representation of HLA-DR3 (left) and the solvent accessible surface representation of HLA-DR10 (right). Residues are color coded by charge: blue, positively charged; red, negatively charge; green, polar and not charged; white, hydrophobic. Note that the site 1 binding cavity is filled in for HLA-DR3, whereas it is a well-defined cavity in HLA-DR10.

Close modal

Molecular docking was used to computationally screen the MDL Available Chemical Directory of small molecules (∼200,000 compounds at the time the ligands were screened) to identify the top ranked molecules predicted to bind to site 1 and site 2. All compound conformations were evaluated by an energy score. While the molecules were ranked based on their energies, the scoring function does not predict the actual binding affinities. The top 1% of scored compounds (∼2,500) was examined visually. Specific interactions, such as charge and hydrophobic interactions, were qualitatively noted for each compound. A variety of molecules representing the spectrum of available compounds were selected for experimental analysis.

Isolation and purification of HLA-DR10 and its β subunit. The HLA-DR10 protein, which is a heterodimer composed of α and β subunits with a total apparent molecular weight of 55 kDa, was isolated from cultured Raji Burkitt's lymphoma cells and purified on a Lym-1 antibody affinity column as described previously (42). The β subunit of HLA-DR10 was expressed in Escherichia coli. The protein was isolated as inclusion bodies, dissolved, and refolded to yield protein that remained soluble in aqueous buffers in the absence of a detergent. Circular dichroism spectra and small molecule ligand binding were used to confirm that the recombinant β subunit was properly folded.

Experimental confirmation of ligand binding to HLA-DR10. Transferred NOE spectroscopy experiments (2327) and water LOGSy experiments (28) were carried at 30°C, using mixtures of several ligands per experiment, to identify those ligands that bind to native HLA-DR10. The proteins used in the nuclear magnetic resonance screening experiments ranged in concentration from 0.7 to 1.8 mg/mL. Small volumes of ligand stock solution in D2O or d6-DMSO were added sequentially to an aqueous HLA-DR10 solution (containing 10-50% D2O for the lock signal) so that the molar ratio of each ligand/protein was ∼20:1. The total concentration of d6-DMSO never exceeded 3% (v/v).

SHAL synthesis. First generation SHALs containing deoxycholate and leu-enkephalin (SHAL1-SHAL3) were manually synthesized using standard Fmoc solid phase synthesis on chlorotritylchloride resin in a 5-mL disposable polyethylene column (Pierce Biotechnology, Inc.). After the coupling of Fmoc-Lys (t-butoxycarbonyl)-OH, Fmoc-mini-polyethylene glycol (PEG)-OH, and the addition of the second lysine (Dde-Lys(Fmoc)-OH) to the resin, the Fmoc protecting group on the ε-amine of the second lysine residue was removed with 20% piperidine. The leu-enkephalin was then synthesized on the resulting resin by sequential addition of Fmoc-Ile-OH, Fmoc-Phe-OH, Fmoc-Ala-OH, Fmoc-Ala-OH, and Fmoc-Tyr-OH. The resin was submitted to a washing and filtration cycle with DMF (2 mL, thrice, 1 min) between every deprotection and coupling step. After acetylating the tyrosine residue on leu-enkephalin with acetic anhydride, deoxycholate was added to the α-amine of the second lysine using NHS ester chemistry after removing the (1-(4,4-dimethyl-2,6-dioxocyclohexylidene)ethyl) Dde-protecting group with 4% hydrazine. In those SHALs with linkers containing one or two mini-PEGs, one or two cycles of Fmoc-mini-PEG-OH were added before deoxycholate addition. The final resin was then treated with 20% trifluoroacetic and 1% triethylsilane in dichloromethane. The trifluoroacetic acid esters formed on the primary alcohols of deoxycholate during cleavage from the resin were removed by stirring in ammonium bicarbonate. The SHALs were purified using semipreparative high-pressure liquid chromatography (HPLC), and the product fractions were lyophilized to dryness. Analytic HPLC trace electrospray ionization-mass spectrometry and 1H–nuclear magnetic resonance spectra were obtained for each SHAL to assess their purity. The nuclear magnetic resonance spectra obtained for each purified SHAL (data not shown) was consistent with the intended structure. The yields of purified product (SHAL1-SHAL3) and observed masses are shown in Table 1.

Table 1.

Yields and molecular masses obtained by electrospray ionization-mass spectrometry for each of the HPLC purified SHALs

SHALYield (mg)(M + 1H)1+ observed (expected)(M + 2H)2+ observed (expected)(M + 3H)3+ observed (expected)(M + 4H)4+ observed (expected)(M + 5H)5+ observed (expected)(M + 6H)6+ observed (expected)
17.0 1601.14 (1601.05) 800.67 (801.03)     
11.0 1746.06 (1746.21) 873.26 (873.61)     
17.0 1891.09 (1891.36) 945.87 (946.19) 630.69 (631.13)    
6.5 1718.20 (1717.99) 859.46 (859.49) 573.29 (573.33)    
3.5  1660.92 (1660.93) 1107.62 (1107.60) 830.99 (830.97) 664.95 (664.58) 554.26 (553.98) 
8.0  1805.78 (1806.09) 1204.41 (1204.40) 903.61 (903.55) 723.04 (723.04) 602.64 (602.70) 
4.5   1301.19 (1301.17) 976.19 (976.13) 781.12 (781.10) 651.03 (651.09) 
SHALYield (mg)(M + 1H)1+ observed (expected)(M + 2H)2+ observed (expected)(M + 3H)3+ observed (expected)(M + 4H)4+ observed (expected)(M + 5H)5+ observed (expected)(M + 6H)6+ observed (expected)
17.0 1601.14 (1601.05) 800.67 (801.03)     
11.0 1746.06 (1746.21) 873.26 (873.61)     
17.0 1891.09 (1891.36) 945.87 (946.19) 630.69 (631.13)    
6.5 1718.20 (1717.99) 859.46 (859.49) 573.29 (573.33)    
3.5  1660.92 (1660.93) 1107.62 (1107.60) 830.99 (830.97) 664.95 (664.58) 554.26 (553.98) 
8.0  1805.78 (1806.09) 1204.41 (1204.40) 903.61 (903.55) 723.04 (723.04) 602.64 (602.70) 
4.5   1301.19 (1301.17) 976.19 (976.13) 781.12 (781.10) 651.03 (651.09) 

Second generation SHALs containing N-benzoyl-l-arginyl-4-amino-benzoic acid and dabsyl-l-valine (SHAL4-SHAL7) were also synthesized on a chlorotritylchloride resin and cleaved using similar methods. The bisbidentate SHALs were created using the same chemistry as the bidentate SHALs. SHAL4 molecules were synthesized at the ends of a pair of mini-PEG chains (two to four PEG units, depending on the desired linker length) conjugated through the α-amine and ε-amine of the N-terminal lysine in lysyl-lysine. 1,4,7,10-Tetra-azacyclododecane-N,N′,N″,N‴-tetraacetic acid SHALs were synthesized using solution phase coupling of the free amine SHALs with 1,4,7,10-tetra-azacyclododecane- N,N′,N″,N‴-tetraacetic acid NHS ester of PF6 salt in DMF and DIEA. These SHALs were purified using reverse phase HPLC and characterized using electrospray ionization-mass spectrometry. Analytic HPLC was carried out on an Agilent 1100 instrument (Waters Symmetry C18 column), and preparative HPLC was carried out on a Waters preparative HPLC (Waters Symmetry prep C18 column). The nuclear magnetic resonance spectra obtained for each purified second generation SHAL (data not shown) was consistent with the intended structure. The observed yields and masses obtained for the purified SHAL4 to SHAL7 are shown in Table 1.

Analysis of SHAL binding to isolated HLA-DR10, live cells, and fixed tissue. The binding of SHALs to the recombinant HLA-DR10 β subunit was analyzed using surface plasmon resonance on a BIAcore 3000 (Biacore International AB). Equivalent amounts of an antipolyhistidine monoclonal antibody (R&D Systems, Inc.) were immobilized on two flow cells of a CM5 chip using standard amine coupling, and the recombinant HLA-DR10 β subunit was captured on one of the two cells. The cell without the β subunit was used as a reference or “blank.” Individual SHALs prepared at the same molar concentration were injected sequentially into both cells in 1.0× PBS, 5% DMSO for 1 min at a flow rate of 50 μL/min. The binding and dissociation of each SHAL was recorded by monitoring the change in response units observed for each cell. Three minutes were allowed for the first SHAL to dissociate and the signal to return to baseline before the next SHAL was injected and the process repeated. This approach allowed the binding of a set of SHALs to the same β-subunit sample to be compared. The resulting curves were double referenced, molecular weight was adjusted, and solvent was corrected. In those experiments used to determine if SHAL bound to the same site as Lym-1, Lym-1 was added to the captured β subunit before SHAL addition. Those SHALs binding within the Lym-1 epitope no longer bound after Lym-1 addition.

Experiments conducted to determine if SHAL binds to live Raji, Ramos, and other cells expressing HLA-DR10 and a variety of cell lines known to not contain HLA-DR10 were done in an ELISA format in 96-well plates. The plates were coated with streptavidin, the biotinylated SHAL was added, and after rinsing to remove unbound SHAL, suspensions of cells (0.5-1.0 × 106 per well) were added to the wells and incubated for 1 h at 37°C. Cells were also added to wells prepared by coating with Lym-1 antibody to compare SHAL and Lym-1 binding to each cell line. Excess and unbound cells were removed by gently rinsing the wells thrice with PBS containing 5% bovine serum albumin, each well was fixed for 10 min with 10% buffered formalin and subsequently rinsed with PBS. The cells remaining bound were stained with Cresyl violet, the plates were rinsed to remove excess stain, and the stained cells bound to the well were visualized and photographed under the microscope.

Immunohistochemistry was done on formalin-fixed, paraffin-embedded lymphoma tissue sections from three patients after deparaffinization, clearing inhibition of endogenous peroxidase activity in a solution of 3% hydrogen peroxide in methanol and hydration in decreasing concentrations of alcohol to distilled water. Before incubation with SHAL or Lym-1, antigen retrieval was done by high-temperature (microwave) incubation in 0.01 mol/L citric acid buffer (pH 6.0) for 3 × 4 min. After cooling and transfer to PBS (pH 7.4; 2 × 5 min each), 10% normal horse serum (Vector Laboratories) was applied to the sections and incubated for 20 min in a humidified chamber at room temperature. Slides were covered with primary SHAL and allowed to sit overnight. The Vectastain ABC Elite kit (Vector Laboratories) was used as amplification system according to the manufacturer's instructions. Slides were counterstained in Mayer's hematoxylin, dehydrated, cleared, and coverslipped. One negative control slide per sample run was run without primary antibody.

Competitive binding with Scatchard analysis was used to estimate the binding constants for SHAL4 to SHAL7 to Raji cell HLA-DR10 antigen. One million HLA-DR10–positive (Raji) and HLA-DR10–negative (CEM) cells were incubated with 10−10 to 2 × 10−13 mol/L SHAL in a total volume of 150 μL of 5% bovine serum albumin in PBS for 1 h. The bound counts (cell pellet) and free counts (supernatant) were collected by centrifugation and counted in a calibrated gamma well counter. Data for the CEM cells were subtracted from that of the Raji cells to obtain the cell-specific binding constants.

Unique HLA-DR10–specific ligand binding sites. Lym-1 is a murine IgG2a monoclonal antibody that selectively binds an abundant protein located on the surface of malignant human B-cell lymphocytes (16). This protein, HLA-DR10, has an epitope that contains several amino acid substitutions not found in other HLA-DR subtypes (17, 18). The epitope, which seems to be localized to the β subunit, has been shown to contain two amino acid residues, R70 and R71, both of which are required for Lym-1 binding. HLA-DR molecules with amino acid substitutions at either position do not bind the antibody.

To create an HLA-DR10–specific SHAL that binds to the surface of the protein in the proximity of these two residues, it was necessary to first develop a three-dimensional model of the HLA-DR10 protein. Crystal structures for four different closely related HLA-DR molecules (HLA-DR1–HLA-DR4) were identified as candidate structural templates (1922), and the protein sequences of these four proteins were aligned with the HLA-DR10 sequence to locate the variable amino acids and identify those regions of HLA-DR10 that contribute to the critical epitope of the Lym-1 antibody (17, 18). This alignment showed that all five proteins exhibited a high degree of sequence similarity (93-97%) and confirmed that a reasonably accurate three-dimensional model of the HLA-DR10 β subunit could be generated by homology modeling. The results of the modeling revealed that amino acid residues that contribute to Lym-1 binding were located in the same β-subunit α-helix. R70 and R71 were located near the center of the helix, whereas another Lym-1 reactive residue, V85, was positioned at the end of the helix directly adjacent to the hinge that links the two structural domains (Fig. 1A) of the β subunit.

Cavities on the solvent accessible surface of HLA-DR10 were identified and used as potential binding sites for ligands. Several sites surrounding the key amino acids (R70 and R71) in the Lym-1 epitope were examined. Two pockets (site 1 and site 2) in the modeled HLA-DR10 surface, flanking both sides of R70, were selected as appropriate sites for ligand binding (Fig. 1B) because of their proximity to each other, their position relative to the Lym-1 reactive residues, and the unique features (shape of cavity and nature of surrounding amino acid side chains) that define each pocket (Fig. 1C).

Identification of small molecule ligands that bind to HLA-DR10. Virtual screening of a database of ∼200,000 small molecules identified the top ranked 2,500 molecules predicted to bind in site 1 and site 2. After visual examination of the results, 35 compounds representing the full range of structural families present in each ligand set were selected for experimental testing to determine if they actually bind to the protein. Transferred nuclear Overhauser effect spectroscopy (2327) and water LOGSY experiments (28) were used to screen a subset (based on cost and availability) of the top 35 predicted site 1 and site 2 ligands, in mixtures of several ligands per experiment, to determine which compounds bound to the HLA-DR10 protein isolated from Raji cells. Eleven of thirty molecules tested were found to bind to the isolated HLA-DR10 protein (Table 2). Six of the molecules were ligands predicted to bind to site 1; five were ligands predicted to bind to site 2. Competition experiments showed that each of the five site 2 ligands bound to a different site than deoxycholate (one of the site 1 ligands), thereby making it a suitable partner for linking to deoxycholate.

Table 2.

Ligands determined to bind to isolated HLA-DR10 using nuclear magnetic resonance spectroscopy

Site 1
• Methidiumpropyl EDTA 
• Deoxycholic acid 
• 5(6)-Carboxytetramethylrhodamine N-succinimidyl ester 
N-(9-Fluorenyl)methoxycarbonyl-aspartic acid(O-benzyl)-OH 
• 4-Dimethylaminoazobenzene-4′-sulfonyl-l-valine 
N,N′-bis-(4-amino-2-chloro-phenyl)-terephthalamide-4-[[5-(trifluoromethyl)pyridin-2-yl]oxy]phenyl N-phenylcarbamate 
 
Site 2
 
• Leu-enkephalin 
• Asp-arg-val-tyr 
N-α-benzoyl-arginine-4-amino benzoic acid 
N-α-N-omega-dicarbobenzoxyarginine 
• Bis-t-butoxycarbonyl-l-arginine 
Site 1
• Methidiumpropyl EDTA 
• Deoxycholic acid 
• 5(6)-Carboxytetramethylrhodamine N-succinimidyl ester 
N-(9-Fluorenyl)methoxycarbonyl-aspartic acid(O-benzyl)-OH 
• 4-Dimethylaminoazobenzene-4′-sulfonyl-l-valine 
N,N′-bis-(4-amino-2-chloro-phenyl)-terephthalamide-4-[[5-(trifluoromethyl)pyridin-2-yl]oxy]phenyl N-phenylcarbamate 
 
Site 2
 
• Leu-enkephalin 
• Asp-arg-val-tyr 
N-α-benzoyl-arginine-4-amino benzoic acid 
N-α-N-omega-dicarbobenzoxyarginine 
• Bis-t-butoxycarbonyl-l-arginine 

First generation bidentate SHALs. Three first generation SHALs were synthesized by linking together the two ligands leu-enkephalin and deoxycholate (Fig. 2A). The optimal linker length was estimated by measuring the distance between the two bound ligands in the computer model of the relevant ligand-protein complexes generated during the docking runs, and the combination of lysine and PEG monomers were used to create a SHAL set with ligands separated by the predicted length (lysine-PEG linker) and one PEG monomer longer and shorter. Biotin was incorporated into SHAL to facilitate the in vitro binding studies of SHAL to HLA-DR10 and to test the selectivity of SHAL for binding to live cells and tissue sections.

Fig. 2.

Structure of the first generation human lymphoma SHALs (A) and their relative binding to the HLA-DR10 β-subunit (B) as determined by surface plasmon resonance. The recombinant β subunit of HLA-DR10 was attached to a CM5 chip via its (his)6 terminal tag, which was captured by an immobilized polyhistidine antibody. The binding curves were collected using a BIAcore 3000 after injection of equimolar concentrations of SHAL1 to SHAL3.

Fig. 2.

Structure of the first generation human lymphoma SHALs (A) and their relative binding to the HLA-DR10 β-subunit (B) as determined by surface plasmon resonance. The recombinant β subunit of HLA-DR10 was attached to a CM5 chip via its (his)6 terminal tag, which was captured by an immobilized polyhistidine antibody. The binding curves were collected using a BIAcore 3000 after injection of equimolar concentrations of SHAL1 to SHAL3.

Close modal

Surface plasmon resonance binding experiments were conducted to confirm that SHALs bound to the β subunit of HLA-DR10. In these experiments, a recombinant form of the HLA-DR10 β subunit with a COOH terminus histidine tag was captured by an immobilized polyhistidine monoclonal antibody on an surface plasmon resonance chip, and the binding of each SHAL to the protein was measured by passing each SHAL over the antibody-protein complex and a control surface containing only the antibody. All three SHALs were observed to bind to the β subunit (Fig. 2B).

To determine if SHAL would bind to live human lymphoma cells and to assess the selectivity of binding, ELISA plates were coated with streptavidin, SHAL2 (LePLDB) was introduced and allowed to bind to the streptavidin, and the plates were rinsed to remove unbound SHAL. Suspensions of cultured lymphoma and other cancer cells were added to the plates, and the plates were incubated, then washed and stained with Cresyl violet to visualize and quantify the amount of cell binding to the immobilized SHAL on the plates. As shown in Fig. 3, SHAL2 bound to nine different HLA-DR10 expressing human lymphoma cell lines, but it did not bind to cells that were known to lack HLA-DR10.

Fig. 3.

Selectivity of SHAL binding to human B-cell lymphoma cell lines. ELISA plates were coated with streptavidin, SHAL 2 was added to bind to the SA, and the plates were rinsed to remove unbound SHAL. Suspensions of cultured lymphoma and other cancer cells were added to the plates, and the plates were incubated, then washed and stained with Cresyl violet to visualize and quantify the amount of cell binding to the immobilized SHAL on the plates.

Fig. 3.

Selectivity of SHAL binding to human B-cell lymphoma cell lines. ELISA plates were coated with streptavidin, SHAL 2 was added to bind to the SA, and the plates were rinsed to remove unbound SHAL. Suspensions of cultured lymphoma and other cancer cells were added to the plates, and the plates were incubated, then washed and stained with Cresyl violet to visualize and quantify the amount of cell binding to the immobilized SHAL on the plates.

Close modal

Experiments carried out using horseradish peroxidase-tagged streptavidin–conjugated SHAL2 to stain frozen human lymphoma tissue sections also revealed that SHAL binds to human lymphoma cells with a pattern of binding that mimics that of the Lym-1 antibody. In particular, the larger, more undifferentiated lymphoma cells exhibited more pronounced staining than the small, more differentiated lymphoma cells (Fig. 4). SHAL binding to these B-cell lymphoma sections was not affected by tissue fixation or embedding in paraffin. SHAL2 bound equally well to frozen tissue, formalin-fixed tissue, and fixed tissue after antigen retrieval, whereas Lym-1 antibody only bound well to fresh or frozen B-cell lymphoma tissue.

Fig. 4.

Binding of SHAL2 and Lym-1 antibody to B-cell lymphoma tissue from patients. SHAL2 was preincubated with horseradish peroxidase-tagged streptavidin and binding to the cells in the tissue was detected by 3,3′-diaminobenzidine reagent. Lym-1 binding was detected with a biotinylated antimouse monoclonal antibody, followed by horseradish peroxidase-tagged streptavidin and 3,3′-diaminobenzidine. A, SHAL 2 on large cell lymphoma. B, SHAL 2 on small cell lymphoma. C, Lym-1 antibody on large cell lymphoma. D, Lym-1 antibody on small cell lymphoma.

Fig. 4.

Binding of SHAL2 and Lym-1 antibody to B-cell lymphoma tissue from patients. SHAL2 was preincubated with horseradish peroxidase-tagged streptavidin and binding to the cells in the tissue was detected by 3,3′-diaminobenzidine reagent. Lym-1 binding was detected with a biotinylated antimouse monoclonal antibody, followed by horseradish peroxidase-tagged streptavidin and 3,3′-diaminobenzidine. A, SHAL 2 on large cell lymphoma. B, SHAL 2 on small cell lymphoma. C, Lym-1 antibody on large cell lymphoma. D, Lym-1 antibody on small cell lymphoma.

Close modal

Second generation bidentate SHALs produced by ligand substitution. To maximize the stability of SHAL in vivo (leu-enkephalin has a half life of <15 min when it enters the blood; ref. 29), minimize its uptake by the liver (deoxycholate is similar in structure to bile acids), and increase SHAL's affinity for HLA-DR10, a series of second generation SHALs were created by successively replacing deoxycholate and leu-enkephalin with other site 1 and site 2 ligands. The same linker, a lysine conjugated to a single PEG monomer, was used to connect each of these new ligand combinations, and the chelator 1,4,7,10-tetra-azacyclododecane-N,N′,N″,N‴-tetraacetic acid was attached to the third position on the linker in place of the biotin used in the first generation SHALs. This enabled SHAL to be radiolabeled with 111In and to be used in binding experiments to determine their affinities for whole-lymphoma cells. The best of these new SHALs, which contained dabsylvaline, as the site 1 ligand, and N-benzoyl-l-arginyl-4-amino-benzoic acid, as the site 2 ligand (SHAL4; Fig. 5), bound to HLA-DR10–expressing Raji cells with a Kd of ∼80 nmol/L.

Fig. 5.

A, structure of the second generation bidentate SHAL4 (DvLPBaPLDo) and B, one of the three bisbidentate SHALs, SHAL7 [(DvLPBaPPPP)2LLDo].

Fig. 5.

A, structure of the second generation bidentate SHAL4 (DvLPBaPLDo) and B, one of the three bisbidentate SHALs, SHAL7 [(DvLPBaPPPP)2LLDo].

Close modal

Bis-bidentate SHALs. One important structural feature of all antibodies is that they contain two identical antigen binding sites. Whereas many antibodies use only one of these sites to bind to their targets, those that bind bivalently exhibit much higher affinities. To improve the avidity of these second generation SHALs for lymphoma cells, a series of bisbidentate SHALs were synthesized in which two bidentate SHAL4 molecules were linked together using lysine residues and a series of PEG monomers to produce a SHAL capable of simultaneously binding to two neighboring HLA-DR10 molecules in a manner similar to the bivalent binding exhibited by some of the best antibodies. The length of the linker needed to span the distance between two HLA-DR10 proteins on the cell surface, about 50 to 90 Å, was estimated using the physical dimensions of an IgG antibody and the calculated distance of separation between HLA-DR10 molecules on the Raji cell surface. Three bisbidentate SHALs were subsequently synthesized with 53 Å (SHAL5), 75 Å (SHAL6), and 93 Å (SHAL7) linker lengths. The structure of the bisbidentate SHAL with the longest linker, SHAL7, is shown in Fig. 5.

Whole Raji cell binding experiments carried out using the three 111In-labeled bisbidentate SHALs showed that the process of linking pairs of SHALs together to obtain molecules capable of binding to two neighboring HLA-DR10 molecules increased the avidity of SHAL for cells expressing HLA-DR10 by 1,000-fold to 4,000-fold over that observed for the bidentate SHAL. Picomolar cell-specific Kd's were obtained for all three bisbidentate SHALs (Table 3).

Table 3.

Avidities of bis-bidentate SHALs obtained from Raji (and CEM) cell binding experiments

SHALKd
SHAL4: DvLPBaPLDo 80 nmol/L 
SHAL5: (DvLPBaPP)2LLDo 93 pmol/L 
SHAL6: (DvLPBaPPP)2LLDo 51 pmol/L 
SHAL7: (DvLPBaPPPP)2LLDo 18 pmol/L 
SHALKd
SHAL4: DvLPBaPLDo 80 nmol/L 
SHAL5: (DvLPBaPP)2LLDo 93 pmol/L 
SHAL6: (DvLPBaPPP)2LLDo 51 pmol/L 
SHAL7: (DvLPBaPPPP)2LLDo 18 pmol/L 

Comparative analyses of normal cells and their transformed counterparts have identified a number of differences in cell surface proteins that can be exploited for both the identification and treatment of a variety of cancers. B-cell lymphocytes and the human lymphomas that develop from these lymphocytes, for example, have been shown to contain an abundant sequence variant of HLA-DR, HLA-DR10 (1618), that does not seem to be present on the surface of other cells. Furthermore, the mucins found on breast and prostate cancer cells exhibit significant differences in the extent of glycosylation and level of branching of their carbohydrates (30, 31). Other tumors have also been determined to have higher densities of various cell surface proteins, including prostate-specific membrane antigen (3235), CD20 (36), CD22 (37), and CD43 (38). These differences in protein sequence, posttranslational modification, and concentration on the cell surface have enabled the development of a number of tumor-specific antibodies (16, 3943) for use both as diagnostics and as therapeutics. Whereas the application of some of these antibodies as direct acting therapeutics has been successful, their use as carriers for cytocidal agents, such as radionuclides in radioimmunotherapy, has shown even greater promise. However, the limitations that have been experienced in using antibodies as therapeutics or targeting agents suggest that smaller nonprotein-based reagents are needed as replacements for antibodies if the field is to move forward in more than incremental steps. Recent successes in creating high-affinity synthetic ligands by linking together small molecules that bind to known sites on proteins (1014) indicate that this should be possible, provided reasonably selective ligands that bind to functionally important or structurally unique features on the protein's surface can be identified.

Experiments conducted with the present SHALs show that small synthetic molecules (<4,000 Da) can be created that bind to unique structural features located on the surface of proteins within epitopes targeted by antibodies. The HLA-DR10 site selected for SHAL binding is located within a region of the β subunit containing the three amino acid residues determined previously (17, 18) to be important for the binding of the antibody Lym-1. The side chain of one of these amino acids, R70, is positioned between the two cavities selected for binding the two ligands that were linked together to create SHAL. The side chains of the other two residues, A74 and R71, change the shape of the cavities and charge distribution surrounding them sufficiently to allow the identification of small molecules that bind selectively to HLA-DR10.

These SHALs were developed using a virtual ligand screening approach (molecular docking) to select candidate small molecules predicted to bind to the two cavities and to reduce the number of ligands that needed to be tested experimentally. Whereas it is recognized that the force fields and energetics used by the present docking programs are less than ideal, the results we report here for HLA-DR10 and our previous studies with the C fragment of tetanus neurotoxin (24, 44) show that these methods can provide sufficient discrimination to effectively screen very large libraries of small molecules and identify a small set of compounds for experimental testing. In each of these studies, <40 compounds were screened for binding to the target proteins and 25% to 56% of the compounds tested were experimentally determined to bind to the protein.

The binding of the first generation SHAL2 to nine cell lines expressing HLA-DR10, as well as the absence of its binding to seven cell lines known to lack HLA-DR10, confirms its selectivity for cells containing HLA-DR10 and the SHAL's potential utility as a targeting agent or diagnostic for lymphoma and leukemias. The absence of SHAL2 binding to the cell lines lacking HLA-DR10 also provides evidence that SHALs do not bind with significant affinities to other cell surface proteins. The observed subtle differences in staining of the larger, more undifferentiated lymphoma cells and the smaller, more differentiated lymphoma cells observed in SHAL-treated lymphoma biopsy sections not only show the extent to which SHAL mimics the binding that has been observed when using the antibody Lym-1, but the successful staining of fixed, paraffin-embedded tissue sections also points out another useful feature of small molecule ligands. Because small molecules bind to a much smaller region on the protein surface, their binding is less likely to be disrupted by fixation techniques that chemically modify selected amino acids (e.g., aldehyde modification of lysine).

The demonstration that these first SHALs mimic so closely the binding selectivity of the Lym-1 antibody is encouraging and suggests that these prototype molecules may be useful future therapeutics for non–Hodgkin's lymphoma. HLA-DR10, the antigen containing the epitope recognized by Lym-1, is an abundant (∼106 molecules per cell) cell surface marker that is present on ∼80% of lymphoma cells. In contrast, only ∼2% of healthy B cells express the HLA-DR10 protein, and the level of expression is ∼10-fold less than on tumor cell lines. This indicates that SHALs targeting HLA-DR10 should have a limited effect on the healthy B-cell population and allow them to continue making antibodies to fight infection.

Clearly, one of the most exciting aspects of this work is the demonstration that pairs of bidentate SHALs can be linked together to create bisbidentate molecules 1/50th the size of an antibody that appear to bind bivalently to two neighboring HLA-DR10 molecules on the surface of lymphoma (Raji) cells. The bisbidentate ligands generated by linking together two SHAL4 molecules resulted in a several thousand-fold increase in avidity for cells expressing HLA-DR10, from the near nanomolar Kd obtained for the bidentate SHAL4 to picomolar Kd's obtained for the bisbidentate versions of the same SHAL. This result substantiates the concept that small molecules can be created to bind to unique structural features on the surfaces of proteins with affinities similar to or exceeding those of antibodies.

This ability to create small protein-specific ligands that bind to selected sites on a protein's surface provides a powerful new approach that may be exploited to modulate protein function and enable the development of a new spectrum of therapeutics and diagnostic reagents. Small molecules identified to bind to unique cavities located near the active site of an enzyme, for example, could be used to develop bidentate inhibitor analogues (new ligands linked to existing inhibitors) that exhibit increased affinities and selectivities. The approach also makes it possible to consider designing steric inhibitors that block catalytic reactions without actually binding within the active site of the enzyme. Other inhibitors could be developed that block intermolecular associations between proteins by binding tightly to the sites on the surfaces of the proteins required for complex formation. In certain cases, it might even become possible to design SHALs that bind the most subtle structural features on proteins, such as structural alterations in the surface of the protein induced by posttranslational modifications or the natural binding of cofactors or other small molecule ligands.

Grant support: National Cancer Institute PO1-CA47829 (G. DeNardo) and Lawrence Livermore National Laboratory LDRD awards 01-ERD-111 (R. Balhorn and J. Perkins), 01-ERD-046 (M. Cosman), and 01-SI-012 (F. C. Lightstone).

Presented at the Eleventh Conference on Cancer Therapy with Antibodies and Immunoconjugates, Parsippany, New Jersey, USA, October 12-14, 2006.

Note: The work at Lawrence Livermore National Laboratory was done under the auspices of the U.S. Department of Energy under contract W-7405-ENG-48.

1
Cheson B. Radioimmunotherapy of non-Hodgkin's lymphomas.
Blood
2003
;
101
:
391
–8.
2
DeNardo GL, O'Donnell RT, Oldham RK, DeNardo SJ. A revolution in the treatment of non-Hodgkin's lymphoma.
Cancer Biother Radiopharm
1998
;
13
:
213
–23.
3
DeNardo GL. Treatment of non-Hodgkin's lymphoma (NHL) with radiolabeled antibodies (mAbs).
Semin Nucl Med
2005
;
35
:
202
–11.
4
Juweid ME. Radioimmunotherapy of B-cell non-Hodgkin's lymphoma: from clinical trials to clinical practice.
J Nucl Med
2002
;
43
:
1507
–29.
5
Witzig TE, Gordon LI, Cabanillas F, et al. Randomized controlled trial of yttrium-90-labeled ibritumomab tiuxetan radioimmunotherapy versus rituximab immunotherapy for patients with relapsed or refractory low-grade, follicular, or transformed B-cell non-Hodgkin's lymphoma.
J Clin Oncol
2002
;
20
:
2453
–63.
6
Juweid ME, Swayne L, Sharkey RM, et al. Prospects of radioimmunotherapy in epithelial ovarian cancer: results with 131I-labeled murine and humanized MN-14 anti-carcinoembryonic antigen (CEA) monoclonal antibodies.
Gynecol Oncol
1997
;
67
:
259
–71.
7
Riva P, Franceschi G, Frattarelli M, et al. 131I radioconjugated antibodies for the locoregional radioimmunotherapy of high-grade malignant glioma-phase I, phase II. study.
Acta Oncol
1999
;
38
:
351
–9.
8
Paganelli G, Grana C, Chinol M, et al. Antibody guided three step therapy for high grade glioma with yttrium-90 biotin.
Eur J Nucl Med
1999
;
26
:
348
–57.
9
Wong JYC, Chu DZ, Yamauchi DM, et al. A phase I radioimmunotherapy trial evaluating 90yttrium-labeled anti-carcinoembryonic antigen (CEA) chimeric T84.66 in patients with metastatic CEA-producing malignancies.
Clin Cancer Res
2000
;
6
:
3855
–63.
10
Shuker SB, Hajduk PJ, Meadows RP, Fesik SW. Discovering high-affinity ligands for proteins: SAR by NMR.
Science
1996
;
174
:
531
–4.
11
Kramer RH, Karpen JW. Spanning binding sites on allosteric proteins with polymer-linked ligand dimmers.
Nature
1998
;
395
:
710
–3.
12
Maly DJ, Choong IC, Ellman JA. Combinatorial target-guided ligand assembly: identification of potent subtype-selective c-Src inhibitors.
Proc Natl Acad Sci U S A
2000
;
97
:
2419
–24.
13
Kitov PI, Sadowska JM, Mulvey G, et al. Shiga-like toxins are neutralized by tailored multivalent carbohydrate ligands.
Nature
2000
;
403
:
669
–72.
14
Fan E, Zhang Z, Minke WE, Hou Z, Verlinde CLMJ, Hol WG. High-affinity pentavalent ligands of Escherichia coli heat-labile enterotoxin by modular structure-based design.
J Am Chem Soc
2000
;
122
:
2663
–4.
15
Mammen M, Choi, S-K, Whitesides GM. Polyvalent interactions in biological systems: implications for design and use of multivalent ligands and inhibitors.
Angew Chem Int Ed
1998
;
37
:
2754
–94.
16
Epstein AL, Marder RJ, Winter JN, et al. Two new monoclonal antibodies, Lym-1 and Lym-2, reactive with human-B-lymphocytes and derived tumors, with immunodiagnostic and immunotherapeutic potential.
Cancer Res
1987
;
47
:
830
–40.
17
Rose LM, Gunasekera AH, DeNardo SJ, DeNardo GL, Meares CF. Lymphoma-selective antibody Lym-1 recognizes a discrete epitope on the light chain of HLA-DR10.
Cancer Immunol Immunother
1996
;
43
:
26
–30.
18
Rose LM, Deng CT, Scott SL, et al. Critical lym-1 Binding residues on polymorphic HLA-DR molecules.
Mol Immunol
1999
;
36
:
789
–97.
19
Ghosh P, Amaya M, Mellins E, Wiley DC. The structure of an intermediate in class II MHC maturation: CLIP bound to HLA-DR3.
Nature
1995
;
378
:
457
–62.
20
Smith KJ, Pyrdol J, Gauthier L, Wiley DC, Wucherpfennig KW. Crystal structure of HLA-DR2 (DRA*0101, DRB1*1501) complexed with a peptide from human myelin basic protein.
J Exp Med
1998
;
188
:
1511
–20.
21
Bolin DR, Swain AL, Sarabu R, et al. Peptide and peptide mimetic inhibitors of antigen presentation by HLA-DR class II MHC molecules. Design, structure-activity relationships, and X-ray crystal structures.
J Med Chem
2000
;
43
:
2135
–48.
22
Jardetzky TS, Brown JH, Gorga JC, et al. 3-dimensional structure of a human class-II histocompatibility molecule complexed with superantigen.
Nature
1994
;
368
:
711
–8.
23
Henrichsen D, Ernst B, Magnani JL, Wang W, Meyer B, Peters T. Bioaffinity NMR spectroscopy identification of an E-selection antagonist in a substance mixture by transfer NOE.
Angew Chem Int Ed
1999
;
38
:
98
–102.
24
Cosman M, Lightstone FC, Krishnan VV, et al. Identification of novel small molecules that bind to two different sites on the surface of tetanus toxin C fragment.
Chem Res Toxicol
2002
;
15
:
1218
–28.
25
Clore GM, Gronenborn AM, Mitchinson C, Green NM. 1H-NMR studies on nucleotide binding to the sarcoplasmic reticulum Ca2+ ATPase. Determination of the conformations of bound nucleotides by the measurement of proton-proton transferred nuclear Overhauser enhancements.
Eur J Biochem
1982
;
128
:
113
–7.
26
Gronenborn AM, Clore GM. Conformation of NAD+ bound to yeast and horse liver alcohol dehydrogenase in solution. The use of the proton-proton transferred nuclear Overhauser enhancement.
J Mol Biol
1982
;
157
:
155
–60.
27
Roberts GCK. The other kind of biological NMR - studies of enzyme substrate interactions.
Neurochem Res
1996
;
21
:
1117
–24.
28
Dalvit C, Fogliatto G, Stewart A, Veronesi M, Stockman B. WaterLOGSY as a method for primary NMR screening.
J Biomol NMR
2001
;
21
:
349
–59.
29
Mosnaim AD, Puente J, Saavedra R, Diamond S, Wolf ME. In vitro human plasma leucine(5)-enkephalin degradation is inhibited by a select number of drugs with the phenothiazine molecule in their chemical structure.
Pharmacol
2003
;
67
:
6
–13.
30
Hull SR, Bright A, Carraway KL, Abe M, Hayes DF, Kufe DW. Oligosaccharide differences in the DF3 sialomucin antigen from normal human milk and the BT-20 human breast carcinoma cell line.
Cancer Commun
1989
;
1
:
261
–7.
31
Lloyd KO, Burchell J, Kudryashov V, Yin BW, Taylor-Papadimitriou J. Demonstration of simpler and fewer glycan chains in tumor cells. Comparison of O-linked carbohydrate chains in MUC-1 mucin from normal breast epithelial cell lines and breast carcinoma cell lines. Demonstration of simpler and fewer glycan chains in tumor cells.
J Biol Chem
1996
;
271
:
33325
–34.
32
Murphy GP, Elgamal AA, Su SL, Bostwick DG, Holmes EH. Current evaluation of the tissue localization and diagnostic utility of prostate specific membrane antigen.
Cancer
1998
;
83
:
2259
–69.
33
Sweat SD, Pacelli A, Murphy GP, Bostwick DG. Prostate-specific membrane antigen expression is greatest in prostate adenocarcinoma and lymph node metastases.
Urology
1998
;
52
:
637
–40.
34
Lui H, Moy P, Kim S, et al. Monoclonal antibodies to the extracellular domain of prostate-specific membrane antigen also react with tumor vascular epithelium.
Cancer Res
1997
;
57
:
3629
–34.
35
Chang SS, Reuter VE, Heston WDW, Bander NH, Grauer LS, Gaudin PB. Five different anti-prostate-specific membrane antigen (PMSA) antibodies confirm PSMA expression in tumor-associated neovasculature.
Cancer Res
1999
;
59
:
3192
–8.
36
Lundin J, Osterborg A. Advances in the use of monoclonal antibodies in the therapy of chronic lymphocytic leukemia.
Semin Hematol
2004
;
41
:
234
–45.
37
Mason DY, Stein H, Gerdes J, et al. Value of monoclonal anti-CD22 (p135) antibodies for the detection of normal and neoplastic B lymphoid cells.
Blood
1987
;
69
:
836
–40.
38
Fernandez-Rodriguez J, Andersson CX, Laos S, et al. The leukocyte antigen CD43 is expressed in different cell lines of nonhematopoietic origin.
Tumour Biol
2002
;
23
:
193
–201.
39
Leonard JP, Link BK. Immunotherapy of Non-Hodgkin's lymphoma with hLL2 (Epratuzumab, an anti-CD22 monoclonal antibody) and HuID10 (Apolizumab).
Semin Oncol
2002
;
29
:
81
–6.
40
Link BK, Kostelny SA, Cole MS, Fusselman WP, Tso JY, Weiner GJ. Anti-CD3-based bispecific antibody designed for therapy of human B-cell malignancy can induce T-cell activation by antigen-dependent and antigen-independent mechanisms.
Int J Cancer
1998
;
77
:
251
–6.
41
Smith-Jones PM, Vallabahajosula S, Goldsmith SJ, et al. In vitro characterization of radiolabeled monoclonal antibodies specific for the extracellular domain of prostate-specific membrane antigen.
Cancer Res
2000
;
60
:
5237
–43.
42
DeNardo GL, O'Donnell RT, Rose LM, Mirick GR, Kroger LA, DeNardo, SJ. Milestones in the development of Lym-1 therapy.
Hybridoma
1999
;
18
:
1
–11.
43
Wilder RB, DeNardo GL, DeNardo SJ. Radioimmunotherapy: recent results and future directions.
J Clin Oncol
1996
;
14
:
1383
–400.
44
Lightstone FC, Prieto MC, Singh AK, et al. The identification of novel small molecule ligands that bind to tetanus toxin.
Chem Res Toxicol
2000
;
13
:
356
–62.
45
Venclovas C. Comparative modeling of CASP4 target proteins: combining results of sequence search with three-dimensional structure assessment.
Proteins Suppl
2001
;
5
:
47
–54.