Comparative properties of the single chain antibody and Fv derivatives of mAb 4-4-20. Relationship between interdomain interactions and the high affinity for fluorescein ligand.

Recombinant Fv derivative of the high affinity murine anti-fluorescein monoclonal antibody 4-4-20 was constructed and expressed in high yields, relative to the single chain antibody (SCA) derivative (2 3-fold), in Escherichia coli. Both variable heavy (VH) and variable light (VL) domains, that accumulated as insoluble inclusion bodies, were isolated, denatured, mixed, refolded, and affinity-purified to yield active Fv 4-4-20. Affinity-purified Fv 4-4-20 showed identical ligand binding properties compared with the SCA construct, both were slightly lower than the affinities expressed by Fab or IgG 4-4-20. Proper protein folding was shown to be domain-independent by in vitro mixing of individually refolded variable domains to yield functional Fv protein. In solid phase and solution phase assays, Fv 4-4-20 closely approximated the SCA derivative in terms of both idiotype and metatype, confirming identical active site structures and conformations. The equilibrium dissociation constant (Kd) for the VL/VH association (1.43 x 10(-7) M), which was determined using the change in fluorescein spectral properties upon ligand binding, was relatively low considering the high affinity displayed by the Fv protein for fluorescein (Kd, 2.9 x 10(-10) M). Thus, domain-domain stability in the Fv and SCA 4-4-20 proteins cannot be the sole cause of reduced affinity (2-3-fold) for fluorescein as compared with the Fab or IgG form of 4-4-20. With their identical ligand binding and structural properties, the decreased SCA or Fv affinity for fluorescein must be an ultimate consequence of deletion of the CH1 and CL constant domains. Collectively, these results verify the importance of constant domain interactions in antibody variable domain structure-function analyses and future antibody engineering endeavors.

Antibody Fv fragments are composed of heavy chain (V H ) 1 and light chain (V L ) variable domains. These two domains associate noncovalently to form the smallest functional antibody protein capable of antigen binding that most closely approximates the Ig molecule (1,2). These proteins have been previously found to be less stable in terms of domain-domain association than Fab fragments due to the lack of covalent bonds between the two variable domains (3,4). Single chain antibody (SCA) molecules have been produced to diminish this instability by the introduction of an interdomain linker peptide (5,6). SCA proteins often mimic the parent antibody active site in terms of antigen binding and structural properties with usually some reduction in affinity for antigen (7)(8)(9). Recently, Fv proteins have been engineered to possess an interdomain disulfide linkage, effectively disallowing dissociation of the two domains (10). Due to their small size and amenability to genetic engineering, recombinant Fv proteins have been widely applied in the study of antibody active site structure-function (7,(11)(12)(13), idiotypy and metatypy (14 -17), antibody bivalency and bispecificity (18 -21), and in vivo immunodiagnostics and therapy (10,22,23).
Fv molecules have been efficacious proteins in the study of antibody active site structure-function and protein stability. Studies involving comparative analysis of Fv protein with other immunoglobulin constructs afford unique opportunities for determining domain-domain interactions and the effects these interactions exert upon the intrinsic conformational and antigen binding properties of the variable domains. Being covalently coupled by an interdomain linker, SCA proteins have been suggested to possess greater interdomain stability than their Fv counterparts due to the favorable entropic effect of domain coupling (6,8). This would, in turn, suggest that in the appropriate Fv molecule (one with high affinity for antigen), interdomain associative properties would dictate the overall affinity displayed for antigen because only associated V L /V H proteins would bind antigen. In previous studies, dissociation constants for the V L /V H association in Fv molecules varied from 10 Ϫ7 to Ͼ10 Ϫ9 M (3, 24 -26). These Fv molecules also displayed similar dissociation constants for their respective antigens (10 Ϫ7 to Ͼ10 Ϫ9 M), further supporting some correlation between interdomain and active site/antigen interactions. Further analysis of V L /V H association constants in relation to antigen affinity would allow identification of components necessary for the production of stable Fv molecules and novel variable domain proteins.
Fv molecules have been especially useful in the study of idiotypy and metatypy. Antibody idiotype and metatype are immunologically resolved markers of active site structural and conformational determinants in the unliganded and liganded state, respectively (review in Refs. 27 and 28). Indeed, the transition between the idiotypic and metatypic states upon ligand binding emphasizes the dynamic properties of antibody proteins (15). The principle of antibody dynamics is governed by the process of structural fluctuation in both the unliganded and liganded states until the most energetically favorable state is established (29,30). Thus, understanding how structural and dynamic features are exerted within antibody variable domains will directly influence our understanding of antibody active site/ligand binding properties. Small differences in antigen binding affinities have been found between certain SCA and IgG despite their structural relatedness (7). The possibility exists that constant/variable domain interactions results in more than structural rigidity, but are responsible for restricted variable domain conformational dynamics favoring antigen binding. Indeed, studies with MOPC 315, a murine anti-nitrophenyl IgA antibody, have indicated that the C H 1 domain in Fab fragments influences idiotypic expression of the antibody through a dynamic effect on the variable domain structure (31). In other antibodies, studies have demonstrated the importance of cooperative interface interactions (cis and trans) between the variable and constant domains regarding variable domain stability and antigen binding (7,32). To examine this, Fv and SCA proteins, due to differences in interdomain associative properties with similar ligand binding to IgG, represent critical reagents to study ligand binding properties and variable region structural features as influenced by antibody constant domains.
To study the relation between interdomain association and affinity for antigen, an antibody protein must be available in many derivative forms (e.g. with and without constant domain, with and without an intervariable domain linker). To satisfy these criteria, the Fv derivative of mAb 4-4-20, a high affinity murine anti-fluorescein antibody, has been synthesized. mAb 4-4-20 was a suitable antibody for this study due to its high degree of structural characterization (7,33,34) and the previous construction and characterization of SCA 4-4-20 (7,35). SCA 4-4-20 has been studied extensively in terms of active site environment (36,37), antigen binding structure (13), and thermodynamics (8,38). Comparison of SCA with Fab 4-4-20 showed almost identical guanidine-induced denaturation profiles, idiotype and metatype expression, yet a 2-3-fold reduction in affinity for fluorescein (7). The structural and antigen binding properties required for fluorescein binding and quenching by SCA and mAb 4-4-20 have also been extensively analyzed in relation to the remainder of the 4-4-20 idiotype antifluorescein antibody family (11). Comparison of Fv 4-4-20 with SCA, Fab, and IgG may assist in understanding the basis for this difference in affinity for antigen upon removal of the 4-4-20 constant domains. Studies reported here show that Fv 4-4-20 possessed identical structural, idiotypic, metatypic, and ligand binding properties as SCA 4-4-20. With confirmation of identical ligand binding and structural characteristics between Fv and SCA, the dissociation constant (K d ) for the V L /V H association was determined and analyzed in relation to the Fv affinity for fluorescein. Such analyses implicated the necessity of constant domain/variable domain association in the formation of the high affinity liganded state. Overall, results indicated that the reduced affinity of Fv and SCA 4-4-20 did not correlate with reduced variable domain association, but with the absence of antibody constant domains, emphasizing their role in antibody/ antigen interactions.
Strains, Plasmids, and Media-Escherichia coli strain GX6712 (F galk2 rpsL cI 857 ) and plasmid pGX8773 were provided originally by the Genex Corp. (now Enzon, Inc.). Expression vector pGX8773 encodes SCA 4-4-20/212 fused to the OmpA signal sequence and containing the interdomain 212 linker (GSTSGSGKSSEGKG) (6,45). The expression vector utilizes a hybrid O L /P R promoter with protein expression initiated by temperature shift from 30°C to 42°C in E. coli strain GX6712 (46). Reactions were incubated in a thermal cycler (MJ Research) using the following program: 92°C for 5 min, 53-67°C (depending on primer sequence) for 5 min, 72°C for 1 min, followed by 30 cycles of 72°C for 1 min, 92°C for 1 min, and 53-67°C for 1 min. The V L and V H genes were amplified separately to contain the 5Ј OmpA signal sequence and the 3Ј terminator sequences necessary for expression from pGX8773.
Following amplification, V L and V H PCR products were purified in low melting temperature agarose (Seaplaque, FMC) and cloned into SmaI digested pTZ18u (48). Correct clones were identified by restriction length analysis and verified by dideoxy sequencing. To construct the V L gene, PCR was used to add the transcription stop codons by 3Ј primer overhang (Fig. 1). To incorporate the signal sequence in the V H gene, PCR was used to amplify the OmpA sequence with the addition of the V H 5Ј sequence to the 3Ј end of OmpA. The resulting PCR product was then used as the 5Ј primer to incorporate the signal sequence to the V H gene (Fig. 1). Both amplified genes were cloned into pTZ18u to form pJWc2-2 and pJWc1-5, respectively. Following verification of proper primary sequence, the V L and V H genes were excised using ClaI-BamHI and cloned into pGX8773 for expression.
Sequence Determination-Following cloning, sequences of the PCR products were determined by the dideoxy chain termination procedure using a double-stranded plasmid DNA template (49) and Sequenase® (U. S. Biochemical Corp.).
Large Scale Expression of Fv 4-4-20 -Fv 4-4-20 was expressed in E. coli, denatured, and refolded using a modified version of the protocol used in Denzin et al. (45) and Rumbley et. al. (12). The procedure was modified in two ways: 1) in scale, to accommodate 1 liter of bacterial culture instead of 12 liters, and 2) in molecular weight cut-off size, all concentration and dialysis steps were performed using molecular weight cutoff of 3 kDa. Denatured V L and V H inclusions bodies were combined in a 1:1 mass ratio in the refolding solution to produce associated active Fv. Concentration of the diluted protein was accomplished using an Amicon Ultrafiltration Cell. Monomer V L and V H were produced by denaturation and renaturation in the absence of the other protein.
Anti-fluorescein Solid Phase ELISA Binding Assay-To demonstrate anti-fluorescein activity, a solid phase ELISA assay similar to that in Denzin et. al. (45) was performed. After addition of Fv 4-4-20 to fluorescein-BSA-coated wells and extensive washing, 50 l of 10 g/ml hamster anti-4-4-20 variable light domain-specific peptide antibody (3A5-1) (63) was added. Bound antibody was detected using horseradish peroxidase-labeled anti-hamster antibodies and 3,3Ј,5,5Ј-tetramethylbenzidine (Pierce). Substrate was added and incubated at room temperature for 30 min. Enzyme reactions were terminated with 2 N H 2 SO 4 and optical densities determined using a Dynatech MR500 automatic plate reader.
Dissociation Rate Kinetic Assay-Ligand dissociation rates of Fv 4-4-20 (90% liganded with fluorescein) were determined at 4°C as described (51). Affinity constants were determined from the dissociation rate using the previously determined association rate of 5 ϫ 10 6 M Ϫ1 s Ϫ1 for anti-fluorescyl antibodies to calculate intrinsic affinities (K a ϭ k 1 /k 2 ) (52, 53). Ligand dissociation rates were also performed in the presence of both polyclonal and monoclonal anti-metatype antibodies (54). Dissociation assays were performed as above, but liganded Fv 4-4-20 was preincubated with an active site molar excess of anti-metatype antibody active sites for 15-20 min at 4°C.
Fluorescein Fluorescence Quenching Assay-Fluorescence quenching measurements of antibody-bound ligand were performed as described by Watt and Voss (55). Fluorescein fluorescence quenching by affinitypurified Fv and by equal optical density amount of preincubated V L and V H protein was compared with SCA 4-4-20.
Circular Dichroism (CD) Studies-CD spectra of the antibody derivatives were recorded on a Jasco model JA720 instrument. A 0.1-cm cuvette (Hellma) was used for all measurements, and all spectra were averaged five times at room temperature. The bandwidth used was 2 nm, to a resolution of 0.1 nm, in all experiments. The buffer used in all experiments was PBS.
Hydrostatic Pressure Fluorescence Measurements-Fv and SCA 4-4-20 were compared by pressure-induced fluorescein dissociation. Hydrostatic pressure in the range of 1 bar to 2.4 kbar was achieved with a pressure cell as described by Paladini and Weber (56). Samples were excited at 480 nm (slit width of 8 nm) and emission spectra recorded in the range of 500 -600 nm (slit width of 8 nm). The intensity data were collected on an ISS GREG PC photon counting spectrofluorometer (ISS, Champaign, IL), and the intensity at each pressure was acquired by integrating the area under the emission spectra. Temperature was regulated with a circulating water bath and monitored by a thermocouple in direct contact with the stainless steel pressure cell. The temperature of the pressure cell was allowed to equilibrate for 1 h after temperature readings stabilized. Protein samples were prepared in 20 mM Tris-HCl, pH 8.0, and contained 0.127 M antibody active sites and 0.02 M fluorescein. The protein pressure samples were allowed to equilibrate for 4 min after each pressure change before spectra measurements were taken. tration. For anisotropy based measurements, affinity-labeled protein was serially diluted (1:2) over the concentration range of 23.5 M to 4 nM. Studies were performed using excitation at 480 nm and emission at 525 nm with slit widths at 1 and 2 nm. Steady state fluorescence anisotropy data were analyzed using Delta Graph Professional (Delta Point, Monterey, CA) as described (57) and K d values were determined from binding curves. Differences between quantum yield of bound and free fluorescein (due to domain dissociation upon dilution) result in differently weighted fluorescence anisotropy values (58). In order to correct for this, anisotropy data were analyzed in terms of degree dissociation (␣) at each protein concentration (56) as calculated from the following equation, where Q is the ratio between quantum yields of free and bound ligand, r f is the anisotropy value for fluorescein-labeled variable domain (0.02), and r b is the anisotropy for totally bound fluorescein (0.32). Protein concentrations were corrected for the R value of Fv affinity labeling.

Construction and Expression of Fv 4-4-20 -
Individual V L and V H gene constructs were assembled using both PCR and conventional cloning techniques (see "Materials and Methods"). The 3Ј stop codons and 3Ј BamHI restriction site were added to V L 4-4-20 by PCR technology. The OmpA signal sequence was added to V H 4-4-20 using a modified version of the megaprimer method of mutagenesis (59,60). Briefly, the OmpA sequence was amplified from SCA 4-4-20 incorporating a portion of the V H 5Ј sequence to the 3Ј end of OmpA. This PCR product was then used to amplify the entire V H gene, resulting in addition of the signal sequence. Protein yields from 1 liter of E. coli cultures were from 3 to 4 mg of affinity-purified Fv. This represented a 2-3-fold increase of active anti-fluorescein protein as compared with expression yields of SCA 4-4-20. Protein concentrations were calculated from absorption spectra at 240 -350 nm (61) using a Beckman DU-64 spectrophotometer. Extinction coefficients (A 278 nm mg/ml ) of 2.2, 2.1, 1.5 and 2.7 for SCA, Fv, V L , and V H proteins, respectively, were calculated from chromophore content (62).
Polyacrylamide Gel Analysis-Fv 4-4-20 was purified by affinity chromatography using fluorescein-Sepharose as described (see "Materials and Methods"). SDS-polyacrylamide gel electrophoresis analysis showed the purified Fv protein consisted of two detectable bands (14.0 and 12.5 kDa) corresponding to V H And V L proteins. Migration patterns indicated actual molecular weights for the two domains were in good agreement with their calculated values based on amino acid content (data not shown). Additionally, the affinity-purified material was shown to be Ͼ90% pure. Similar SDS-polyacrylamide gel electrophoresis analysis on individually refolded domain proteins showed that the V L and V H proteins were the major detectable band found in their respective samples (data not shown).
Anti-metatype Reactivity-To compare the degree of structural relatedness between the liganded states of SCA, Fab, and Fv 4-4-20, these proteins were used as polyclonal anti-metatype/liganded Fab 4-4-20-soluble inhibitors. All 4-4-20 proteins were affinity-labeled as described previously (Ref. 7; see "Materials and Methods"). Fig. 2B compares the inhibition titrations of affinity-labeled Fv with similarly labeled Fab and SCA 4-4-20. Unliganded Fab was also tested to determine the amount of anti-idiotype and anti-constant domain activity present in the anti-metatype reagent. The anti-metatype reagent was not passed over an unliganded IgG 4-4-20 adsorbent to remove such activity prior to this experiment. Results indicated that the Fv 4-4-20 possessed a similar anti-metatype inhibition profile as SCA 4-4-20, implying an overall structural similarity between their liganded states. Comparison of the unliganded and liganded Fab curves suggested the presence of anti-constant domain activity in the anti-metatype reagent, but confirmed specificity for the liganded state of the 4-4-20 active site.
Anti-idiotype Reactivity-In terms of a polyclonal anti-idiotype reagent, comparative inhibition studies revealed identical patterns of SCA and Fv anti-idiotype recognition. Results suggested that SCA and Fv 4-4-20 were idiotypically identical (Fig.  2C). Previous idiotypic analysis of SCA with mAb 4-4-20 indicated that the two were idiotypically identical (7).
Spectral Properties of Fluorescein Bound to Fv 4-4-20 -Antifluorescein antibodies have been characterized by their ability to quench (Q max ) the fluorescence of fluorescein (64). Q max for Fv (87.5 Ϯ 0.9%) compared well with SCA (85.9 Ϯ 0.5%) ( Table I). Identical Q max properties were found when affinitypurified Fv was compared with an equal optical density (278 nm) mixture of refolded V L and V H protein (Fig. 3). This suggested that each individually refolded domain protein had formed a dimerization competent structure in the absence of the other domain protein. Identical Q max values confirmed the similar active site environments displayed by SCA and Fv 4-4-20.
Affinity Measurements-Affinity-purified Fv 4-4-20 (liganded with fluorescein) was examined by dissociation rate fluorescence analysis. Fv 4-4-20 showed an affinity for fluorescein (3.5 ϫ 10 9 M Ϫ1 ) that was nearly identical to SCA (4.9 ϫ 10 9 M Ϫ1 ) within error limitation of the experiment (Table I). Similar affinity determinations were performed in the presence of excess polyclonal and monoclonal anti-metatype antibodies (reviewed in Refs. 28 and 54). These antibodies characteristically delay the fluorescein dissociation rate from the antibody active site against which they were raised. Both polyclonal and monoclonal anti-metatype reagents caused similar changes in the determined affinity values for fluorescein for SCA and Fv 4-4-20 (Table I). Cumulatively, binding data indicated that the Fv molecule effectively mimics the SCA, proving that the 212 linker peptide was not responsible for the original affinity decrease found in SCA as compared with mAb 4-4-20.
CD Spectra of Fv 4-4-20 - Fig. 4 shows the CD spectra recovered for Fv and mAb 4-4-20. Results are expressed in terms of mean residue weight ellipticity ([] ϫ 10 3 (degree cm 2 dmol Ϫ1 )). Analyses of CD spectra were carried out using previously computed CD spectra for poly-L-lysine containing varying amounts of ␣-helix, ␤-sheet, and random coil segments (65) as well as the previously determined CD spectra for SCA and mAb 4-4-20 (36). These analyses enabled estimation of the general secondary structure characteristics as a means of qualitative comparison of Fv with SCA and mAb 4-4-20. Fv 4-4-20 showed the same positive extremum (204 nm) and negative extrema (217 and 230 nm) as reported previously for SCA (Fig.  4). The Fv protein also displayed the slight shift in extrema, as well as the pronounced negative value at 230 nm, that SCA did in comparison with mAb 4-4-20. Similar CD spectra were recorded for samples of refolded V L and V H proteins (data not shown).
Pressure-induced Dissociation of Fluorescein from Fv 4-4-20 -Further structural comparison of Fv 4-4-20 with SCA was accomplished by measuring their hydrostatic-induced fluorescein dissociation parameters. Hydrostatic pressure has been shown to cause conformational changes (independent of protein tertiary structure) in proteins which can promote ligand dissociation (38,66).

DISCUSSION
In terms of structure-function relationships, recombinant Fv proteins have been invaluable tools for experimental studies of immunoglobulins. More recent endeavors involving these recombinant proteins have included their engineering with specialized effector functions for in vitro and in vivo immunodiagnostic and therapeutic roles. A common characteristic upon production of these diminutive antibody proteins is that their affinity for antigen is often reduced (or abrogated) as compared with the parental IgG. The reduced affinity has been attributed to changes in the active site structure or variable domain associative properties upon removal of the constant domains (3). If the initial decrease in Fv affinity for antigen was due to decreased domain-domain interactions, the properties governing stable variable domain association in relation to antigen binding must be identified. As such antibody proteins continue to be modified and applied to different systems (reviewed in Refs. 67 and 68), the nature of this affinity decrease, including how V L /V H affinity correlates with antigen binding affinity,  3.5 Ϯ .7 ϫ 10 9 M Ϫ1 K a (with polyclonal anti-metatype reagent) d 2.0 Ϯ 1.0 ϫ 10 10 M Ϫ1 3.0 Ϯ 1.1 ϫ 10 10 M Ϫ1 K a (with mAb 3A5-1 anti-metatype antibody) e 9.1 Ϯ 1.0 ϫ 10 9 M Ϫ1 6.3 Ϯ 0.8 ϫ 10 9 M Ϫ1 a Quenching of fluorescein fluorescence determined as described in . Briefly, antibody protein was added sequentially to a fluorescein solution and total fluorescence quenching monitored.
b Absorption maximum of fluorescein fluorescence determined as described in . Briefly, the absorption maximum wavelength for fluorescein was determined for a solution of 100% liganded fluorescein.
c Protein affinity for fluorescein determined as described in . Briefly, the dissociation of fluorescein from the active site was monitored after the addition of a large excess (Ͼ100-fold) of fluorescein amine. Affinity constant was determined from the dissociation rate using the previously determined association rate of 5 ϫ 10 6 M Ϫ1 s Ϫ1 for anti-fluorescyl antibodies.
d Affinity determination measurements were performed as above (Footnote c) with the addition of a large (Ͼ100-fold molar active site) excess of polyclonal anti-metatype antibodies.
e Affinity determination measurements were performed as above (Footnote c) with the addition of a large (Ͼ10-fold molar active site) excess of monoclonal anti-metatype antibody 3A5-1. must be defined and exploited. The well characterized 4-4-20/ fluorescein system presented an ideal method to study this phenomenon, based on the fact that SCA 4-4-20 exhibits a slight decrease in affinity for antigen compared with IgG (7). This study addressed the question by production and characterization of the Fv analogue of the 4-4-20 active site. These studies were based on the premise that comparative analysis provided clarification of the correlation between antibody constant domains, variable domain stability, and affinity for antigen.
Using similar expression conditions for SCA, purified Fv 4-4-20 demonstrated nearly identical anti-fluorescein activity as SCA ( Fig. 2A). Polyacrylamide gel analyses confirmed that the purified Fv protein contained only V L and V H domain proteins (data not shown). In terms of expression yield, E. coli cultures producing V L and V H protein consistently yielded 2-3fold more active Fv protein than similar cultures producing SCA upon refolding and affinity purification. The fact that improper disulfide bonds could not form between variable domain proteins during expression and refolding was most likely responsible for this result (4). In terms of idiotypy and metatypy, Fv 4-4-20 showed properties identical to SCA 4-4-20 when examined with polyclonal 4-4-20 variable domain-specific antibodies (Fig. 2, B and C). These results suggested that despite the dependence on noncovalent interactions for association, Fv 4-4-20 closely approximated the SCA molecule in terms of unliganded and liganded state structure.
Ligand binding affinities and ligand-related spectral measurements were made to assess Fv homology to the SCA molecule (in terms of the initial decrease in affinity for antigen). Such spectral measurements involving fluorescein/anti-fluorescein antibodies are characteristic of the specific anti-fluorescein active site environment which are relatively independent of affinity (69). Fv 4-4-20 showed almost identical ligand-related spectral properties (Q max and max ) and affinity for antigen relative to SCA (Table I). Anti-metatype antibodies, both polyclonal and monoclonal, characteristically enhance the affinity for fluorescein displayed by the anti-fluorescein active site for which they are specific (15,54). Fluorescein affinity measurements were repeated for Fv and SCA 4-4-20 in the presence of anti-metatype reagents to assess their relationship in terms of ligand binding kinetics and liganded state conformation. Fv and SCA showed similar (proportional) increases in affinity in the presence of anti-metatype antibodies, confirming that both active site structures possess the same conformational perturbations upon ligand binding (Table I).
In addition, CD analysis suggested identical overall secondary structures for Fv and SCA 4-4-20. of negative extrema at 217 nm, indicative of ␤-sheet structure, but also showed negative values at 204 nm, possibly due to a higher degree of random structure (65) (data not shown). The shoulder at 230 nm in the CD spectra of both V L and V H proteins was reduced compared with the Fv, suggesting a possible re-orientation of tryptophan and tyrosine side chains in their respective environments (37). This would indicate that isolated domain proteins undergo dynamic secondary structure rearrangement in order to dimerize and form active Fv protein.
To support this result, comparative fluorescein quenching studies were performed using affinity-purified Fv and associated V L and V H proteins. Associated protein showed almost identical fluorescein quenching properties as compared with an equal optical density solution of Fv (Fig. 3). It was also demonstrated that the liganded V L /V H dimers responded similarly to affinitypurified Fv when affinity measurements were determined in the presence of anti-metatype reagents (data not shown). Collectively, these results indicated that 1) Fv, V L , and V H proteins consisted of mostly ␤-sheet structure and some random coil, 2) upon V L and V H association some conformational changes are necessary for proper dimerization and active site formation, and 3) individually refolded domains maintain a dimerization competent form in the absence of constant domains which can form the proper active site environment for fluorescein binding and quenching.
As previously stated, hydrostatic pressure does not promote changes on the tertiary structure of proteins, but alters regions of secondary structure responsible for global protein conformation (38,66). A comparison of the pressure induced dissociation of fluorescein profiles for Fv and SCA would be a definitive evaluation of their dynamic similarity. Identical fluorescein fluorescence profiles were recovered for the two proteins when exposed to increasing hydrostatic pressure (Fig. 5). This indicated that Fv 4-4-20 displayed the same standard volume change (⌬ diss : Ϫ50 ml/mol) upon fluorescein dissociation as SCA (38). Seeing that their structures were apparently identical, this suggested that the Fv 4-4-20 must have increased conformational dynamics relative to the IgG molecule (⌬ diss : Ϫ5 ml/mol) as originally postulated for SCA (15,38,41). This indicated that increased dynamics were responsible for the decreased affinity for antigen displayed by Fv and SCA. Determination of the Fv interdomain dissociation constant (1.43 ϫ 10 Ϫ7 M) showed that despite the relatively low associative affinity, the high affinity fluorescein interaction was unchanged relative to the SCA (Fig. 6). This excluded the possibility that the initial decrease in the affinity for fluorescein upon removal of the constant domains was due to decreased domain-domain stability. The large difference between V L /V H and Fv/fluorescein K d values (ϳ400-fold) suggested that in terms of 4-4-20, there was little or no quantitative correlation between interdomain stability and antigen affinity. Seeing that individual variable domain proteins showed no affinity for antigen (data not shown), this confirmed that there was no coupling of fluorescein binding or domain association free energy in the formation of the Fv 4-4-20 (72). Thus, Fv structural characteristics responsible for interdomain association were independent of the structural features necessary for high affinity antigen binding. Collectively, results indicated that the absence of constant domains caused increased dynamic flexibility, not reduced variable domain associative affinity, in Fv and SCA 4-4-20 and resulted in decreased affinity for antigen.
Previous studies have demonstrated that heavy chain isotype (i.e. constant domain structure) influences antibody functional affinity against multivalent antigen (73,74). The effect of constant domains reported in these studies, which depended on high multivalent antigen concentrations, suggested that the change in functional affinity was due to change in segmental flexibility of the IgG molecule. Antibody isotype was, however, implicated in the expression of idiotopes on the variable domains of an anti-nitrophenyl antibody MOPC 315 (31). Idiotopes represent structural markers on the antibody active site which are sensitive to conformational fluctuations due to either ligand binding or natural protein dynamics (29,30,75). Such relationships would support the hypothesis that the interaction between the variable and first constant domains are necessary for proper variable domain conformational dynamics and not rigid structural features (Fig. 7). Results presented here support this hypothesis by demonstrating how the absence of constant domains influences active site/antigen interactions. In the case of 4-4-20, the binding of fluorescein can be considered a perturbation of the active site conformation which the constant domains can restrict to maintain the high affinity interaction. Removal of the constant domains from the SCA and Fv constructs resulted in the removal of this "dynamic buffering" effect. The ensuing increased domain dynamics translated into an increased dissociation rate of fluorescein from the active site. As studies progress on the re-engineering of antibody proteins, care must be taken to assess the importance of constant domain interactions for proper variable domain function. Methods which can both stabilize the active site structure and maintain wild type conformational dynamics may be necessary to ensure the success in producing recombinant Fv proteins which mimic parental IgG affinities.