Identification of residues that stabilize the single-chain Fv of monoclonal antibodies B3.

B3(Fv)-PE38 is a recombinant single-chain immunotoxin in which the Fv portion of the B3 antibody in a single-chain form, which serves as the targeting moiety, is fused to PE38, a truncated form of Pseudomonas exotoxin A, which serves as the cytotoxic moiety. B3(Fv)-PE38 is specifically cytotoxic to many human cancer cell lines and is currently evaluated in a clinical trial. Monoclonal antibodies B3 (IgG1k) and B5 (IgMk) recognize related carbohydrate epitopes on human carcinoma cells. The Fv regions of these antibodies were previously cloned and expressed as the single-chain Fv-immunotoxins B3(Fv)-PE38 and B5(Fv)-PE38, respectively. The B3(Fv)-PE38 immunotoxin binds to antigen-positive cancer cells with a higher affinity than B5(Fv)-PE38 and is a more potent cytotoxic agent than B5(Fv)-PE38. However, it is less stable and rapidly aggregates upon incubation at 37°C. The VL domains of the two Fvs are very similar, differing by only three residues, the fourth and seventh Fr1 residues and the fifth CDR1 residue. The VH domains of the two Fvs vary considerably. To investigate whether any of the different VL residues may influence the stability of the B3(Fv), we constructed a chimeric immunotoxin containing the B3VH and the B5VL. This chimera had an improved stability and a higher apparent antigen binding affinity and cytotoxic activity when compared with B3(Fv)-PE38. Site-specific mutagenesis was used to show that the VL M4L mutation has an important role in stabilizing B3(Fv), although residues VL Ser-7 and VL Ile-28 also play a role in the increased stability. When tested in an in vivo model system, the chimera containing the B3VH and the B5VL had an improved antitumor activity in a human xenograft mouse model. These studies indicate that the common use of degenerate (“family-specific”) primers to clone Fv fragments may introduce destabilizing mutations.

B3(Fv)-PE38 is a recombinant single-chain immunotoxin in which the Fv portion of the B3 antibody in a single-chain form, which serves as the targeting moiety, is fused to PE38, a truncated form of Pseudomonas exotoxin A, which serves as the cytotoxic moiety. B3(Fv)-PE38 is specifically cytotoxic to many human cancer cell lines and is currently evaluated in a clinical trial. Monoclonal antibodies B3 (IgG1k) and B5 (IgMk) recognize related carbohydrate epitopes on human carcinoma cells. The Fv regions of these antibodies were previously cloned and expressed as the single-chain Fv-immunotoxins B3(Fv)-PE38 and B5(Fv)-PE38, respectively. The B3(Fv)-PE38 immunotoxin binds to antigen-positive cancer cells with a higher affinity than B5(Fv)-PE38 and is a more potent cytotoxic agent than B5(Fv)-PE38. However, it is less stable and rapidly aggregates upon incubation at 37°C. The V L domains of the two Fvs are very similar, differing by only three residues, the fourth and seventh Fr1 residues and the fifth CDR1 residue. The V H domains of the two Fvs vary considerably. To investigate whether any of the different V L residues may influence the stability of the B3(Fv), we constructed a chimeric immunotoxin containing the B3V H and the B5V L . This chimera had an improved stability and a higher apparent antigen binding affinity and cytotoxic activity when compared with B3(Fv)-PE38. Site-specific mutagenesis was used to show that the V L M4L mutation has an important role in stabilizing B3(Fv), although residues V L Ser-7 and V L Ile-28 also play a role in the increased stability. When tested in an in vivo model system, the chimera containing the B3V H and the B5V L had an improved antitumor activity in a human xenograft mouse model. These studies indicate that the common use of degenerate ("family-specific") primers to clone Fv fragments may introduce destabilizing mutations.
Monoclonal antibodies B3 and B5 are murine antibodies directed against Lewis Y -related carbohydrate antigens, which are abundant on the surface of many carcinomas (1). The B3 IgG or its fragments are currently used as the targeting moiety of immunotoxins that are being developed as anticancer agents. Both conventional whole IgG conjugates and singlechain recombinant immunotoxins have been prepared (1)(2)(3)(4). The single-chain Fv 1 immunotoxin of B3 is unstable at 37°C; it undergoes inactivation mainly by aggregation, especially upon incubation in PBS or in cell culture medium. In contrast, the B5(Fv)-PE38 immunotoxin is less susceptible to inactivation under those conditions, but it has lower apparent antigen binding affinity and cytotoxicity (5). We reasoned that we might be able to combine the advantages of each Fv by chimerization of their variable domains, since the Fvs of monoclonal antibody B3 and B5 bind the same carbohydrate antigen and are homologous in sequence (particularly in the V L domain, where 109 of 112 residues are identical). Therefore, we sought to gain insight on the possible involvement of individual residues of the light chain on the stability and the binding affinity of the B3(Fv). In this paper we have characterized the specific cytotoxicity, stability, and binding properties of single-chain Fv immunotoxins whose Fvs are B3-B5 chimeras or B3(Fv)s carrying point mutations in the V L domain. We found that the chimera with B3V H and B5V L was the most stable and potent of all the molecules tested in the in vitro assays, and we have also compared it with the parental B3(Fv)-PE38 molecule in an in vivo antitumor activity assay.

MATERIALS AND METHODS
Cloning of DNA Fragments Encoding the Heavy and Light Fv Segments of Monoclonal Antibodies B1 and B5-The B5 V H and V L DNA fragments were PCR-amplified independently, and the resulting PCR products were used as "primers" in a "domain shuffling" scheme where they replaced the corresponding B3 V H or V L or both, generating scFvtoxin expression plasmids having B3V H -B5V L , B5V H -B3V L , and B5Fv as described (5) (Fig. 1). Site-specific mutagenesis (6), with pULI7 uracil-containing single-stranded DNA as template and the oligodeoxynucleotide 5Ј AAGTAAACTGAATGGAG(T/A)TTGGGTCA(A/T)CA-GCACATCGCTTCC 3Ј as a primer was used to prepare plasmids expressing B3(Fv)PE38 derivatives in carrying V L mutations M4L and S7T in combination or separately.
Expression and Purification of Recombinant Proteins-Plasmids encoding B3(Fv)-PE38, B3V H -B5V L -PE38, B5V H -B3V L -PE38, or B5(Fv)-PE38 were expressed in Escherichia coli. The single-chain immunotoxins were obtained by solubilization and refolding of inclusion body protein as described (7). Properly refolded proteins were separated from contaminating proteins and aggregates by sequential ion exchange chromatography on Q-Sepharose and Mono Q (Pharmacia Biotech Inc.) followed by size exclusion chromatography on a TSK G3000SW (Toso-Haas) column.
Cytotoxicity of Recombinant Immunotoxins-The cytotoxic activity of immunotoxins was tested by determination of their ability to inhibit the incorporation of [ 3 H]leucine into cell protein, reflecting inhibition of protein synthesis in cultured cells as described (3).
Stability Assays-The stability of the immunotoxins in PBS at 37°C was determined by incubation at 0.2 mg/ml in PBS for 1, 2, or 4 h, followed by analytical chromatography on a TSK G3000SW (TosoHaas) column, to separate the monomers from the aggregates. Cytotoxic activities of aliquots of these immunotoxins were determined as described * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
above, and compared with the activities of the immunotoxins that were not incubated at 37°C.
The stability of the immunotoxins in human serum at 37°C was determined by incubation at 0.02 mg/ml in serum for 1, 2, or 4 h. Cytotoxic activities of aliquots of these immunotoxins were determined as described above and compared with the activities of the immunotoxins that were not incubated at 37°C.
Fluorescence Measurements-Solutions of immunotoxins at 10 g/ml were prepared in PBS containing 0, 0.5, 1, or 2 M urea. The intrinsic tryptophan fluorescence (8) (excitation at 295 nM) of each solution was determined at 23°C in a Perkin-Elmer LS50B luminescence spectrometer using a 1-cm path length.
Toxicity and Pharmacokinetics in Mice-Female BALB/c mice (6 -8 weeks old, ϳ20 g) were used for all the in vivo experiments. The single-dose mouse LD 50 was determined by a single intravenous injection of different doses of B3(Fv)-PE38 or B3V H -B5V L -PE38 diluted in 200 l of PBS-HSA. Mice were monitored for weight loss or death for at least 2 weeks postinjection. For pharmacokinetics the mice were given a single intravenous dose of 10 g B3(Fv)-PE38 or B3V H -B5V L -PE38 diluted in 200 l of PBS-HSA. Blood samples were collected at 2, 5, 10, 20, 30, 60, 120, and 240 min postinjection. The concentration of toxin was determined by incubating serial dilutions of the serum with A431 cells and measuring its ability to inhibit protein synthesis. A standard curve, obtained by incubating serial dilutions of the injected toxins on A431 cells, was used to determine the toxin concentration in each serum sample.
Antitumor Activity-Athymic (Nu/Nu) mice were injected subcutaneously on day 0 with 3 ϫ 10 6 A431 cells suspended in 0.2 ml of PBS. By day 4, tumors were about 30 -40 mm 3 in size. Mice were treated on days 4, 6, and 8 by intravenous injections of different doses of immunotoxins diluted in PBS-HSA. Tumors were measured with a caliper at 2-day intervals, and the tumor volumes were calculated using the following formula: volume ϭ length ϫ width 2 ϫ 0.4.

Cloning of the Variable Regions of B3 and B3-B5 Chimera as
Single-chain Immunotoxins-DNA encoding the variable regions of the heavy and light chains of B5 mRNA was prepared from B5 hybridoma cells as described (5). To generate singlechain immunotoxins in which the Fv of B5 in a single-chain form, is fused to PE38, the V H and V L fragments were PCRamplified using phosphorylated primers to enable the ligation of extended PCR products. The PCR products were purified and annealed to uracil-containing single-stranded DNA that was obtained by rescue of the phagemid pULI7, which encodes the single-chain immunotoxin B3(Fv)-PE38 with the helper phage M13KO7. The primer sequences, the extension of the templateprimer, ligation, transformation, and analysis of clones were described (5). This procedure resulted in the generation of plasmids for expression in E. coli in which either the V H or the V L domain of B3 was replaced by the corresponding domain from B5 ( Fig. 1), namely B3V H -B5V L -PE38 and B5V H -B3V L -PE38.
Expression, Purification, and Cytotoxic Activity of B3(Fv)-PE38 and B3-B5 Chimeras-Cultures of BL21(DE3) (9) transformed with each expression plasmid were used to produce immunotoxins. Following isopropyl-1-thio-␤-D-galactopyranoside induction, the overproduced proteins accumulated in inclusion bodies, which were isolated, and the recombinant protein was solubilized, reduced, and refolded as described previously (3). Active immunotoxins were recovered by ion exchange and size exclusion chromatography, as described under "Materials and Methods." Fig. 2 shows that the purified monomeric proteins were over 95% pure. Next, the cytotoxic activity of B3(Fv)-PE38 and of B3V H -B5V L -PE38 was assessed by measuring the incorporation of [ 3 H]leucine by various human carcinoma cell lines after treatment with serial dilutions of the immunotoxin as described previously (3). Activities of the immunotoxins were compared following 2 or 20 h of incubation on the cells. As shown in Fig. 3A, when tested on A431 cells, which strongly bind B3 IgG, B3(Fv)-PE38 has an IC 50 of 2.8 and 2.6 ng/ml following 2 or 20 h of incubation, respectively. B3V H -B5V L -PE38 has IC 50 values of 1.0 and 0.3 ng/ml following 2 and 20 h of incubation, respectively. B5(Fv)-PE38 has an IC 50 of 120 and 20 ng/ml following 2 or 20 h of incubation, respectively (5). B5V H -B3V L -PE38 also had low activity, with IC 50 values of 200 and 120 ng/ml following 2 or 20 h of incubation, respectively (data not shown). This data shows B3V H -B5V L -PE38 is the most active immunotoxin, and the fact that its activity increased 3-fold, from 1.0 to 0.3 ng/ml when the incubation with cells was extended from 2 to 20 h suggests it is quite stable.
To test the specificity of the immunotoxins, the same cytotoxic assay was repeated on additional cell lines, which differ in their level of B3 antigen expression (1, 10). As shown in Table  I, B3(Fv)-PE38, B3V H -B5V L -PE38, and B5(Fv)-PE38 had the same spectrum of recognition of the cancer cell lines tested, although they had different levels of specific cytotoxic activity. Also the cytotoxic activity of each correlates with its binding affinity.
Stability of Immunotoxins-The stabilities of B3(Fv)-PE38 and B3V H -B5V L -PE38 in PBS were tested by determination of their respective levels of aggregation and inactivation at 37°C as described under "Materials and Methods." As shown in Fig.  4 and Table II, both immunotoxins were mainly monomeric before the incubation. After 1 h of incubation in PBS at 37°C, about 30% of B3V H -B5V L -PE38 was aggregated, whereas B3(Fv)-PE38 was 78% aggregated. After 2 h of incubation in PBS at 37°C, 50% of B3V H -B5V L -PE38 was aggregated, whereas B3(Fv)-PE38 was about 90% aggregated. After 4 h of incubation in PBS at 37°C, B3V H -B5V L -PE38 was about 60% aggregated, whereas B3(Fv)-PE38 was about 95% aggregated. The aggregates were recovered in the volume excluded from the column and are multimeric forms larger than 300 kDa (Fig. 4). These aggregates have very little cytotoxic activity (data not shown). Fig. 5 shows results from cytotoxicity assays that were performed with aliquots of the immunotoxins that had been incubated in PBS as 37°C. Before treatment, B3(Fv)-PE38 had an IC 50 of 2.2 ng/ml, and it retained 23, 10, and 5% of its cytotoxic activity following 1, 2, and 4 h of incubation in PBS at 37°C, respectively. The B3V H -B5V L -PE38 chimera had an IC 50 of 0.4 ng/ml, and it retained 78, 56, and 35% of its cytotoxic activity following 1, 2, and 4 h, respectively, of incubation in PBS at 37°C. When incubated in human serum, B3(Fv)-PE38 retained 50, 25, and 12% of its cytotoxic activity following 1, 2, and 4 h at 37°C, respectively, and B3V H -B5V L -PE38 retained 80, 66, and 43% at these time points. Table II compares the percentage of monomeric immunotoxin with the percentage of cytotoxic activity remaining following incubation in PBS at 37°C as well as the percentage of active immunotoxin surviving after incubation in human serum at 37°C. It demonstrates that upon incubation in PBS, the residual cytotoxic activity correlates strongly with the relative amount of immunotoxin monomer that survived the 37°C incubation and that B3V H -B5V L -PE38 is more stable than B3(Fv)-PE38 with about a 4-fold longer half-life. The inactivation rate of the immunotoxins in human serum is slower than in PBS as we have observed previously with other immunotoxins (11). 2 However, the 4:1 ratio between the half-life of the more stable B3V H -B5V L -PE38 and the less stable B3(Fv)-PE38 is maintained.
Intrinsic tryptophan fluorescence was also used to determine the relative stability of B3(Fv)-PE38 and B3V H -B5V L -PE38 to denaturation by urea. In PBS, both immunotoxins have an emission peak at 330 nm and a second partially overlapping peak at 343 nm. The 330-nm peak remains unchanged up to 2 M urea and probably reflects the emission of the PE38 part of the molecule, which requires higher urea concentration for denaturation. 2 However, the peak at 343 nm increases with urea concentration. The relative fluorescence change is shown in Fig. 6; the peak at 343 nm increased more for B3(Fv)-PE38 2 I. Benhar and I. Pastan, unpublished observations. than for B3V H -B5V L -PE38, indicating that the former undergoes a conformational change, which is reflected by the fluorescence change under a lower denaturant concentration than the latter. Role of Specific Residues in the Light Chain Fv-Alignment of the amino acid sequences of the B3 and B5 V L domains (Fig.  7), reveals that they are very similar, with 109/112 identical residues; they differ only at positions 4 and 7, which are in Fr1 and at position 28, which is the fifth residue in CDR1. To identify the residue or residues that are responsible for the stabilization of the chimeric B3V H -B5V L -PE38 immunotoxin, we performed a molecular dissection in which we analyzed the properties of B3(Fv)-PE38 derivatives in which V L residues 4 and 7 were mutated to the corresponding B5 residues (Fig. 8).
We did not mutate the CDR residue to minimize the risk of altering the binding specificity of B3(Fv). Plasmids expressing B3(Fv)-PE38 derivatives carrying V L mutations M4L and S7T together or separately were prepared by site-specific mutagenesis using uracil-containing pULI7 single-stranded DNA as template and oligonucleotide primers. The plasmids were expressed in E. coli, and the immunotoxins were purified (Fig. 2).
As shown in Fig. 3B, B3(Fv)PE38 V L M4L/S7T was the most potent of the three, with IC 50 values of 1.2 and 0.6 ng/ml following 2 or 20 h, respectively, of incubation on A431 cells. Its cytotoxic activity was almost indistinguishable from that of the chimeric B3V H -B5V L -PE38 (Fig. 3A). B3(Fv)PE38 V L M4L had IC 50 values of 1.8 and 1.2 ng/ml following 2 and 20 h, respectively, of incubation on A431 cells. B3(Fv)PE38 V L S7T had the lowest activity, with IC 50 values of 6.0 and 5.0 ng/ml following 2 and 20 h, respectively, of incubation on A431 cells. Thus the immunotoxin with V L S7T was much less potent than the parental B3(Fv)-PE38 immunotoxin. This mutant was not analyzed further.
As shown in Figs. 4 and 5 and Table II, both the immunotoxin with the V L M4L/S7T mutation and the V L M4L mutation alone aggregated in PBS at 37°C to a lesser extent than B3(Fv)PE38 and were also more resistant to loss of cytotoxicity at 37°C than B3(Fv)-PE38. B3(Fv)-PE38 VL M4L/S7T was 40, 56, and 75% aggregated following 1, 2, and 4 h in PBS at 37°C, respectively, and retained 55, 28, and 15% of its cytotoxic activity at these time points. B3(Fv)-PE38 V L M4L was 55, 64, and 81% aggregated following 1, 2, and 4 h, respectively, in PBS at 37°C, and retained 50, 25, and 20% of its cytotoxic activity at these time points. When incubated in human serum, B3(Fv)-PE38 V L M4L/S7T retained 63, 38, and 33% of its cytotoxic activity following 1, 2, and 4 h at 37°C, respectively, and B3(Fv)-PE38 V L M4L retained 66, 45, and 33% of its activity at these time points. When the intrinsic tryptophan fluorescence of PE38 V L M4L/S7T and of PE38 V L M4L was determined, the spectra obtained were similar to the spectra obtained for B3(Fv)-PE38 and B3V H -B5V L -PE38, with a peak at 330 nm that does not change and a second peak at 343 nm that increased with increasing urea concentration. As shown in Fig. 8, the increases at 343 nm measured for B3(Fv)-PE38 V L M4L/S7T and B3(Fv)-PE38 V L M4L were similar to each other and were higher than the increase observed for B3V H -B5V L -PE38 but were lower than the increase observed for B3(Fv)-PE38, indicating that the sensitivity of PE38 V L M4L/S7T and PE38 V L M4L to undergo a conformational change (which is reflected by the fluorescence change) following exposure to urea is greater than that of B3V H -B5V L -PE38 but smaller than that of the parental molecule B3(Fv)-PE38. Taken together, these data show that while more stable than B3(Fv)-PE38 and having similar stabilities to each other, both mutated immunotoxins V L M4L/S7T and V L M4L are somewhat less stable than the chimeric B3V H -B5V L -PE38. These results further suggest that most of the stabilizing effect of B5V L on the Fv immunotoxin stems from its having a leucine rather than a methionine at position 4.
Toxicity and Pharmacokinetics of Immunotoxins-Since the in vitro assays demonstrated that B3V H -B5V L -PE38 was the most stable and most potent of the immunotoxins tested, it was chosen for further characterization in vivo, where it was compared with our prototype anti-Le Y immunotoxin B3(Fv)-PE38. The toxicity of the immunotoxins was evaluated by intravenous injections of different doses of immunotoxin into BALB/c mice. The mice were observed for 7-10 days postinjection. As shown in Table III, the LD 50 values of the immunotoxins were similar, with B3(Fv)-PE38 having an LD 50 of about 0.5 mg/kg and B3V H -B5V L -PE38 having an LD 50 of about 0.6 mg/kg. This value is in agreement with the published LD 50 of B3(Fv)-PE38 (0.5 mg/kg) (11). The pharmacokinetics of B3(Fv)-PE38 and B3V H -B5V L -PE38 were determined by measuring the immunotoxin levels in blood samples drawn from mice at various time points following an intravenous injection of each immunotoxin. The immunotoxin levels in the samples were determined by the ability of the serum or its dilutions to inhibit protein synthesis on A431 cells in comparison with standard curves obtained with the immunotoxins themselves. As shown in Fig. 9, both immunotoxins had a similar pharmacokinetic behavior when compared with each other, with a t1 ⁄2 of about 22 min. This value is in agreement with published pharmacokinetics of B3(Fv)-PE38 (11).
Antitumor Activity of Immunotoxins-The in vivo potency of the immunotoxins was evaluated by assessing their ability to cause regressions of established human carcinoma xenografts in nude mice. Tumors were induced by subcutaneous injection of 3 ϫ 10 6 A431 cells on day 0. Treatment was initiated on day 4 when the tumors averaged 30 -40 mm 3 in volume, and it consisted of three intravenous injections on days 4, 6, and 8 of various doses of immunotoxin. Control mice were treated with PBS-HSA only. As shown in Fig. 10, both immunotoxins had an antitumor effect that was dose-dependent. Mice treated with B3(Fv)-PE38 had complete regressions of the tumors at a dose of 0.1 mg/kg given intravenously every other day ϫ 3. At 0.05 mg/kg ϫ 3, the tumors showed complete regressions until day 16, but tumor growth resumed thereafter. In mice treated with 0.025 mg/kg ϫ 3, tumor growth was arrested for the duration of the treatment and resumed after its completion. This is in agreement with published data (11). B3V H -B5V L -PE38 had about 2-fold better antitumor activity; tumors showed complete regressions at doses of 0.1 and 0.05 mg/kg intravenously every other day ϫ 3. At 0.025 mg/kg intravenously, the tumors were in complete regression until day 14, and tumor growth resumed thereafter. In mice treated with 0.0125 mg/kg, tumor growth was arrested for the duration of the treatment and resumed after its completion. Control animals developed large tumors and were euthanized on day 18, when the tumor size was about 1.0 cm 3 . These data show that B3V H -B5V L -PE38 is a more potent immunotoxin in vivo because it has a 2-fold better therapeutic index. DISCUSSION We have cloned DNA fragments encoding the variable domains of the anticarbohydrate monoclonal antibodies B3 and B5 as single-chain Fv immunotoxins. We employed the method of "variable domain shuffling" (5), which allowed us to obtain Fv chimeras having V H and V L segments from the two different antibodies. The activities of the immunotoxins varied, with B3V H -B5V L -PE38 being the most potent, with an IC 50 of 0.3 ng/ml and B5V H -B3V L -PE38 being the least potent, with an IC 50 of 120 ng/ml following 20 h of incubation on A431 cells.  a M, percentage of monomeric immunotoxin remaining after incubation at 37°C for the indicated period was calculated by comparing the area under the monomer and aggregate peaks of the original fast protein liquid chromatography A 280 recorder graphs that were used to prepare Fig. 4. b % A: percentage of cytotoxic activity of immunotoxins aliquots incubated in PBS at 37°C for the indicated period (calculated from Fig. 5).
c % S, percentage of cytotoxic activity of immunotoxins aliquots incubated in human serum at 37°C for the indicated period.
The similar spectrum of recognition of cell lines that differ in the level of B3 antigen expression (Table I) suggests that there is no change in specificity following chimerization of B3 and B5. We have shown previously that the parental monoclonal antibodies recognize the same carbohydrate antigen (5). Since the B3V H -B5V L -PE38 chimera was the most stable and potent of the immunotoxins tested here, it was chosen for in vivo characterization in comparison with B3(Fv)-PE38. When tested in mice, the chimera did not differ significantly from the B3(Fv)-PE38 immunotoxin in pharmacokinetics or toxicity (Fig. 9, Table III). However, it had a better antitumor activity in a human xenograft nude mouse model (Fig. 10), where it was 2-fold more potent than B3(Fv)-PE38. The improved antitumor activity correlates with the 3-fold difference in cytotoxicity between the two immunotoxins when tested on A431 cells, which were also used to establish the tumors in the mice. The higher cytotoxic activities (lower IC 50 values) of the B5V H -B3V L -PE38 chimera and the B3(Fv)-PE38 V L M4L/S7T and B3(Fv)-PE38 V L M4L mutants and their higher apparent antigen binding affinity (not shown) may be explained by the fact that both are more stable than the wild type B3(Fv)-PE38. This improved stability is evident from their slower aggregation and loss of cytotoxic activity upon incubation in PBS or in human serum at 37°C; very little B3(Fv)-PE38 monomer survives a 4-h-long incubation, whereas the stabilized variants (the chimera and the mutants) retain significant cytotoxic activity. This accounts for the fact that there is very little difference in the cytotoxic activity of B3(Fv)-PE38 when incubated with A431 cells for 2 or 20 h, whereas with the stabilized variants there is a 2-3-fold increase upon a 20-h incubation with A431 cells. Furthermore, the intrinsic tryptophan fluorescence data (Fig. 6) provides additional evidence that B3(Fv)-PE38 undergoes a conformational change (unfolding; Ref. 8) under less severe chaotropic conditions than the chimera or the mutants, reflecting its inferior stability.
Site-specific mutagenesis followed by stability and cytotoxicity assays were used to identify which of the three V L residues that differ between B3 and B5 (Fig. 7) are responsible for the stabilizing effect. Since the B3V H -B5V L -PE38 and B3(Fv)-PE38 V L M4L/S7T (which differs from the chimera only at the fifth CDR1 residue) had similar characteristics in the cytotoxicity and stability assays, we conclude that the CDR residue does not play a major role in stability. Although, judging from the differences in stability between the B3V H -B5V L -PE38 chimera and the B3(Fv)-PE38 variants carrying B5V L residues at positions 4 and 7 or position 4 alone, the CDR residue may also contribute to the Fv immunotoxins' stability. Yasui et al. (13) have recently reported that mutations in CDR residues can influence the stability of Fvs analysis of B3(Fv)-PE38 derivatives carrying mutations V L M4L or V L S7T separately showed that replacing V L methionine 4 with leucine stabilized the immunotoxin almost as much as the B3V H -B5V L -PE38 combination, whereas replacing V L serine 7 with threonine had no stabilizing effect and was, in fact, less active than B3(Fv)-PE38 (Fig. 3).
Examination of a structural model of B3(Fv) (Fig. 8) (10), reveals that the side chain of V L serine 7 is exposed to the solvent and does not appear to interact directly with any other part of the molecule. The bulkier side chain of a threonine at the same position would probably also be exposed to solvent. V L methionine 4 is a buried residue (Fig. 7) as would probably be a leucine at the same position. Methionine 4 does not appear to interact directly with any of the V H residues, and the stabilizing effect resulting from its replacement with leucine would have to be explained by an effect on the independent folding of the V L domain. Creamer et. al. (14) have observed that there is a diminished entropy loss when leucine replaces methionine and is folded into the buried part of the protein. Eriksson et al. (15) have observed that replacing leucine at buried positions of lysozyme with other hydrophobic residues that have smaller side chains (including methionine) destabilized the protein.  Leucine is somewhat more hydrophobic than methionine, so it may be favored in a buried position (16 -18). In the studies cited above, there is a small but significant difference in the stability of the protein, and the methionine-containing variant is less stable than the leucine-containing variant (14,15). The high tendency of B3(Fv)s to aggregate may be improved by a small but favorable alteration in its refolding thermodynamics due to V L M4L replacement.
B3(Fv)-PE38 is less stable than B5(Fv)-PE38 (5), as were the B3V H -B5V L -PE38 chimera, B3(Fv)-PE38 V L M4L/S7T and B3(Fv)-PE38 V L M4L (not shown). It is thus clear that part of the stability difference results from differences in the V H domains. We have identified the B3V H residue, which is responsible for its relative instability. But that residue cannot be mutated without a severe loss in binding affinity, and this has important implications on the mechanism of antigen binding by the B3 antibody. 3 Finally, we wish to emphasize that the state of the art in cloning antibody fragments in E. coli is the use of "family-specific primers" for the PCR amplification of the variable region segments from cDNA (19 -22). Our data demonstrates that one must be cautious when such techniques are used, because the N termini of the obtained clones may not match the original protein sequence, and this alteration can lead to Fvs with altered stabilities and affinities.