OF BIOLOGICAL CHEMISTRY
`THE JOURNAL
`
`Val. 269, No. 28, Issue of July 15, pp. 18327-18331, 1994
`Printed in U.S.A.
`
`Improved Binding and Antitumor Activity of a Recombinant
`Anti-erbB2 Immunotoxin by Disulfide Stabilization
`of the Fv Fragment*
`
`(Received for publication, January 11, 1994, and in revised form, April 11, 1994)
`
`Yoram ReiterS, Ulrich Brinkmann, Sun-Hee Jung, Byungkook Lee, Philip G. Kasprzykg,
`C. Richter Kings, and Ira Pastanll
`From the Laboratory of Molecular Biology, Division of Cancer Biology, Diagnosis and Centers, National Cancer Institute,
`National Institutes of Health, Bethesda, Maryland 20892 and SOncologix Inc., Gaithersburg, Maryland 20878
`
`e23(dsFv)-PE38KDEL is a recombinant immunotoxin
`composed of the Fv region of anti-erbB2 monoclonal an-
`tibody e23 connected to a truncated form of Pseudomo-
`nus exotoxin (PE3SKDEL), in which the inherently un-
`stable Fv heterodimer (composed of V, and V,) is
`stabilized by a disulfide bond engineered between struc-
`turally conserved framework positions of V, and V,.
`We have now found that e23(dsFv)-PE38KDEL is con-
`siderably more cytotoxic to antigen-positive cell lines
`than the corresponding single-chain immunotoxin. The
`basis for the enhanced cytotoxic activity is that the e23
`dsFv-immunotoxin binds to erbB2 with greater affinity
`than the single-chain counterpart. The dsFv-immuno-
`toxin had 4-fold increased binding compared to the scFv
`and almost identical to the binding affinity of e23 Fab.
`e23(dsFv)-PE38KDEL was also considerably more stable
`at 37 “C than the single-chain immunotoxin.
`The therapeutic potential of the disulfide-stabilized
`immunotoxin was compared with its single-chain coun-
`terpart using two animal models of immunodeficient
`mice bearing subcutaneous tumor xenografts of human
`gastric tumor NS7 cells or human A431 epidermoid car-
`cinoma cells.
`The antitumor effect of e23(dsFv)-PE38KDEL was sig-
`nificantly better than that of the single-chain immuno-
`toxin. e23(dsFv)-PE3SKDEL caused complete regression
`of tumors at doses which caused no toxic effects in mice,
`whereas the single-chain immunotoxin did not cause
`complete regressions at the same doses.
`
`Until recently, this was the only available method to generally
`stabilize Fv fragments. In many cases such single-chain Fv
`fragments (scFv) retain the specificity and affinity of the anti-
`body. Single-chain Fv fragments have been successfully used
`for tumor imaging and as the basis for construction of multi-
`functional fusion proteins (4-8). Fv domains have been fused to
`toxins to make immunotoxins in which the Fv domain serves as
`a cell-targeting moiety for a potent toxin (5-8). Recombinant
`immunotoxins selectively bind to and kill cells that are recog-
`i.e. the Fv moiety. Several
`nized by the antigen binding domain,
`potent single-chain immunotoxins have been made that are
`specifically cytotoxic to antigen bearing tumor cells and cause
`complete or partial regression of human tumor xenografts in
`nude mice (7, 8). One example is e23(Fv)-PE38KDEL directed
`against the erbB2 proto-oncogene product (7). In these immu-
`notoxins, the scFv is connected to a truncated form of Pseudo-
`monas exotoxin (PE38KDEL) which contains the PE protein
`domains responsible for intracellular translocation, ADP-ribo-
`sylation activity, and a mutant carboxyl-terminal sequence,
`KDEL (10). These proteins lack the PE protein domain respon-
`sible for cell binding. Some scFv fragments, such as e23(Fv),
`have a lower affinity for antigen than the Fab counterpart. This
`lower affinity could result from the peptide linker somehow
`interfering with antigen binding or the linker may be unable to
`stabilize the Fv sufficiently.
`Recently, we identified an alternative method of stabilizing
`the Fv moiety in recombinant immunotoxins. In this approach,
`the Fv fragment is stabilized by a disulfide bond that is engi-
`neered between the framework regions of the two Fv domains
`and the toxin is fused to either of the Fv domains (11, 13). The
`appropriate disulfide positions were identified
`by molecular
`modeling techniques (12). These positions are in the conserved
`framework regions, distant from CDRs, and therefore are gen-
`erally applicable to many Fv fragments without affecting anti-
`gen binding. Several disulfide-stabilized Fv (dsFv) immunotox-
`ins have been made and tested, and they have about the same
`cytotoxicity as their single-chain counterparts and also similar
`antitumor activities (14, 15). However, the initial in vitro char-
`acterization of e23(dsFv)-PE38KDEL showed that it is more
`active than its single-chain counterpart, but the basis of this
`increased activity was not examined. We report here that the
`dsFv-immunotoxin has significantly improved binding and an-
`titumor activity compared with its single-chain counterpart.
`The dsFv-immunotoxin is also more stable in vitro. Our results
`suggest that in some cases disulfide stabilization of Fv frag-
`ments can improve binding and activity of Fv fragments.
`MATERIALS AND METHODS
`Construction of Plasmids for Expression of the dsFu-immunotoxin
`and Production of Recombinant Proteins: e23(dsFv)-PE38KDEL-The
`introduction of cysteines by site-directed mutagenesis in between e23
`18327
`
`Fv fragments of antibodies are heterodimers of the heavy
`chain variable domain (VJ’ and the light chain variable do-
`main (VJ They are the smallest functional modules of anti-
`bodies required for high affinity antigen binding. The polypep-
`joined by a
`tide chains of whole IgG or Fab fragments are
`disulfide bond. Fv fragments have no such interchain disulfide
`bridge and are therefore unstable (1-9). Stable Fv fragments
`can be produced by making recombinant molecules in which
`the V, and V, domains are connected by a peptide linker so that
`is regenerated in a single protein (2,3).
`the antigen binding site
`
`* The costs of publication of this article were defrayed in part by the
`payment of page charges. This article must therefore be hereby marked
`“aduertzsement” in accordance with 18 U.S.C. Section 1734 solely to
`indicate this fact.
`$ Supported by a grant from the Rothschild Foundation.
`7 To whom all correspondence should be sent. Tel.: 301-496-4797; Fax:
`301-402-1344.
`The abbreviations used are: V, variable; H, heavy; L, light; scFv and
`dsFv, single-chain and disulfide-stabilized
`Fv, respectively; CDR,
`complementary-determining region; PE, Pseudomonas exotoxin; MES,
`4-morpholineethanesulfonic acid; PBS, phosphate-buffered saline.
`
`IMMUNOGEN 2046, pg. 1
`Phigenix v. Immunogen
`IPR2014-00676
`
`

`

`18328
`
`Disulfide-stabilized Fu in a Recombinant Immunotoxin
`
`A L
`
`PE36KbEL
`SC(FV) immunotoxin
`
`ds(Fv) immunotoxin
`
`B
`pAK3 e23
`
`Nqel
`
`(VL) H e23 (VH) H PE38KDEL
`
`Ecoql
`f
`
`TABLE I
`Activity of e23(Fv)-PE38KDEL immunotoxins on various cell lines
`Cytotoxicity assays were performed by measuring incorporation of
`L3H1leucine into cell proteins as described (8). Data are given as ID,,
`values, the concentration of immunotoxin that causes a 50% inhibition
`of protein synthesis after a 24-h incubation with immunotoxin. Immu-
`notoxins tested were: single-chain e23(Fv)-PE38KDEL (scFv) and di-
`sulfide-stabilized e23(dsFv)-PE38KDEL (dsFv). ND, not determined.
`
`Cell line
`
`5 P e
`
`1%
`
`scFv
`
`dsFv
`
`-Fold
`Statistical
`
`
`increase significance
`
`Cytotoxicity
`
`V, and V, for stabilization of e23(Fv) and the amino acid sequence of
`e23(Fv) has been described previously (13). In the dsFv
`of e23(VH)
`and Glym of e23(VL) are changed to cysteines. The plasmids
`for the
`components of e23(dsFv)-PE38KDEL were made by subcloning as de-
`scribed (13).
`Recombinant proteins were produced from inclusion bodies as de-
`scribed previously (11, 13, 16). Properly folded disulfide-stabilized and
`single-chain immunotoxins were purified
`as described previously (8,
`11).
`cytotoxic activity of immunotoxins was
`Cytotoxicity Assays-Specific
`determined by inhibition of protein synthesis as described (8).
`
`Binding Assays-In binding assays 9-labeled e23 Fab was added to
`lo5 N87 cells as a tracer with various concentrations of the competitor.
`The binding assays were performed at 4 “C for 2 h in RPMI containing
`1% bovine serum albumin and 50 mM MES (Sigma) as described (7).
`Stability Assays-The
`stability of e23(dsFv)-PE38KDEL and
`e23(Fv)-PE38KDEL immunotoxins was determined by incubating them
`at 10 pg/ml at 37 “C in PBS. Active immunotoxin remaining after 5 h of
`incubation was determined by cytotoxicity assays on N87 cells.
`Antitumor Activity of dsFu-immunotoxin in Nude Mice-Antitumor
`activity of e23(dsFv)-PE38KDEL was determined in nude mice bearing
`human gastric (N87) or epidermoid carcinoma (A431) tumors. N87 tu-
`mor cells (5 x lo6) were injected subcutaneously on day 0 into immu-
`nodeficient “nude” mice. Tumors about 5 x 5 mm in size developed in all
`FIG. 1. Disulfide connection between V, and V, of recombinant
`animals by day 7. Starting on day 7 after tumor implantation, animals
`immunotoxin and plasmids for expression of dsFv-immuno-
`were treated with intravenous injections of e23(dsFv)-PE38KDEL or
`toxin. A, design of a disulfide-stabilized Fv-immunotoxin. Positions of
`e23(Fv)-PE38KDEL. Therapy was given every other day. For A431 tu-
`mors cells (3 x lo6) were injected and tumors (5 x 5 mm) were developed
`cysteine replacement in framework region of e23(Fv) are
`+ Cys in
`V, and Glyg9 + Cys in V, as described previously (13). B , pAK3 codes for
`by day 4. Starting on day 4 animals were treated with immunotoxins.
`the scFv immunotoxin e23(Fv)-PE38KDEL and is the parent plasmid
`Each treatment group consisted of five animals. Tumors were measured
`for the generation of plasmids encoding the components of the disulfide-
`with a caliper every 2 days and the volume of the tumor was calculated
`stabilized immunotoxin (7). In this molecule, the V, and VL domains of
`by using the formula: tumor volume (in mm3) = length x (width)’ x 0.4.
`monoclonal antibody e23 are held together by a peptide linker (Gly4 +
`
`Ser)3 and then fused to the PE38KDEL gene encoding the translocation
`and ADP-ribosylation domains of PE. pYR39 and pYR40 encoding
`RESULTS
`
`e23(VH C ~ S ~ ~ ) - P E ~ ~ K D E L and e23(VL Cys’’) are the expression plas-
`the Disulfide-stabilized
`Improved Cytotoxic Activity of
`the components of the dsFv-immunotoxin e23(dsFv)-
`mids for
`Immunotoxin-The goal of this study is to characterize the
`PE38KDEL and are derived from pAK3 by site-directed mutagenesis
`binding and in vivo activity of a recombinant anti-erbB2 im-
`and subcloning as described (13).
`munotoxin, e23(dsFv)-PE38KDEL, in which the Fv fragment is
`stabilized by a designed disulfide bond (Ref. 13; Fig. L4). The
`e23(Fv)-immunotoxins are PE-derived immunotoxin, highly cy-
`totoxic agents whose specificity is mediated by specific binding
`of the Fv component to target cells and toxicity by PE38KDEL
`which is the translocation and ADP-ribosylation domains of
`PE. Therefore, the specific Fv-mediated cytotoxicity of dsFv
`and scFv could be used to compare directly the binding of these
`Fv derivatives. Analysis of the cytotoxic activity of the e23dsFv
`and scFv immunotoxins shows that they bind to and kill the
`same spectrum of antigen-positive cells, but do not bind and kill
`antigen-negative cells. We found the dsFv-immunotoxin to be
`3-10-fold more active than the scFv-immunotoxin, depending
`on the cell line used using immunotoxin preparations with the
`same punty (13). Repeated determinations of the activity of the
`dsFv-immunotoxin on N87 and A431 cells indicate a statisti-
`cally different activity (p < 0.0001) (Table I).
`Improved Binding of the dsFv-immunotoxin-To determine
`whether the improved cytotoxic activity of the dsFv-immuno-
`toxin is due to improved binding, a competitive binding analy-
`sis was performed using lZ5I-labeled e23 Fab as a tracer and
`increasing concentrations of the e23dsFv- or scFv-immuno-
`toxin. Fig. 2 shows that e23(dsFv)-PE38KDEL binds to N87
`cells with 4-fold greater affinity compared with the single-chain
`immunotoxin. The single-chain immunotoxin binds
`with a
`4-fold lower affinity than the Fab
`(40 nM
`for e23(Fv)-
`PE38KDEL compared with 8 nM for e23 Fab) (Fig. 2). In con-
`trast, the
`binding of
`the dsFv-immunotoxin, e23(dsFv)-
`to that of e23 Fab (10 nM).
`PE38KDEL, is almost identical
`These results indicate that disulfide stabilization of e23 Fv
`significantly improves its binding and suggest that the peptide
`linker in the
`scFv somehow interferes with binding of the
`single-chain immunotoxin or the linker does not sufficiently
`stabilize the Fv heterodimer structure.
`
`have
`Improved Stability of the dsFu-immunotoxin-We
`shown previously for other dsFv-immunotoxins that disulfide
`stabilization of the Fv improves its stability in human serum
`and in buffers (11, 13). Fig. 3 shows a stability analysis of
`e23(Fv)-immunotoxins in which their binding and cytotoxic ac-
`tivity were tested on N87 cells after incubation in PBS for 5 h
`at 37 “C. It is evident that the single-chain immunotoxin is
`unstable at 37 “C, and almost all its cytotoxic activity is lost
`after 5 h in PBS. The loss of cytotoxicity is paralleled by a loss
`of binding activity (Fig. 3). In marked contrast, the dsFv-im-
`munotoxin, e23(dsFv)-PE38KDEL, is stable. Its cytotoxic activ-
`ity is reduced only slightly and its binding to N87 cells is almost
`unchanged after 5 h incubation in PBS.
`Improved Antitumor Activity of e23(dsFu)-PE38KDEL-To
`determine whether the improved binding and cytotoxic activity
`in vitro is accompanied by an increase in in vivo activity, we
`
`Breast
`HTB2O
`N87
`
`Gastric 0.3
`
`
`HEPG-2 Hepatic 1.2 0.3
`
`
`
`A431 Epidermoid
`KB3-1
`
`0.4
`
`nglml
`0.07
`0.06
`
`1.0 10.0
`
`10
`Epidermoid >lo00 >lo00
`
`5.7
`5.0
`4.0
`
`p < 0.0001
`p < 0.0001
`ND
`p < 0.0001
`ND
`
`IMMUNOGEN 2046, pg. 2
`Phigenix v. Immunogen
`IPR2014-00676
`
`

`

`
`
`
`
`Recombinant Disulfide-stabilized Fu in a Immunotoxin
`
`18329
`
`
`assayed the antitumor activity of the dsFv- and the scFv-im-
`munotoxin against two different models of tumor xenografts in
`nude mice. The first model is one in which N87 human gastric
`tumors are grown in nude mice. To obtain tumors, N87 cells (5
`x lo6) were injected
`subcutaneously into 4-6-week-old 20-g
`mice (day 0). Treatment was started
`on day 7 after tumor
`implantation when tumors measured about 50-70 mm'. Ani-
`mals were treated intravenously on days 7,9, and 11. Shown in
`Fig. 4 is the antitumor activity of the disulfide-stabilized im-
`munotoxin (e23(dsFv)-PE38KDEL) and its single-chain coun-
`terpart (e23(Fv)-PE38KDEL). Compared with untreated mice,
`the treated mice showed significant dose-related tumor regres-
`sions. The antitumor activity of the dsFv-immunotoxin was
`significantly better than that of the scFv-immunotoxin when
`equivalent doses of immunotoxin were compared. The mean
`tumor volume in mice treated with the dsFv-immunotoxin was
`2-3-fold smaller than tumor size of the mice treated with the
`scFv-immunotoxin, depending on the dose used (Table 11). In
`this model, the tumors do not usually completely regress, but
`still it is clear that the dsFv-immunotoxin is more active.
`
`The other tumor model that we used was human epidermoid
`cancer cell line A431, which rapidly forms large tumors in nude
`mice (18) and toward which the dsFv-immunotoxin is 10-fold
`more active in vitro than the scFv-immunotoxin (ID5,, of 1 ng/ml
`for dsFv-immunotoxin and 10 ng/ml for scFv-immunotoxin,
`Table I). Accordingly, 3 x lo6 A431 cells were injected subcuta-
`neously in nude mice on day 0. Animals were treated intrave-
`nously on days 4 (tumor size 50-70 mm3), 6, and 8. As shown in
`Fig. 5, both immunotoxins have significant dose-related anti-
`tumor activity. However, whereas the
`scFv-immunotoxin,
`e23(Fv)-PE38KDEL, caused regression of A431 tumors at 1.5-
`or 2.5-pg dose levels, the dsFv-immunotoxin at the same dose
`levels caused complete remission of the tumors after treatment
`with both dose levels of e23(dsFv)-PE38KDEL. Furthermore,
`the tumors treated with the
`scFv-immunotoxin regrew after
`treatment while cures lasting a t least 2 months were observed
`in 9 out of 10 animals treated with the
`dsFv-immunotoxin.
`These results indicate that the dsFv-immunotoxin has signifi-
`cantly better antitumor
`activity compared with the single-
`chain immunotoxin. Thus, the improved binding and cytotoxic
`improved in vivo antitumor
`activity in vitro translates into
`activity.
`
`e23 Fab
`-0-
`-m-
`e23(Fv)-Toxin
`"c e23(dsFv)-Toxin
`
`8 0 0 0
`
`I n
`0
`
`4000 2ooj
`
`,
`
`, , _ , _ _
`
`. .
`1
`
`1WO
`
`1w
`10
`Competitor, nM
`FIG. 2. Binding of e23(Fv)-PE38KDEL and e23(dsFv)-
`PE38KDEL to N87 cells. Competitive binding analysis of the ability of
`purified e23(Fv)-PE38KDEL and e23(dsFv)-PE38KDEL to inhibit the
`binding of '251-labeled e23 Fab to cells overexpressing erbB2. The con-
`centration of competitor which caused 50% inhibition of the binding of
`lZ6I-e23Fab was 3 1 1 ~ for e23 IgG, 8 1 1 ~ for e23 Fab, 40 nM for e23
`single-chain Fv-immunotoxin, and 10 nM for disulfide-stabilized e23Fv-
`immunotoxin.
`
`DISCUSSION
`In this study, we have shown that a recombinant immuno-
`toxin composed of a disulfide-stabilized Fv fragment of mono-
`clonal antibody e23 and a truncated form of Pseudomonas exo-
`significant
`toxin is a very potent antitumor agent with
`improved characteristics in vitro and in vivo compared with its
`single-chain linker-stabilized counterpart. This study demon-
`strates for the first time that disulfide stabilization of an Fv can
`improve its activity compared with the linker-stabilized single-
`chain Fv. The disulfide stabilization of e23(Fv) in this recom-
`binant immunotoxin improves its binding, cytotoxic activity in
`vitro, stability in vitro, and most important its antitumor ac-
`tivity. In contrast to the corresponding single-chain immuno-
`toxin, the disulfide-stabilized immunotoxin can cause complete
`regression of human tumor xenografts in nude mice.
`Improved Binding of e23(dsFv)-The
`disulfide-stabilized
`e23(Fv)-immunotoxin, e23(dsFv)-PE38KDEL, is directed to the
`proto-oncogene product erbB2.
`In this immunotoxin the Fv
`
`FIG. 3. Stability of e23(dsFv) and
`e23(Fv)-immunotoxins in buffered
`saline. The stability of e23(dsfi)-
`PE38KDEL and e23(Fv)-PE38KDEL was
`determined by incubating them for 5 h at
`immunotoxin re-
`37 "C in PBS. Active
`maining aRer incubation was determined
`by cytotoxicity assays on N87 cells. The
`binding of immunotoxins after incubation
`in PBS was determined by competitive
`binding analysis as described in the leg-
`end to Fig. 2.
`
`4000
`
`A. Binding
`
`e23(dsFv)-PE38KDEL
`
`loooi
`
`-0-
`
`Control(0hr)
`
`1
`
`.1
`
`1
`
`100
`
`1000
`
`1 0
`n M
`
`""
`
`3 0 0 0 -
`2 2 0 0 0 -
`
`0
`
`1000-
`
`) 1
`
`-o-
`Control (0 hr)
`"0- 5hrsPBS
`
`"1 \\
`
`I
`
`E 6 0
`
`E 0 4 ........ .-
`I=
`.01
`.1
`2
`=.
`g 100
`
`0
`
`B. Activity
`
`e23(dsFv)-PE38KDEL
`
`-
`
`-0-
`
`Control(0hr)
`5 h r s ~ ~ S
`
`......., ......
`100
`1 0
`ng/ml
`
`1
`
`1000
`
`.01
`
`.1
`
`1
`
`1 0
`ng/ml
`
`100 1000
`
`IMMUNOGEN 2046, pg. 3
`Phigenix v. Immunogen
`IPR2014-00676
`
`

`

`Disulfide-stabilized Fv in a Recombinant Irnrnunotoxin
`I
`
`e23(Fv)-PE38KDEL
`
`"t Control
`
`"-c 1.5 pg
`"f 2.5 pg
`Control
`
`18330
`
`600
`
`500
`
`400
`
`300
`
`200
`
`100
`
`0
`
`0
`
`
`
` 2 4
`
`6
`
`8
`
`1 0 1 2 1 4 1 6 1 8
`days
`
`e23(dsFv)-PE38KDEL
`
`0.5 pg -
`1 Irg -
`
`"c Control
`-0-
`
`2 pg
`
`600 -
`500 -
`
`4 0 0 -
`
`0
`
`2 4
`
`6
`
`8
`
`1 0 1 2 1 4 1 6 1 8 2 0
`days
`FIG. 4. Effect of e23(dsFv)-PE38KDEL and e23(Fv)-PE38KDEL
`on the growth of N87 tumors in nude mice. Groups of five animals
`were injected with 5 x lo6 N87 cells and treated intravenously on days
`7, 9, and 11. Mice were given 0.5 pg (01, 1 pg (01, or 2 pg (0) of
`e23(dsFv)-PE38KDEL or e23(Fv)-PE38KDEL. Control groups received
`PBS, 0.2% human serum albumin (m). Error burs indicate S.E.
`
`TABLE I1
`Effect of e23(Fu)-PE38KDEL immunotoxins on the regression of
`N87 tumor xenografts in nude mice
`Tumor size (mean + S.E.) was measured on day 14-15 after tumor
`implantation in groups of five mice. The mice were treated with three
`doses given every other day of e23(Fv)-PE38KDEL (scFv) or e23(dsFv)-
`PE38KDEL (dsFv).
`
`0
`
`0 2 4
`
`6 8 1 0 1 2 1 4 1 6 1 8 2 0 2 2 2 4
`day
`FIG. 5. Antitumor effect of e23(dsFv)-PE38KDEL and e23(Fv)-
`PE38KDEL in nude mice bearing A431 tumors. Groups of five
`animals were injected with 3 x 106A431 cells. The animals were treated
`intravenously on days 4, 6, and 8 with 1.5 pg (0) or 2.5 pg (m) of
`e23(dsFv)-PE38KDEL or e23(Fv)-PE38KDEL. Control groups received
`PBS, 0.2% human serum albumin. Error burs indicate S.E.
`
`part. The enhanced activity of e23(dsFv)- over e23(scFv)-im-
`munotoxin is due to improved binding of the dsFv-immuno-
`toxin. The binding
`of e23(dsFv)-PE38KDEL is almost
`indistinguishable from the binding of e23 Fab and is improved
`4-fold when compared with the single-chain Fv.
`disulfide-
`Improved Stability of dsFv-immunotoxins-The
`stabilized e23(Fv)-immunotoxin is also more stable than the
`single-chain irnmunotoxin. Its binding and cytotoxic activity
`are almost fully retained after incubation in PBS
`at 37 "C,
`whereas the single-chain immunotoxin loses all its binding ca-
`pacity and cytotoxic activity after the same period of incuba-
`tion. We have shown this increased stability for all disulfide-
`stabilized Fv immunotoxins produced so far (13-15). The dsFv-
`immunotoxins are also more stable in human serum (14, 15).
`The reason for the increased stability is because dsFv-immu-
`notoxins have a decreased
`tendency to aggregate compared
`with single-chain immunotoxins (17). The increased stability of
`the dsFv-immunotoxin should facilitate large scale production
`and better handling of the molecule in clinical applications.
`Improved Antitumor Activity of e23(dsFv)-PE38KDEL-To
`compare the antitumor activity of the single-chain and disul-
`fide-stabilized form, we used two tumor models which allowed
`a quantitative comparison of e23(Fv)-PE38KDEL and
`e23(dsFv)-PE38KDEL. One model consists of N87 human gas-
`tric tumor xenografts. N87 cells have a gene amplification of
`the erbB2 gene and express a large amount of erbB2 antigen
`(7). We compared the extent of tumor regression by measuring
`mean tumor volume and found that the dsFv-immunotoxin is
`2-3-fold more potent in antitumor activity.
`In the A431 human epidermoid carcinoma model, we com-
`pared the ability of the dsFv- and scFv-immunotoxins to cause
`complete regressions of the tumors. We found that the scFv-
`immunotoxin causes significant tumor regression but not com-
`plete remissions. However, the disulfide-stabilized immuno-
`toxin caused complete remissions
`in all animals. Thus, the
`dsFv-immunotoxin is a more potent antitumor agent. The re-
`sponse of the two in vivo tumor models to the same immuno-
`
`e23(Fv)-PE38KDEL
`immunotoxin
`
`Exp. 1
`
`Exp. 2
`
`scFv
`dsFv
`scFv
`dsFv
`
`o pg
`
`Mean tumor size after treatment
`with indicated doses
`0.5 pg
`
`1 pg
`mm3
`480275 180+38 120+34 87231
`
`480275 120+15
`
`68+13 3126
`300265 170+24 73+36 29210
`300265 110+32 3 2 + 1 7 8 2 5
`
`2 pg
`
`fragment is stabilized by an interchain disulfide bond instead
`of the linker peptide used to stabilize single-chain Fv frag-
`ments. The engineered disulfide bond is at positions within the
`framework region of the Fv and thus is generally applicable for
`the stabilization of many Fv fragments, because they are in
`conserved regions of Fv fragments and are distant from CDRs
`such that antigen binding should be not affected (13). Some
`single-chain Fv fragments have a reduced afflnity for the an-
`tigen when compared with the corresponding IgGs or Fab
`fragment (7,9), probably because the peptide linker interferes
`with binding or the linker does not sufficiently stabilize the
`Fv heterodimer structure. The
`single-chain
`e23(Fv)-
`PE38KDEL is an example of this; its binding to erbB2 is re-
`duced 4-fold compared with the Fab fragment or whole anti-
`body (Fig, 2 and Ref. 7). In contrast, the disulfide-stabilized
`immunotoxin, e23(dsFv)-PE38KDEL, is 3-10-fold more active
`on various cell lines compared with its single-chain counter-
`
`IMMUNOGEN 2046, pg. 4
`Phigenix v. Immunogen
`IPR2014-00676
`
`

`

`
`
`
`
`Disulfide-stabilized Fv in Recombinant a Immunotoxin
`
`
`
`
`
`18331
`
`toxin was found to be different. The dsFv immunotoxin caused
`3. Bird, R. E., Hardman, K. D., Jacobson, J. W., Johnson, S., b u f m a n , B. M., h e ,
`S.-M., Lee, T., Pope, S. H., Riordan, G. S., and Withlow, M. (1988) Science
`complete regression of A431 tumors with IC,, of 1 ng/ml, but
`242,423-426
`not of N87 tumors (0.06 ndml
`Several factors may ac-
`4. Milenic, D. E., Yokota, T., Filpula, D. R., Finkelman, M. A. J., Dodd, S. W.,
`count for this result. N87 cells might shed antigen more than
`Wood, J. F., Whitlow, M., Snoy, P., and Schlom, J. (1991) Cancer Res. 61,
`6363-6371
`(which express low levels Of erbB2) and
`A431
`im-
`5, Pastan, I,, and FitzGerald, D, (1991) Science 254, 1173-1177
`munotoxin activity. Another possibility is poor penetration Of
`6. Chaudhary, V. K., Queen, C., Junghans, R. P., Waldmann, T. A,, FitzGerald, D.
`immunotoxin into N87 tumors, perhaps because of the rela-
`J., and Pastan, I. (1989) Nature 339,394-397
`tively poor vascula~zation of this tumor
`with ~ 4 3 1 7. Batra, J. K., bsprzyk, P. G., Bird, R. E., Pastan, I., and King, C. R. (1992)
`Proc. Natl. Acad. Sci. U. S. A. 89, 5867-5871
`tumors. we found no si@’ificant
`difference in the
`‘Ox-
`8. Brinkmann, U., Pai, L. H., FitzGerald, D. J., Willingham, M. C., and Pastan,
`icity of the scFv- and dsFv-immunotoxin. The LD,, of a single
`I. (1991) Proc. Natl. Acad. Sci. U. S. A. 88, 861G8620
`dose of immunotoxin was 0.175 mgflrg for the scFv-immuno-
`9. Glockshuher, R., Malia, M., Pfitzinger, I., and Pluckthun, A. (1990) Biochern-
`istry 29,1362-1367
`for the dsFv-immunotoxin (data not
`and o’2 mgkg
`‘Oxin
`10. Seetharam, S., Chaudhary, V. K., FitzGerald, D., and Pastan, I. (1991) J. Biol.
`shown). In summary, the results presented here indicate that
`Chem. 266, 17376-17381
`the improved binding of e23dsFv results in both increased cy-
`11. Brinkmann, U., Reiter, Y., Jung, S.-H., Lee, B. K., and Pastan, I. (1993) Proc.
`Natl. Acad. Sci. U. S. A. 90, 7538-7542
`activity in
`activity in
`and increased
`12. Jung, S.-H., Lee, B. K., and Pastan,
`I. (1994) Proteins Struct. Funct. Genet. 19,
`3 5 4 7
`13. Reiter, Y., Brinkmann, U., Kreitman, R. J., Jung, S.-H., Lee, B. K., and Pastan,
`I. (1994) Biochemistry 33, 5454-5459
`14. Rei% Y., Psi, L., Brinkmann, U., and Pastan, 1. (1994) Cancer Res. 54,
`2714-2718
`15. Reiter, Y., Kreitman, R. J. Brinkmann, U., and Pastan, I. (1994) Int. J. Cancer,
`in press
`16. Buchner, J., Pastan, I., and Brinkmann, U. (1992) Anal. Biochern. 206,263-
`270
`
`I. (1994) Protein Eng. 7,697-704
`18. Pai, L., FitzGerald, D. J., Willingham, M. C., and Pastan, I. (1991) Proc. Natl.
`Acad. Sci. U. S. A. 88, 33583362
`
`REFERENCES
`1. Yokota, T., Milenic, D. E., Whitlow, M., and Schlom, J. (1992) Cancer Res. 62,
`3402-3408
`2. Huston, J. S., Levinson, D., Mudgett-Hunter, M., Tai, M. S., Novotny, J.,
`Margolies, M. N., Ridge, R. J., Bruccoleri, R. E., Haber, H., Crea, R., and
`U. S. A. 86,5879-5883
`Oppermann, H. (1988) Proc. Natl. Acad. Sci.
`
`
`
`17. Reiter, Y., Brinkmann, U., Webher, K. O., Jung, S.-H., Lee, B. K., and Pastan,
`
`mice.
`
`thank E. Lovelace and A. Harris for cell cul-
`Acknowledgments-We
`ture assistance and p, bdryszak, A, Jackson, and J. E~~~~ for
`assistance.
`
`IMMUNOGEN 2046, pg. 5
`Phigenix v. Immunogen
`IPR2014-00676
`
`