www.nature.com/naturebiotechnology • APRIL 2003 • VOLUME 21 • nature biotechnology

Genetic engineering of an immunotoxin toeliminate pulmonary vascular leak in mice

Joan E. Smallshaw1, Victor Ghetie1,2, Jose Rizo3, John R. Fulmer1, Linda L. Trahan1,Maria-Ana Ghetie1,2, and Ellen S. Vitetta1,2*

Published online 10 March 2003; doi:10.1038/nbt800

Vascular leak syndrome is a major and often dose-limiting side effect of immunotoxins and cytokines. We pos-tulated that this syndrome is initiated by damage to vascular endothelial cells. Our earlier studies identified athree–amino acid motif that is shared by toxins, ribosome-inactivating proteins, and interleukin-2, all of whichcause this problem. We have now generated a panel of recombinant ricin A chains with mutations in thissequence or in amino acids flanking it in the three-dimensional structure. These have been evaluated aloneand as immunotoxins for activity, ability to induce pulmonary vascular leak in mice, pharmacokinetics, andactivity in tumor-xenografted mice. One mutant was comparable to the ricin A chain used before in all respectsexcept that it did not cause vascular leak at the same dose and, when used as an immunotoxin, was moreeffective in xenografted SCID mice.

RESEARCH ARTICLE

Immunotoxins are hybrid molecules consisting of monoclonal anti-bodies (mAbs) or other cell-binding ligands that are biochemically orgenetically linked to toxins, toxin subunits, or ribosome-inactivatingproteins (RIPs) from plants, fungi, or bacteria1. During the pasttwo decades, immunotoxins containing deglycosylated ricin toxinA chain (dgRTA) have been developed structurally, optimized forin vitro stability and activity, and evaluated in vivo in rodents, mon-keys, and humans1–3. Four mAbs linked to dgRTA have completedtesting in phase 1 trials in >250 patients with relapsed chemorefrac-tory lymphoma, myeloma, or graft-versus-host disease4–10 (P. Martinand E. Vitetta, unpublished data). These immunotoxins have shownno evidence of myelotoxicity or hepatotoxicity, but all have inducedvascular leak syndrome (VLS) as defined by hypoalbuminemia,weight gain, and in the most severe cases, pulmonary edema andhypotension4–8,10. In addition, they have induced myalgia and, in 3%of patients, rhabdomyolysis or aphasia have defined the maximumtolerated dose (MTD)4–7. These dose-limiting side effects may also berelated to VLS and result from muscle edema or edema in the cere-bral microvasculature, respectively. VLS has been a problem (albeitnot always dose limiting) with many immunotoxins thus far tested inhumans3, as well as with cytokines such as interleukin-2 (IL-2) (ref.11). Prior radiotherapy and the lack of circulating tumor cells pre-dispose patients treated with dgRTA-containing immunotoxins toVLS, whereas circulating tumor cells (CTCs) protect them8,10,12. WithdgRTA immunotoxins, despite this dose-limiting toxicity (DLT),clinical responses have been encouraging, and 15–30% of patientswith chemorefractory relapsed lymphoma who have been treatedwith our anti-CD22 (RFB4)-dgRTA have experienced objective par-tial or complete responses in phase 1 clinical trials2,13. It is likely thatresponse rates would be even higher in less advanced or minimalresidual disease. Hence, further development of dgRTA immunotox-ins in particular, and other immunotoxins and cytokines in general,would be greatly facilitated by the elimination of VLS.

Several years ago, we postulated that dgRTA immunotoxinsinduce VLS by damaging vascular endothelial cells (VECs). Indeed,studies using human umbilical vein VECs (HUVECs) demonstratedthat dgRTA or immunotoxins prepared with dgRTA damaged thesecells within 1 hour14, although the inhibition of protein synthesisrequired ≥4 hours. We therefore postulated that either another por-tion of the RTA molecule was responsible for inducing VEC damageor the active site had an additional activity. Working from thishypothesis, we considered the possibility that RTA and other mole-cules that cause VLS, including toxins, RIPs, and IL-2, might sharestructural motifs responsible for interfering with cell-cell and cell-matrix interactions and thereby damage human VECs15.

In comparing the published sequences of VLS-inducing toxins,RIPs, and IL-2, we identified an xAspy consensus motif, where xcould be Leu, Ile, Gly, or Val and y could be Val, Leu, or Ser15. In thecase of RTA and IL-2, molecular modeling suggested that thesemotifs were partially exposed on the surface of their respective mol-ecules15. Of interest, a similar motif is shared by viral disintegrins,which disrupt the function of integrins16. This led us to postulatethat RTA might be a disintegrin15. To determine whether this motifwas responsible for VEC damage, we generated short LeuAspVal(LDV)- or LeuAspLeu (LDL)-containing peptides from RTA or IL-2,respectively, attached these peptides to the mAb, RFB4, and studiedthe ability of these conjugates to bind to and damage HUVECs invitro and to damage mouse lung vasculature and human vasculaturein vivo15. The human vasculature was studied in a SCID mousexenograft model of human skin17. The constructs containing theLDV from RTA, but not an altered or deleted LDV sequence, dam-aged VECs in all three models of vascular damage15. These resultssupport the hypothesis that the VLS-inducing site does not requirethe enzymatically active site of RTA, even though, in the case of RTA,these sites are in very close proximity in the folded protein15. We fur-ther showed that the LDV site of RTA activated caspases and induced

1Cancer Immunobiology Center, 2Department of Microbiology, and 3Department of Biochemistry, University of Texas Southwestern Medical Center at Dallas,Dallas, TX 75390-8576. *Corresponding author (ellen.vitetta@utsouthwestern.edu).

387

©20

03 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://w

ww

.nat

ure

.co

m/n

atu

reb

iote

chn

olo

gy

apoptosis in VECs18. This is important because it has been reportedthat many toxins and immunotoxins both induce apoptosis andinhibit protein synthesis. The relationship between these two eventshas, however, been unclear. Our results suggest that different por-tions of the RTA molecule may be responsible for these two effects.

To generate an effective RTA-containing immunotoxin that doesnot cause VLS, we have determined whether the putative VLS-inducing region in RTA, including the LDV sequence or residuesadjacent to it, can be mutated to reduce or eliminate VLS withoutaffecting the activity of the enzymatic site.

ResultsExperimental strategy. We generated a panel of recombinant (r)RTA mutants with single-residue changes either in the LDVsequence or in one of two residues adjacent to this sequence in itsthree-dimensional structure (Arg48 or Asn97). Our panel ofmutants was evaluated and narrowed down by a series of tests toidentify the best candidates. The successful candidates had to meetthe following criteria: (i) high-yield production and long-termstability, (ii) full enzymatic activity in a cell-free assay, (iii) goodcytotoxicity as an immunotoxin, (iv) failure to induce pulmonaryvascular leak (PVL) in SCID mice when used as an immunotoxin atthe same dose as an immunotoxin containing dgRTA, (v) a half-maximal lethal dose (LD50) higher than that of the dgRTA-containingimmunotoxin, (vi) pharmacokinetics in mice that are similar tothose of the immunotoxin containing dgRTA, and (vii) improvedefficacy as compared to the dgRTA immunotoxin at equitoxic dosesin our SCID-Daudi tumor model.

The in vitro activity of rRTAs and immunotoxins prepared withmutant rRTAs. The plasmid containing wild-type rRTA (in vectorpKK223) was used as our starting material. We generated severalmutants with conservative changes in the LDV sequence, includingL74A, L74M, D75N, D75A, D75E, V76A, and V76M, as well as in sur-face residues adjacent to the LDV sequence and distal from the activesite, R48A and N97A (Fig. 1). Purified rRTA preparations were >90%pure (data not shown). The enzymatic activities of these mutantrRTAs, wild-type rRTA, and dgRTA were then analyzed in a cell-freerabbit reticulocyte assay19. They were also conjugated to the mouseIgG1 anti–human CD22 mAb RFB4 (ref. 20), and were again evaluat-ed in the cell-free reticulocyte assay after reduction. Finally, theimmunotoxins were tested on CD22+ Daudi cells in a standard invitro cytotoxicity assay21.

When tested in the reticulocyte assay, the wild-type rRTA anddgRTA had very similar activities, although the wild-type rRTA wasslightly more active (Table 1). After coupling to the RFB4 mAb, both

retained their activity in the reticulocyte assay, and theyhad the same relative activity in the Daudi cell cytotoxici-ty assay. Six RTA mutants, R48A, L74A, L74M, V76A,V76M, and N97A, were as active or more active thandgRTA in the reticulocyte assays. They retained this activ-ity after they were coupled to RFB4, reduced, and retest-ed. With the exception of L74A, the same immunotoxinswere highly active in the Daudi cell cytotoxicity assay.

In contrast to these six mutants, rRTA mutants con-taining D75A, D75E, and D75N were two- to nine-foldless active in the reticulocyte assay and >200-fold lessactive as immunotoxins in the Daudi cell cytotoxicityassay. This suggests that Asp75 may be particularly criti-cal for internalization, intracellular routing, or intracel-lular activity, in that immunotoxins containing rRTAswith mutations in Asp75 were less active as immunotox-ins in the cellular cytotoxicity assay than their rRTAs orimmunotoxins in the cell-free reticulocyte assay. In lightof these results, we selected the R48A, L74M, V76A,V76M, and N97A mutants for further testing.

The ability of immunotoxins containing RTAs toinduce PVL in vivo. Unlike humans, mice injected withRTA-containing immunotoxins do not manifest sys-temic VLS, but they do show PVL15,22. However, the dosethat causes ‘plateau’ levels of PVL is higher than the dosethat causes serious VLS in humans. This is most likelydue to the greater sensitivity of human as compared tomouse vasculature to RTA-mediated damage. We there-fore injected SCID mice with the immunotoxins pre-pared with the mutant rRTAs that ‘passed’ the Daudi cell

RESEARCH ARTICLE

nature biotechnology • VOLUME 21 • APRIL 2003 • www.nature.com/naturebiotechnology388

Table 1.The enzymatic activity of rRTAs and immunotoxins prepared withthese rRTAs

RTA Cell-free reticulocyte assaya Daudi cell cytotoxicity assayb

RTA before coupling RTA after coupling RFB4-RTAto RFB4 to RFB4

dgRTA (IC50) 1.7 ± 0.6 × 10–11 c 1.3 ± 0.5 × 10–11 d 8.7 ± 8.6 × 10–13 e

Fold reduction Fold reduction Fold reduction

WT 0.7 ± 0.6 0.3 ± 0.1 0.7 ± 0.6L74Af 0.8 ± 0.3 0.4 ± 0.2 18 ± 14L74M 0.6 ± 0.2 0.2 ± 0.1 1.3 ± 0.6D75A 1.6 ± 1.3 1.7 ± 1.1 210 ± 150D75E 2.5 ± 1.1 4.1 ± 3.4 390 ± 310D75N 8.8 ± 5.3 1.8 ± 0.5 480 ± 190V76A 2.7 ± 1.9 0.8 ± 0.5 6.1 ± 4.5V76M 1.0 ± 0.4 0.4 ± 0.1 2.0 ± 1.7R48A 3.8 ± 1.7 0.7 ± 0.2 3.4 ± 0.8N97A 0.8 ± 0.1 0.2 ± 0.1 0.9 ± 0.6

aA cell-free protein synthesis inhibition (reticulocyte) assay was used to measure the activity ofeach RTA preparation before and after conjugation to the antibody, RFB4. The values representthe activity of dgRTA (in M) and the fold-reduction in activity of each rRTA (numbers <1 = anincrease in activity).bA cytotoxicity assay of each conjugated RFB4-RTA preparation performed on CD22+ Daudicells. The values represent the activity of dgRTA (in molar) and the fold reduction in activity ofeach rRTA (numbers <1 signify an increase in activity).cResults of 12 experiments.dResults of 6 experiments.eResults of >20 experiments.fResults of 3–12 experiments per mutant.

Figure 1. Ribbon diagram of RTA. Ribbon representation of the X-raycrystallographic structure of RTA (PDB accession no. 1br5), with theactive-site residue side chains in white, the LDV motif in blue, and R48and N97 in orange. The model was generated using the INSIGHT IIprogram (Micron Separations).

©20

03 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://w

ww

.nat

ure

.co

m/n

atu

reb

iote

chn

olo

gy

RESEARCH ARTICLE

screening assay and monitored PVL23. When mice were injected with15 µg immunotoxin per gram body weight as earlier described15, theimmunotoxins containing wild-type rRTA, dgRTA, L74M, V76A, orV76M caused PVL (Fig. 2). In contrast, the immunotoxins contain-ing R48A or N97A did not induce PVL. In light of our previousobservations that prior radiotherapy can exacerbate VLS inpatients12, we pre-irradiated mice with 150 cGy before immunotoxintherapy but observed no PVL for the immunotoxins prepared witheither R48A or N97A (data not shown).

The therapeutic activity of RFB4-N97A versus RFB4-R48A inSCID-Daudi mice. Because the immunotoxins containing N97A andR48A had virtually full activity in vitro and did not induce PVL invivo, they were assessed in vivo as RFB4 immunotoxins in our SCIDmodel of disseminated Daudi lymphoma24. In these experiments,our usual dose of 5 µg/g was used and the therapeutic activities ofthe two immunotoxins were compared. The SCID-Daudi miceinjected with immunotoxins prepared with R48A or N97A had meanparalysis times (MPTs) of 60 and 72 days, respectively (Table 2).RFB4-R48A showed a statistically significant improvement overRFB4 but was inferior to RFB4-N97A. Hence, RFB4-N97A was cho-sen for further studies.

The LD50 of RFB4 immunotoxins. To compare the therapeuticactivity of RFB4-N97A to that of the immunotoxin used in patientspreviously (RFB4-dgRTA), we first had to determine the LD50 of thetwo immunotoxins so that we could treat mice with equitoxic doses.To this end, groups of five mice were injected with different doses of

the two immunotoxins and weighed daily. When mice lost 20% oftheir body weight they were killed because, in our experience, allwould go on to die. In these experiments, the LD50 values of RFB4-dgRTA and RFB4-N97A were 9–10 µg/g and 55 µg/g, respectively.

Pharmacokinetics of RFB4-dgRTA versus RFB4-N97A. We exam-ined the pharmacokinetics and in vivo stability of both RFB4-N97Aand RFB4-dgRTA, because differences could affect therapeutic activ-ity. The various pharmacokinetic parameters for immunotoxinsshowed no substantial differences (Table 3). Sera from these animalswas collected at the end of the experiment and examined by SDS-PAGE and autoradiography. The immunotoxins were intact withoutobvious aggregation after 7 days in the mice (data not shown).

The therapeutic activity of RFB4-dgRTA versus RFB4-N97A.Having determined that the original and the new immunotoxin hadsimilar activity in vitro and similar pharmacokinetics in vivo, we nextcompared their therapeutic activity in SCID-Daudi mice using anequitoxic dose of each immunotoxin. For RFB4-dgRTA, we used adose corresponding to the human MTD (based on mg/m2), that is,∼ 40% of the murine LD50. For RFB4-N97A we used a five-fold high-er dose, corresponding to its five-fold higher LD50. Mice were treatedas described24 and killed when they showed paralysis of thehindlimbs. RFB4-N97A was superior therapeutically to RFB4-dgRTA (with a P value of 0.036 using a log-rank test) (Fig. 3).

DiscussionImmunotoxins containing dgRTA have shown efficacy in mice andantitumor activity in phase 1 clinical trials in patients withHodgkin’s and non-Hodgkin’s lymphoma. Although VLS is not aproblem when lower doses of the dgRTA immunotoxins are admin-istered, it would be helpful to optimize the therapeutic index. Inaddition, because many therapeutic molecules (toxins, RIPs,cytokines) cause VLS, a solution to this problem might facilitate theclinical development of many other biological agents.

We have addressed the problem of immunotoxin-mediated VLS bydeveloping one in vitro model14,25 and two animal models of VECdamage and vascular damage15,17. We then identified a candidate con-sensus sequence in RTA, RIPs, and IL-2, and demonstrated that anti-body-conjugated peptides from RTA or IL-2 containing thissequence, but not peptides with deleted or altered sequences, inducedendothelial cell damage in vitro and vascular damage in vivo in the

www.nature.com/naturebiotechnology • APRIL 2003 • VOLUME 21 • nature biotechnology 389

Table 2.Therapeutic efficacy of RFB4-R48A and RFB4-N97A inSCID/Daudi micea

Log-rank testc

Treatment MPT± s.d.b vs. RFB4 vs. R48A

PBS 31.6 ± 3.1 – –RFB4 45.4 ± 1.0 – –RFB4-R48A 59.9 ± 10.2 0.0093 –RFB4-N97A 72.1 ± 14.0 0.0018 0.0405

aGroups of 10 SCID-Daudi mice were treated on days 1–4 after tumor cell injec-tion with a total of 5 µg/g of immunotoxin or RFB4 or an equal volume of PBS.bMean paralysis time of 10 mice per group.cP values based on comparing the survival curves of each group.

Figure 2. In vivo PVL of RFB4-RTA immunotoxins.The 125I-labeled albuminretention in the lungs of SCID mice treated with 15 µg/g of the various RTAimmunotoxins was determined.Values were normalized to PBS-treatedmice by weight for comparison (n = number of mice). P values wereobtained by a log-rank test.The P values for RFB4-R48A vs. RFB4-dgRTAand RFB4-N97A vs. RFB4-dgRTA were 0.001 and 0.004, respectively.

Figure 3. Effect of equitoxic doses of immunotoxins prepared with N97Avs. dgRTA on the survival of SCID-Daudi mice. Groups of nine femaleSCID mice (average body weight of 20 g) were inoculated i.v. with 1 × 107

Daudi cells on day 0 and injected i.p. on days 1–4 with a total dose of thefollowing: RFB4 (IgG) (�) (20 µg/g); RFB4-dgA (�) (4 µg/g); and RFB4-N97A (�) (20 µg/g). Mice were killed when paralysis of their hindlimbsoccurred. The survival curves using the two immunotoxins werecompared using a two-tailed log-rank test. The P value for RFB4-N97Avs. RFB4-dgRTA was 0.036.

©20

03 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://w

ww

.nat

ure

.co

m/n

atu

reb

iote

chn

olo

gy

two animal models15,17. From these experiments, we concluded thatthe noncontiguous active site of the RTA (which, in RTA, does notencompass LDV) is not required to damage VECs. In light of theseresults, we have now evaluated several LDV mutants of rRTA and twomutants with changes in single residues adjacent to the LDV site.

The RTA mutants contained changes that, when modeled usingexisting X-ray crystallographic structural data for dgRTA26, wouldnot be expected to affect the active site of the RTA. Immunotoxinsprepared with mutants containing alterations in Asp75 were enzy-matically active but did very poorly in the Daudi cytotoxicity assay.Hence, Asp75 may be involved in internalization, intracellular rout-ing, or intracellular stability of the RTA. Three mutants, L74M,V76A, and V76M, did well in our in vitro assays but induced PVL invivo. Importantly, we found that single amino acid changes in rRTAoutside LDV but within 10 Å in the three-dimensional structure(R48A or N97A) resulted in the expression of highly active rRTAsthat made effective immunotoxins in vitro. At 15 µg/g, neither RFB4-R48A nor RFB4-N97A caused PVL, whereas RFB4-dgRTA did. Inaddition, at the usual dose at which we treat SCID-Daudi mice withthe RFB4-dgRTA, RFB4-N97A was more effective than RFB4-R48Aand thus was selected for further study. As an immunotoxin, RFB4-N97A had an LD50 fivefold higher than that of RFB4-dgRTA. Whencompared to an equitoxic dose of RFB4-dgRTA in the SCID-Daudimice (40% of their respective LD50 values, which for RFB4-dgA is theMTD in humans), RFB4-N97A had improved therapeutic activity.

Taken together, these experiments suggest that N97A should makeeffective, less toxic immunotoxins in humans. On the other hand, areduction in VLS might also reduce the ability of immunotoxins pre-pared with N97A to extravasate into tumors. Hence, its ultimatevalue as an immunotoxin can only be evaluated in clinical trials inhumans. Just as importantly, similar mutations in other toxins, RIPs,and VLS-inducing cytokines might improve the therapeutic activityof these molecules as well.

Experimental protocolPlasmids and mutagenesis. The pKK223 plasmid with wild-type rRTA

under isopropyl-β-D-thiogalactopyranoside (IPTG)-inducible control wasprovided by J. Michael Lord (Department of Biology Sciences, University ofWarwick, Coventry, UK)27,28. All DNA manipulations were done using stan-dard techniques29. Mutations were introduced using QuikChange (Stratagene,LaJolla, CA) and pairs of mutagenic primer as follows (only one primer percomplementary pair is listed, with mutation(s) shown in boldface type):

R48A, 5′-AGTGTTGCCAAACGCAGTTGGTTTGCCTATAAACC-3′;L74A, 5′-CTTTCTGTTACATTAGCCGCGGATGTCACCAATGCATATG-3′;L74M, 5′-GCTTTCTGTTACATTAGCCATGGATGTCACCAATGC-3′;D75A, 5′-GTTACATTAGCCCTGGCTGTCACCAATGCATATG-3′;D75E, 5′-CTGTTACATTAGCCCTGGAAGTCACCAATGCATATG-3′;D75N, 5′-CTGTTACATTAGCCCTGAACGTCACCAATGCATATGTGG-3′;V76A, 5′-GTTACATTAGCCCTGGATGCTACCAATGCATATGTGGTC-3′;V76M, 5′-GTTACATTAGCCCTGGATATGACCAATGCATATGTGGTC-3′;N97A, 5′-TCTTTCATCCTGACGCTCAGGAAGATGCAGAAGC-3′.

Expression of rRTA in Escherichia coli. Overnight cultures of E. coli strainXL1-Blue transformed with pKK223-rRTA grown in Terrific Broth29 con-taining 100 µg/ml ampicillin were used to inoculate 500 ml of the same

medium (in 2 l flasks) or a 5 l fermenter (New BrunswickScientific, Edmon, NJ). Cultures were shaken and aerated at 37 °Cto an OD600 of 0.6–0.8, at which time the temperature wasreduced to 30 °C; cultures were then grown to an OD600 of 1.0.Expression was induced using 1.0 mM IPTG and the culturesgrown overnight (∼ 15 h). Harvested cells were resuspended in2–3 ml PBS (50 mM PBS, pH 7.0) per gram of cell paste, andlysed by sonication (six 30 s bursts). Cell debris was removed bycentrifugation at 27,000g for 30 min; supernatants were filtered(0.45 µm) and stored at –20 °C until purification.

Radioimmunoassay (RIA) of expressed rRTA. The yield ofexpressed rRTA was assayed using a solid-phase RIA as

described5. The amounts of rRTA in the extracts were determined from thestandard curve using purified wild-type rRTA.

Purification of rRTAs. rRTAs were purified from the lysates by ion-exchangechromatography on CM-Sepharose fast flow (Pharmacia, Peapack, NJ) in10 mM sodium phosphate buffer at pH 6.5. The proteins were eluted using a0–300 mM NaCl gradient. Pooled fractions composing the main proteinpeak typically contained 50–80% rRTA. This pool was further purified bychromatography on Blue Sepharose CL-4B; bound rRTA was eluted with 1 MNaCl. The rRTA preparations were concentrated to 3–4 mg/ml, analyzed bySDS-PAGE, and stored at –20 °C in 50% glycerol. The enzymatic activity ofwild-type rRTA, dgRTA, and the mutant rRTAs was determined using a cell-free rabbit reticulocyte assay19.

Preparation of RFB4-RTA. The murine mAb RFB4 (anti–human CD22)was chemically conjugated to rRTAs and dgRTA using N-succinimidyloxy-carbonylmethyl-(2-pyridyldithio) toluene (SMPT; Pierce, Rockford, IL)and purified as described for dgRTA30,31. The enzymatic activity of the RTAin the immunotoxins was tested in the cell-free rabbit reticulocyte assay,after reduction19.

In vitro cytotoxicity assays. The cytotoxic activities of the different RFB4immunotoxins were determined using CD22+ Daudi cells and [3H]leucineincorporation as described21. Concentrations of immunotoxin that reduced[3H]leucine incorporation by 50% relative to untreated control culture weredefined as the IC50.

Radiolabeling of proteins. Albumin, RFB4, and immunotoxins were labeledusing Na125I as described32.

PVL. We have developed one in vitro and two in vivo models of VLS. The invitro model uses HUVECs14, and is excellent but not quantitative. The two invivo models involve extravasation of fluids (and 125I-labeled albumin) into thelungs of SCID mice (PVL) or into human skin xenografts in these mice15,17.Results using the three models correlated well. We now routinely use the PVLmodel because of its reproducibility, lower cost, and shorter time frame. Inthis model, SCID mice (Taconic, Germantown, NY) were injected intraperi-toneally (i.p.) with a total of 15 µg immunotoxin per gram body weight overthe course of 3 d. This is the plateau dose of an immunotoxin that routinelycauses PVL. Higher doses do not alter the amount of PVL, and lower dosesgive measurable but less PVL. After the final immunotoxin injection, the micewere injected intravenously (i.v.) with 5–7 µCi/mouse of 125I-labeled albu-min; 24 h later the 125I content of the right lung was measured and comparedwith a saline-treated control. Levels of radioactivity in the whole body andblood were determined before killing (data not shown).

Determination of the LD50. Groups of five mice were injected i.p. with dif-ferent doses of the immunotoxins prepared with either dgRTA or N97A.Mice were weighed daily and killed when they lost 20% of their body weight.The LD50 was calculated from the curves representing the percentage sur-vival versus the dose.

Therapeutic protocols. SCID mice with disseminated Daudi lymphoma weretreated as described earlier with each of the two immunotoxins, mAb, or PBS24.Mice were monitored and killed when hindlimb paralysis occurred. A compar-ison between the two survival curves was carried out using a log-rank test33,34.

Pharmacokinetics. SCID mice of 18–22 g were given 0.05% (vol/vol) Lugol’ssolution in sweetened drinking water throughout the experiment.Radiolabeled proteins were injected i.v. in the tail vein in a maximum volumeof 100 µl with a radioactive load of 1 × 107–5 × 107 c.p.m. and a dose of 5 µg.

RESEARCH ARTICLE

nature biotechnology • VOLUME 21 • APRIL 2003 • www.nature.com/naturebiotechnology390

Table 3. Pharmacokinetics in mice of immunotoxins constructed with rRTAs

RTA used in Number Time (h) AUCa FCRb (day–1) MRTc (h)the immunotoxin of mice (µg/h/ml)

dgRTA 5 57.9 ± 8.6 183.0 ± 36.1 0.300 ± 0.075 85.1 ± 15.3N97A 5 56.4 ± 1.3 191.8 ± 6.5 0.293 ± 0.006 78.2 ± 2.0

aArea under the curve.bFractional catabolic rate.cMean residence time.

©20

03 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://w

ww

.nat

ure

.co

m/n

atu

reb

iote

chn

olo

gy

RESEARCH ARTICLE

Whole-body radioactivity was counted immediately after injection and on adaily basis for 6 d using an AtomLab 100-dose calibrater (Atomic ProductsCorporation, Shirley, NY). Results were expressed relative to the initialwhole-body radioactivity (%). The pharmacokinetic parameters were deter-mined using a noncompartmental model with the PKCALC program35.

AcknowledgmentsWe thank Lien Le, Ming-Mei Liu, Yuen Chinn, Ana Firan, Steve Ruback, andStephanie Tuggle for exceptional technical assistance. We thank Shannon

Flowers and Linda Owens for secretarial assistance and M. Lord for providingthe rRTA clone. We are indebted to Jonathan Uhr and John Schindler for helpfulcomments concerning the manuscript. This work was supported by US NationalInstitutes of Health grant CA-77701 and a grant from the Higher EducationCoordinating Board of the state of Texas.

Competing interests statementThe authors declare that they have no competing financial interests.

Received 4 December 2002; accepted 16 January 2003

www.nature.com/naturebiotechnology • APRIL 2003 • VOLUME 21 • nature biotechnology 391

1. Farah, R.A., Clinchy, B., Herrera, L. & Vitetta, E.S. The development of monoclon-al antibodies for the therapy of cancer. Crit. Rev. Eukaryotic Gene Expr. 8,321–356 (1998).

2. Sausville, E.A. & Vitetta, E.S. Monoclonal Antibody-Based Therapy of Cancer(Marcel Dekker, Boston, 1997).

3. Frankel, A.E., Kreitman, R.J. & Sausville, E.A. Targeted toxins. Clin. Cancer Res.6, 326–334 (2000).

4. Amlot, P.L. et al. A phase I study of an anti-CD22-deglycosylated ricin A chainimmunotoxin in the treatment of B-cell lymphomas resistant to conventional thera-py. Blood 82, 2624–2633 (1993).

5. Vitetta, E.S. et al. A phase I immunotoxin trial in patients with B-cell lymphoma.Cancer Res. 51, 4052–4058 (1991).

6. Stone, M.J. et al. A Phase I study of bolus versus continuous infusion of the anti-CD19 immunotoxin, IgG-HD37-dgA, in patients with B-cell lymphoma. Blood 88,1188–1197 (1996).

7. Engert, A. et al. A Phase I study of an anti-CD25 ricin A-chain immunotoxin(RFT5- SMPT-dgA) in patients with refractory Hodgkin’s lymphoma. Blood 89,403–410 (1997).

8. Messmann, R.A. et al. A phase I study of combination therapy with immunotoxinsIgG-HD37-deglycosylated ricin A chain (dgA) and IgG-RFB4-dgA (Combotox) inpatients with refractory CD19+, CD22+ B-cell lymphoma. Clin. Cancer Res. 6,1302–1313 (2000).

9. Schnell, R. et al. A Phase I study with an anti-CD30 ricin A chain immunotoxin (Ki-4.dgA). Clin. Cancer Res. 8, 1779–1786 (2003).

10. Sausville, E.A. et al. Continuous infusion of the anti-CD22 immunotoxin IgG-RFB4-SMPT-dgA in patients with B-cell lymphoma: a phase I study. Blood 85,3457–3465 (1995).

11. Rosenberg, S.A. et al. A progress report on the treatment of 157 patients withadvanced cancer using lymphokine-activated killer cells and interleukin-2 or high-dose interleukin-2 alone. N. Engl. J. Med. 316, 889–897 (1987).

12. Schindler, J., Sausville, E., Messmann, R., Uhr, J.W. & Vitetta, E.S. The toxicity ofdeglycosylated ricin A chain-containing immunotoxins in patients with non-Hodgkin’s lymphoma is exacerbated by prior radiotherapy: a retrospective analy-sis of patients in five clinical trials. Clin. Cancer Res. 7, 255–258 (2001).

13. Senderowicz, A.M. et al. Complete sustained response of a refractory, post-trans-plantation, large B-cell lymphoma to an anti-CD22 immunotoxin. Ann. Intern. Med.126, 882–885 (1997).

14. Soler-Rodriguez, A.M., Ghetie, M.A., Oppenheimer-Marks, N., Uhr, J.W. & Vitetta,E.S. Ricin A-chain and ricin A-chain immunotoxins rapidly damage humanendothelial cells: implications for vascular leak syndrome. Exp. Cell Res. 206,227–234 (1993).

15. Baluna, R., Rizo, J., Gordon, B.E., Ghetie, V. & Vitetta, E.S. Evidence for a struc-tural motif in toxins and interleukin-2 that may be responsible for binding toendothelial cells and initiating vascular leak syndrome. Proc. Natl. Acad. Sci. USA96, 3957–3962 (1999).

16. Coulson, B.S., Londrigan, S.L. & Lee, D.J. Rotavirus contains integrin ligandsequences and a disintegrin-like domain that are implicated in virus entry intocells. Proc. Natl. Acad. Sci. USA 94, 5389–5394 (1997).

17. Baluna, R. & Vitetta, E.S. An in vivo model to study immunotoxin-induced vascularleak in human tissue. J. Immunother. 22, 41–47 (1999).

18. Baluna, R., Coleman, E., Jones, C., Ghetie, V. & Vitetta, E.S. The effect of a mono-clonal antibody coupled to ricin A chain-derived peptides on endothelial cells in vitro:insights into toxin-mediated vascular damage. Exp. Cell Res. 258, 417–424 (2000).

19. Press, O.W., Vitetta, E.S., Farr, A.G., Hansen, J.A. & Martin, P.J. Evaluation of ricinA-chain immunotoxins directed against human T cells. Cell. Immunol. 102, 10–20(1986).

20. Shen, G.-L. et al. Evaluation of four CD22 antibodies as ricin A chain-containingimmunotoxins for the in vivo therapy of human B-cell leukemias and lymphomas.Int. J. Cancer 42, 792–797 (1988).

21. Ghetie, M.A. et al. Antitumor activity of Fab′ and IgG-anti-CD22 immunotoxins indisseminated human B lymphomas grown in mice with severe combined immun-odeficiency disease: effect on tumor cells in extranodal sites. Cancer Res. 51,5876–5880 (1991).

22. Soler-Rodriguez, A.M., Uhr, J.W., Richardson, J. & Vitetta, E.S. The toxicity ofchemically deglycosylated ricin A-chain in mice. Int. J. Immunopharm. 14,281–291 (1992).

23. Rosenstein, M., Ettinghausen, S.E. & Rosenberg, S.A. Extravasation of intravas-cular fluid mediated by the systemic administration of recombinant interleukin-2.J. Immunol. 137, 1735–1742 (1986).

24. Ghetie, M.-A., Tucker, K., Richardson, J., Uhr, J. & Vitetta, E.S. The antitumoractivity of an anti-CD22 immunotoxin in SCID mice with disseminated Daudi lym-phoma is enhanced by either an anti-CD19 antibody or an anti-CD19 immunotox-in. Blood 80, 2315–2320 (1992).

25. Baluna, R., Ghetie, V., Oppenheimer-Marks, N. & Vitetta, E.S. Fibronectin inhibitsthe cytotoxic effect of ricin A chain on endothelial cells. Int. J. Immunopharm. 18,355–361 (1996).

26. Yan, X. et al. Structure-based identification of a ricin inhibitor. J. Mol. Biol. 266,1043–1049 (1997).

27. Simpson, J.C., Lord, J.M. & Roberts, L.M. Point mutations in the hydrophobic C-terminal region of ricin A chain indicate that Pro250 plays a key role in membranetranslocation. Eur. J. Biochem. 232, 458–463 (1995).

28. O’Hare, M. et al. Expression of ricin A chain in Escherichia coli. FEBS Lett. 216,73–78 (1987).

29. Sambrook, J., Maniatis, T. & Fritsch, E.F. Molecular Cloning: A Laboratory Manualedn. 2 (Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY, 1989).

30. Ghetie, V. et al. The GLP large scale preparation of immunotoxins containing deg-lycosylated ricin A chain and a hindered disulfide bond. J. Immunol. Methods 142,223–230 (1991).

31. Knowles, P.P. & Thorpe, P.E. Purification of immunotoxins containing ricin A-chainand abrin A-chain using Blue Sepharose CL-6B. Anal. Biochem. 160, 440–443(1987).

32. Fraker, P.J. & Speck, J.C.J. Protein and cell membrane iodinations with a sparing-ly soluble chloramide, 1,3,4,6-tetrachloro-3α,6α-diphenylglycoluril. Biochem.Biophys. Res. Commun. 80, 849–857 (1978).

33. Kalbfleisch, J.D. & Prentice, R.L. The Statistical Analysis of Failure Time Data(Wiley, New York, 1980).

34. Shah, S.A. et al. Anti-B4-blocked ricin immunotoxin shows therapeutic efficacy infour different SCID mouse tumor models. Cancer Res. 53, 1360–1367 (1993).

35. Shumaker, R.C. PKCALC: a basic interactive program for statistic and pharmaco-kinetic analysis of data. Drug Metab. Rev. 17, 331–348 (1986).

©20

03 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://w

ww

.nat

ure

.co

m/n

atu

reb

iote

chn

olo

gy