0161-5890190$3.00 + 0.00 PergamonPressplc

Molecular Immunology, Vol. 21,No. 3,pp.273-282.1990 Printedin Great Britain.

IMMUNOTOXINS OF PSEUDOMONAS EXOTOXIN A (PE): EFFECT OF LINKAGE ON CONJUGATE YIELD, POTENCY, SELECTIVITY AND TOXICITY ALTON C. MORGAN JR,‘.~.~ GOWSALA SIVAM, PAUL BEAUMIER,ROBERT MCINTYRE, MIKE BJORN

and PAUL G. ABRAMS NeoRx Corporation,

410 West Harrison Street, Seattle, Washington 98119, U.S.A.

(First received 25 May 1989; accepted in revisedform

29 August 1989)

Abstract-Conjugates of monoclonal antibodies and Pseudomonas exotoxin A (PE) were formed with disulfide or thioether bonds. Thioether conjugates which formed with succinimidyl4-(N-maleimidomethyl)cyclohexane-I-carboxylate (SMCC) modified PE and reduced antibody formed with an 80% yield of equimolar conjugate within 30 min with an offering of one to one (toxin:antibody). The efficiency and kinetics of thioether formation were much higher with SMCC than with other maleimide reagents as well as more efficient than disulfide linkers. Thioether linkage resulted in immunotoxin consistently more potent and more selective in cifro than disulfide bonded conjugate. Thioether bonded conjugates also proved to have other favorable in vivo properties compared to disulfide conjugates: (1) a longer half-life in serum; (2) increased tumor localization; and (3) reduced toxicity. Toxicity of thioether linked holotoxin conjugates was directed at the liver hepatocyte but was easily monitored by serum liver enzymes. The conjugates are currently undergoing clinical evaluation for treatment of ovarian cancer with intraperitoneal administration. Research is ongoing to further decrease residual toxicity without reducing the potency of the conjugate.

lNTRODUCTlON Conjugates

of

offer considerable tumor

monoclonal potential

antibodies for localizing

with

et al., 1981, 1982, 1985; Hwang et al., 1983; Ramakrishnan and Houston, 1984). Although these conjugates contain the enzymatic (toxic) portion of the protein, they have less potency than the holotoxin. This appears to be due to the lack of the B-chain which facilitates insertion and translocation of the A-chain across membranes (McIntosh et al., 1983; Youle and Neville, 1982; Vittetta et al., 1984). An alternative strategy is to employ intact toxins (holotoxins) such as ricin or abrin or bacterial holotoxins like Pseudomonas exotoxin (PE) conjugated to monoclonal antibody (Thorpe et al., 1984; Fitzgerald et al., 1984). This results in agents of increased potency compared to the A-chain conjugates. This has been demonstrated in tlitro by comparing conjugates formed with the holotoxin PE or recombinant ricin A-chain and antibodies against both breast and ovarian cancer cell lines (Bjorn et al., 1985; Pirker et al., 198.5). Similarly, comparisons have been made with whole ricin and ricin A-chain conjugated to the same antibody (Vallera et al., 1984). The clinical application of holotoxin conjugates is, however, compromised by toxicity, since the conjugated holotoxin still retains some cell binding properties. Several avenues have been explored to reduce toxicity while retaining potency. Ricin B-chain binding can be inhibited by incubation with lactose or galactose or by covalent incorporation of these saccharides into the B-chain lectin site (Quinones et al., 1984; Houston, 1983). Alternatively, holotoxins, with conjugation to antibody, can have reduced toxicity due to steric occlusion of cell binding (Thorpe et al.,

toxins

therapy

to

sites while sparing

normal tissues of toxicity typically encountered with unconjugated cytotoxic agents. Molecules too toxic for use as free drugs can be delivered selectively to tumor cells with reduced toxicity as conjugates. Several strategies to create highly potent and selective immunotoxins have been explored. The most common approach is to link the protein inhibitory portion of plant toxins, namely A-chains or hemitoxins, or naturally occurring A-chain-like molecules called ribosomal inactivating proteins (RIPS), to monoclonal antibodies via disulfide linkages (Thorpe

‘To whom reprint requests should be addressed: Alton C. Morgan Jr. NeoRx Corporation. 22021-20th Avenue S.E., Bothell, WA 98021-4406. U.S.A. ‘Abbreviations used: RIP, ribosomal inactivating protein; PE, Pseudomonas exotoxin A; FPLC. fast protein liquid chromatography; SMCC, succinimidyl 4-(maleimidomethyl)cyclohexane-I-carboxylate; TAC, IL-2 receptor; PBS/BSA. phosphate buffered saline with 1% bovine serum albumin; DTT. dithiothreitol; PIP, paraiodohalf-life; ADP. adenosine diphenyl: T, ?. serum lethal dose 100; LDH, lactate phosphate; LD,,,, dehydrogenase: SGOT, serum glutamic oxalate transaminase: SGPT. serum glutamic pyruvic transaminase; DTNB, 5.5’-dithio-bis-(2-nitrobenzoic acid); MBS. m-maleimidobenzoyl-N-hydroxysuccinimide; SMPB. succinimidyl 4-( p-maleimidophenyl)butyrate. ‘University of Washington, School of Medicine. Department of Radiology. Division of Nuclear Medicine. 273

ALTON C. MORGAN

214

1984; Godal er al., 1986). In our hands, even the last approach produces conjugates with some antibodies which still have a high level of non-specific toxicity (Godal er al., 1986; Morgan et al., 1985). Non-conjugated PE is less toxic to cells than abrin. ricin or diphtheria toxin. In addition, conjugation of PE further decreases its toxicity. The reduction in toxicity, together with high in vitro potency (TD,,, 10 I(’to 10~ ” M) and selectivity (341ogs) when linked to antibody, indicates the potential of this agent for clinical use. However, when conjugated via a disulfide linkage. PE conjugates still exhibit lethal toxicity at relatively low doses (300 pglkg) when injected into primates (this publication). In this study, we examined holotoxin conjugates constructed with thioether linkages rather than conventional disulfide bonds. Contrary to prior reports, demonstrating reduced potency of toxin conjugates with stable linkages (Edwards rc ul.. 1983). thioether linked PE conjugates were equally potent on antigen positive cells and more selective than disulfide bonded conjugates. In addition, higher doses of the thioethcr conjugates could be safely administered to primates. Thus. potency was preserved and toxicity reduced simultaneously. The particular method of constructing the thioether immunoconjugatc was also shown to provide a markedly improved yield, thereby improving the eventual efficiency and cost effectiveness of therapy with these agents. \lATERIALS Monoclord

antibodies

AND METHODS

und fragmentation

Two murine IgG2a antibodies were used in these studies: anti-TAC (antibody recognizing IL-2 receptor, kindly provided by Dr Tom Waldmann, National Cancer Institute. Bethesda. Maryland) (Uchiyama et cd.. 1981a; Uchiyama c’t ul., 198lb): and 9.2.27. antibody recognizing a human melanomaassociated glycoprotein:proteoglycan (Morgan et (II., 1981). Three murine IgG2b antibodies were also used: NR-LU-10. directed to a pan-carcinoma antigen (Okabe rt cd., 1984). OVB-3 (Willingham t’r t/l.. 1987) and NR-ML-05. directed to a different epitope than 9.227 on the human melanoma associated glycoprotein:‘proteoglycan (Woodhouse et cd., 1990). F(ab)i fragments of 9.227 were prepared from whole antibody by digestion with immobilized pepsin. The digest was fractionated by ion exchange chromatography to remove peptides and undigested antibody and then concentrated by ultrafiltration. F(ab)i fragments were homogeneous as assessed by both SDSPAGE and FPLC gel filtration.

Disulfide bonded conjugates were produced according to previous published methodology (Pirker et (II., 1985) with some modification. Briefly. whole antibody and PE were reacted with 2-iminothiolane

Ja CI ul.

(2-IT) (molar ratio 5: 1) at pH 9.5 in sodium bicarbonate buffer (0.1 M. 0.15 M NaCI). Unreacted reagent was removed by gel filtration and derivatized antibody then reacted with DTNB and excess reagent removed. DTNB activated, derivatized antibody was then mixed with 2-IT treated PE at room temperature for up to 4 hr. The conjugate mixture was then fractionated by FPLC gel filtration on a TSK 3000 column in PBS pH 7.2 to remove unreacted PE. Conjugate corresponding in size to a I : I molar ratio of PE to antibody was pooled for subsequent testing. Thioether linked conjugates were prepared by first reacting PE with succinimidyl 4-(N-maleimidomethyl)cyclohexane-I-carboxylate (SMCC), succinimidyl 4-( p-maleimidophenyl)butyrate (SMPB). or m-maleimidobenzoyl-N-hydroxysuccinimide (MBS) at a molar ratio of 10 : I in pH 9.0 sodium bicarbonate buffer and unreacted reagent removed by gel filtration. Whole antibody was reacted with 25 mM dithiothreitol in 0.1 M phosphate buffered saline (PBS) pH 7.5 and excess reducing agent removed by gel filtration. The two conjugate components were then admixed and incubated at room temperature for up to two hours. Disulfide and thioether (SMCC) linked conjugates were also prepared with fragmented antibody. For this. the above procedure was followed except that F(ab)i w’as substituted for whole antibody. Under the same reducing conditions as ahovc. F(ab)i was reduced to F(ab)’ before conjugation. At least 2 batches of each type conjugate were produced with the batch size varying between I and 5 mg. In the case of OVB-3 conjugates the batch sire was 800 mg as this was produced for a clinical trial (Morgan ef ~1.. submitted for publication). Crude conjugate mixtures were first separated by FPLC gel filtration. Fractions corresponding to a I : I molar ratio of conjugate and unconjugatcd antibody were pooled. Unconjugdted antibody was removled by anion exchange chromatography on a Mono Q column (Pharmacia Fine Chemicals, Piscataway, NJ). Conjugate was eluted with a gradient of sodium phosphate starting with 5 mM sodium phosphate pH 7.6 and ending with 0.5 M sodium phosphate buffer pH 7.6 in 30 min (flow rate 0.5 mlmin). Biochemid

utu~l~~sis of’ conjugutr.r

Conjugate species separated by FPLC gel sieving or anion exchange chromatography were analyzed on SDS slab gels (10%). both reduced and non-reduced. Conjugates were also analyzed by analytical isolectric focusing using a pH 3-10 gradient (Pharmacia, Piscataway, NJ). Unmodified 9.227 focused in multiple bands with isoelectric points of 7.4. 7.3. 7.1. 6.9, 6.8 and 6.7 with an average pl of 7.2. Conjugate focused at an average pl of 7.46. Anti-TAC focused in a series of more basic isoelectric points, but following conjugation to PE focused at a pl similar to that of 9.2.27 conjugate. Both NR-LU-IO and NR-ML-05 had slightly acidic pI (6.57.05) and showed little change after conjugation to derivatized PE.

Linkage of PE immunotoxins Tumor cell lines and cytotoxicity

testing

In vitro cytotoxicity testing was performed as previously described using ‘H-leucine incorporation to measure surviving cells (Pavanasasivam et al., 1987; Morgan et al., 1987). 9.2.27 and NR-ML-05 (antimelanoma) conjugates were tested on the human melanoma cell lines, A375 met mix (A375 M/M) which was antigen positive, A375 primary which was antigen negative and the HT-29 colon carcinoma cell line which was also antigen negative. For anti-TAC (anti-IL-2 receptor) conjugates, HUT 102 cells were the antigen positive, and CEM the antigen negative targets, respectively (Fitzgerald et al., 1984). For NR-LU-10 (anti-carcinoma) conjugates, HT-29 was the relevant target and A375 M/M the irrelevant target. All cell lines were pretested for inherent sensitivity to unconjugated PE and showed similar sensitivity (ID,, = 100 ng/ml). ID,, values for conjugates on cells. The ID,, determinations represent the average of at least three separate determinations. Conjugates were tested in two formats, short exposure and long exposure. For short exposure, conjugate was incubated with cells for 1 hr at 37°C the cells gently washed and cultures continued for up to 72 hr before addition of 3H-leucine. For long exposure, cells were exposed to conjugate for the entire 72 hr of the culture period. Immunoreactivity PE conjugates of 9.2.27 and NR-ML-05 were compared to unconjugated antibody for binding to target A375 M/M melanoma using flow cytometry as previously described (Pavanasasivam et al., 1987). Similarly, conjugates of TAC were compared to unconjugated TAC antibody on HUT 102 cells and NR-LU-10 conjugate to unconjugated NR-LU-10 on HT-29 cells. OVB-3 antibody and conjugates were analyzed on ALAB breast carcinoma cells. Briefly, target cells were suspended in PBS/BSA at I x 105cells/ml and incubated at 4’C for 30 min with titrated doses of antibody or conjugate including saturating or subsaturating doses (1 to 0.01 pg/ml). Cells were then washed twice with PBS/BSA, and incubated with 100 ml (1 pg/ml) of fluorescein labeled goat anti-mouse IgG for 30min. Washed cells were then resuspended in PBS/BSA and analyzed for bound fluoresceinated secondary antibody using a cytofluorograph (Coulter Diagnostics, Hialeah, Florida). Immunoreactivity was assessed by comparison of the mean fluorescence index over 1000 channels for positive cells and then compared to a standard curve generated with fluoresceinated beads (Ortho Diagnostics) and expressed as fuorescein equivalents (FE). The percent of FE displayed by the conjugate was then expressed as a percent of the unmodified antibody at a subsaturation level. This type of assessment measures both changes in the percent of immunoreactive antibody molecules as well as alterations in affinity. MlMM 27,3--F

275

ADP ribosylation PE as well as disulfide or thioether linked PEconjugates were compared in a cell-free system as previously described (Vanness et al., 1980). ADPribosylation measures the ability of toxins such as Pseudomonas exotoxin A and diphtheria toxin to transfer labeled ADP-ribose to an intracellular acceptor molecule which in the case of PE is elongation factor-2. Thus, the assay is a measure of enzymatic activity not dependent on cell binding or translocation of the toxin as is required for cell killing. Both disulfide and thioether conjugates were titrated in the assay after treatment with 8 M urea and 1 M DTT. Biodistribution,

serum half-life

and toxicology

Thioether and disulfide linked conjugates of 9.2.27 were compared in a nude mouse xenograft model of human melanoma for tumor localization and biodistribution (Hwang et al., 1985). PE was first radiolabeled with “‘1-para-iodophenyl (PIP) (Wilbur et al., 1986). By this method, the radiolabel is not subject to dehalogenation and thereby more accurately reflects the biodistribution of conjugates. The labeled PE was then incorporated into conjugates and tested for retention of potency. Animals were administered 2-5 pg of labeled conjugate (2-5 PCi) intravenously and sacrificed at 20 hr post injection, organs blotted, weighed and counted as previously described (Wilbur et al., 1986), and per cent dose per gram calculated for each tissue. In addition, serum half-life of radiolabeled conjugate was estimated in non-tumor bearing Balb/c mice by retro-orbital sampling of whole blood (Wilbur et al., 1986). Groups of four mice were used for both biodistribution and serum half-life determinations. Toxicity of disulfide and thioether conjugates was assessed both in mice and in cynomogolus monkeys. Both the 9.2.27 and anti-TAC PE conjugates were assessed in mice; only anti-TAC conjugates were assessed in monkeys. In both cases, comparisons were made to the corresponding disulfide conjugate. Groups of five mice were observed for death or survival as a final endpoint, whereas monkeys were monitored by liver enzyme levels and observation of other relevant symptoms including changes in behavior, appetite, nausea/vomiting and temperature. Lactate dehydrogenase (LDH) levels proved to be the most sensitive monitor of hepatic toxicity in monkeys. LDH levels of treated monkeys were compared to averages obtained from normal monkeys tested at Microbiological Associates (Rockville, MD) over a period of years. RESULTS

Production

and analysis

qfconjugates

Representative FPLC gel filtration profiles of corresponding batches of disulfide and thioether linked conjugates of anti-TAC and PE are shown in Fig. 1.

216

ALTON C. MORGANJR et al.

Retention

time imln!

Fig. I. Comparison of Pseudomonas exotoxin immunoconiugates by FPLC gel filtration. Disulfide and throether linked conjugates were analyzed on a Superos; r 2 colimn at iml:min (Pharmacia). (A) Thioether reaction (SMCC). (B) Thioether reaction (SMPB). (C) Thioether reaction (MBS). CD), Disulfide reaction. Thioether‘reactio&‘&re analyzed 15 min post initiation. Disulfide reaction was analyzed after 4 hr. I, migration position of 1: I molar ratio conjugate; 2, migration position of unconjugated antibody; 3. migration position of unconjugated PE.

Disulfide linkage employing iminothiolane (Panel D) produced conjugate species of varying sizes reflecting a range of molar ratios as reported previously. SDS-PAGE analysis of these fraction pools revealed size ranges corresponding to 2: 1 and higher, and 1: I molar ratios of PE : antibody. In contrast. thioether linkage (Panels A and B) primarily formed conjugate in a single size range which corresponded by SDS-PAGE analysis to a I : 1 molar ratio of PE to antibody (peak 1). Thioether conjugation efficiencies varied according to the reagent used for PE derivatization (Panels A through C). SMCC derivatized PE reacted with reduced antibody resulted in the best yields. e.g. 80% 1 : 1 conjugate with an offering ratio of I : 1 (PE to antibody). In comparison. disulfide conjugation typically gave a 30% yield of 1 : 1 conjugate, with an offering of 10: 1 (PE to antibody). In addition, the best thioether conjugation required only a short incubation, with conjugation complete in 15 to 30 min whereas the disulfide linkage required greater than 4 hours for optimal yields. Interestingly, the elution position of 1: 1 conjugate differed between thioether and disulfide linked conjugate. One to one conjugate with disulfide linkage was well discriminated from unconjugated antibody (compare peaks 1 and 2, Panel D) while 1: 1 conjugate from thioether linkage migrated in the front edge of the antibody peak (compare peaks 1 and 2, Panel A). In fact, conjugation via thioether linkage was more easily monitored by disappearance of the unconjugated PE (peak 3) than formation of conjugate (compare Panel A, essentially complete conjugation, to Panel C. Forming essentially no conjugate formation). thioether linked conjugate with reduced antibody and SMCC-derivatized PE was the most efficient method for forming 1: I conjugate. The combination of iminothiolane-derivatized antibody and SMCCderivatized PE resulted in slower kinetics of formation as well as the formation of higher molecular weight conjugate (not shown). Due to the difficulty in separating thioether linked conjugate from unconjugated antibody by gel filtration, anion exchange chromatography was employed

to further purify conjugate containing fractions (Fig. 2). Conjugate but not antibody was bound to the Mono Q column under the conditions used and eluted with a hypertonic sodium phosphate buffer. Isolated 1: I ratio thioether (SMCC) and disulfide conjugate as analyzed on SDS-PAGE under reducing conditions verified the nature of the bond (Fig. 3). As expected, reduction of disulfide conjugates gave three bands corresponding to PE and antibody heavy and light chains. Reduction of thioether conjugate gave bands corresponding to covalent adducts of PE with heavy and light chains with most of the adducts with the heavy chain as well as free heavy and light chain presumably from unconjugated antibody half molecules. Free PE was not apparent following reduction of thioether conjugate.

Immunoreactivity of antibody was consistently reduced by conjugation with PE (Table I). The loss of immunoreactivity was found regardless of which antibody and which linkage was used. This contrasted to our previous experience with conjugation of A-chains or ribosomal inactivating proteins which did not alter immunoreactivity (Pavanasasivam er ~1.. 1987). Potency and selectivity of thioether and disulfide bonded conjugates were compared in vitro. Titration of ADP-ribosylating activity for disulfide and thioether conjugates of anti-TAC is shown in Fig. 4. In this cell-free system, PE requires unfolding by reducing and denaturing agents to expose the cofactor (NAO) binding site. Conjugates formed with the two types of linkages were equi-potent for enzymatic activity despite the demonstration by SDS--PAGE (above) that the thioether conjugate did not reduce to yield free PE. We also evaluated thioether and disulfide conjugates of 9.2.27 in an in vitro cytotoxicity assay (Table 2). 9.2.27 (whole antibody) thioether conjugate was consistently more potent (and more selective) than its disulfide counterpart. Similar results were obtained in comparing thioether and disulfide linked TAC conju-

Linkage of PE immunotoxins

0’

I

I

I

IO

I

20

277

I

1

30

Retention time (min) 54321

Fig. 2. Separation of PE conjugated from unconjugated antibody by FPLC anion exchange chromatography. Top: chromatographic profile of an exchange column of thioether linked conjugate (Panel A, Fig. 1). Dark, broad line represents the gradient used for elution. Numbers 1, 2 and 3 refer to pools of material. Bottom: SDS-PAGE analysis (non-reduced) of pools from anion exchange chromatography. Lane 1, unconjugated PE; Lane 2, starting material before anion exchange chromatography (after gel filtration); Lane 3, pool 1 from anion exchange column; Lane 4, pool 2 from anion exchange column; Lane 5, pool 3 from anion exchange column.

gates (not shown). Thioether conjugates formed with F(ab)’ were slightly more potent than those made with whole antibody but were considerably less selective. It was also obvious that the use of a thioether linkage enhanced selectivity of both the intact and F(ab)’ forms of conjugates. Thioether-linked conjugates were also evaluated for selectivity using relevant and irrelevant conjugates reciprocally tested on antigen positive and negative cell lines (Fig. 5). Conjugates of both NR-ML-05 (anti-melanoma) and NR-LU-10 (anti-carcinoma) were highly potent when tested on antigen positive targets, approximately IO to 100 fold more potent than unconjugated PE. In contrast, when tested on

antigen negative targets, conjugates had approximately 100 fold less toxicity than free PE. Both target cells were sensitive to the same extent to free PE. Thus, it appears that thioether linkage of PE to four different antibodies preserves the potency of the toxins in both cell-free as well as cytotoxicity assays. Moreover, thioether linkage seems to enhance selectivity, including those conjugates made with F(ab)’ and to be highly efficient in yielding 1 : 1 molar ratio conjugate. Biodistribution and toxicology Toxicity, tumor localization, biodistribution pharmacokinetics were compared for thioether

and and

278

ALTON C. MORGAN JK et trl

Fig. 3. Comparison of Pseudomonas exotoxin immunoconjugates. SDS PAGE analysis. Pools corresponding to fractions from FPLC separation of disulfide and thioether conjugates were analyzed on reducing IO% SDS slab gels after radiolaheling (whole conjugate) with “‘I-PIP. Left law = disuliide, right lane = thioether.

disulfide linked conjugates. Mice were administered different doses of anti-TAC and 9.2.27 conjugates intraperitoneally. Disulfide conjugates exhibited an average LD,,, of 50pggi’kg differing slightly depending on the batch. Thioether conjugates were reproducibly less toxic with an average LD,,,,, of 250 pgS,‘kg. Prior data in monkey studies with disulfidc conjugates (Fitzgerald. Willingham. Pastan. unpublished observations) had suggested that toxicity was due to direct hepatocellular damage. Thus. WC examined serial hepatic transaminase and lactic dehydrogenase

Table

I. Immunoreactiwty con,ugatea

Antibody form” OVB-3

OVB-3’PE

“Four

of thmethcr (OVB-3)

linked

PC

levels in monkeys administered PE conjugate (s(TAC). As shown in Table 3. LDH. the most sensitive indicator of hepatic toxicity. increased significantly after administration of only 300,~gjkg of disulfide linked conjugate. In addition, hepatic transaminases like SGOT and SGPT were also elevated (not shown). Animals also exhibited behavioral evidence of toxic,5-,

8 x F ijII

DUE (/lg,‘ml)

PUXE?lt pwtiw

Ft‘

5-

I 01 0.01 I 0.1 0.01

Y4 YY YX ‘)‘I YX ?I

x21 716 457 66: 656 ‘)v

o-

other anubodies (NR-LU-IO. anti-TAC. and NR-ML-05) conJugaled to PC showd simlar reduction in Immunorextlvlty “Percent of control at wbsatur;ttmn dew = 20.

-?

I -I

/v0

1U’ jga+eI oqJg/‘tlae

9.227

Fig. 4. Comparison of Pseudomonas exotoxin immunoconjugates. ADP rihosylatmn. l disulfide; 0 thioelher.

Linkage of PE immunotoxins Table 2. In airro potency and selectivity of disulfide linked conjugate (9.2.27)

or tbioether

Table

279

3. Hepatic

toxicity

of PE immunotoxins LDH values

Potency Type of linkage

Antibody

form

Dose

Disulfide-whole

Bond (Icg/W

vs Antigen vs Antigen

Positive Negatwe

8 x IO-” 4 x IO_‘0

1.5

vs Antigen vs Antigen

Positwe Negatwe

4 x IO_” 2 x IO-’

3.5

vs Antigen vs Antigen

Positive Negative

3 x 10~” I.6 x IO ”

0.22

vs Antigen YS Antigen

Positive Negative

3.5 x IO- I2 I x lo-‘0

2.3

Thioether-whole’

Disulfide-F(ab)’

Thioether-F(ab)’

“Potency (ID,, vs antigen positive cells as measured in the ‘H-leucine incorporation assay). ‘Difference, extrapolated from a graph, in ID,, vs antigen positive and antigen negative cell (logs). ‘Thioether conjugates using whole antibody were also evaluated for the four other antibodies: NR-LU-IO, OVB-3, TAC and NRML-05. Potency against antigen positive cells varied from 4 x IO-” to 4 x IO-“. Selectivity varied from 3 to 4.5 logs.

ity at the same dose diarrhea. By contrast, up conjugate produced with little change in

with loss of appetite,

nausea and

to 1 mg/kg of thioether linked only marginal elevation in LDH SGOT or SGPT. Animals exhib-

120IOO80 60 L z 2 ”

S-S s-s s-c

300” 600” 1000’

1

2

3

4

5

6

566 1725; 389

2090. 6000; 796*

4359’ ND 333

ND ND 344

715* ND 1500’

ND ND 386

“Lack of appetite, nausea, diarrhea. hLack of appetite, nausea, diarrhea, death. ‘No symptoms, conjugate administered on day 0 and day 4 (1 mg and 2 mg, respectively). *Values were 2 SD above the mean of normal, untreated monkeys (280), range 100-500. ND = Not determined due to death or severe symptoms. Note: each set of values represents results from a single primate.

ited no behavioral changes suggestive of toxicity. The thioether conjugates were thus less toxic than their disulfide counterpart as had been predicted by in vitro cytotoxicity and mouse toxicology studies. Conjugates of 9.2.27, using the two linkages, were also compared in nude mouse xenografts of human melanoma. The serum half-life of thioether linked conjugates T,,, alpha was increased to 3 hr as compared to 90 min for the disulfide linkage (Fig. 6). The beta phase was 10.4 hr for both conjugates. The resultant tumor localization and biodistribution at 20 hr post administration is shown in Fig. 7. With the exception of kidney and intestine, higher levels of labeled PE were observed in tissues with thioether than disulfide linked conjugates, due to higher blood levels. The tumor level of thioether linked conjugate was also increased, primarily reflecting the higher blood levels; although with time this would be expected to result in improved tumor localization.

DISCUSSION

20 40 :Q

28

&

Dose post treatment

Selectiwty”

(IDsa)

(aTAC)-serum

oe -2

-I

0

2

3

4

”

t 120LL .

IOOElo60-

Holotoxin conjugates offer the potential for enhanced potency when compared to hemitoxin or RIP conjugates (Quinones et al., 1984; Godal et al., 1986; Fitzgerald et al., 1984; McIntosh et al., 1983; Youle and Neville, 1982; Vittetta et al., 1984). As shown with ricin and diphtheria toxin, enhanced potency may be a reflection of enhanced internalization due to certain hydrophobic regions within B-chain (McIntosh et al., 1983; Youle and Neville, 1982; Vitetta et

4020-

Concn (log ng/mi

1

Fig. 5. Evaluation of PE conjugates for potency and selectivity. This was assessed by a long exposure format of a 3H-leucine incorporation assay (Materials and Methods). Controls averaged 20,000 cpm/well. Top panel: HT-29 colon carcinoma cells. NR-LU-lO/PE (relevant), 0; unconjugated PE, n ; NR-ML-OS/PE (irrelevant), l. Bottom panel: A375 M/M melanoma cells. NR-LU-lO/PE (irrelevant), 0; unconjugated PE, n ; NR-ML-OS/PE (relevant), l

0

2

4

6

8

IO

12

14

16

I8

20

Time lhrl

Fig. 6. Comparison of Pseudomonas exotoxin immunoconjugates, serum half-life in Balb/c mice. 0, disulfide; A, thioether.

ALTON C. MORGAN

280

V.,

SK

MU

BO

LU

JR et al.

LI

SP

ST

TH

KI

IN

Fig. 7. Comparison of Pseudomonas exotoxin immunoconjugates, tumor localization and biodistribution. Animals were sacrificed and radiolabel in tissues assessed after 24 hr. q, thioether linkage; 0, disulfide linkage. % inj. dose/gram = percent injected dose/gram. BL = blood, TA = tail = injection site, TU = tumor, SK = skin, MU = muscle. BO = bone, LU = lung, LI = liver, SP = spleen, ST = stomach, TH = thyroid. KI = kidney. IN = intestine.

al., 1984). Although no homologous hydrophobic region has been shown to exist in Pseudomonas exotoxin A. other as yet unidentified regions within PE may serve the same function (Allured ef al., 1986). Conjugates of PE are usually more potent than conjugates made with the same antibody and ricin A-chain (Bjorn et al.. 1985; Pirker et al., 1985). This higher potency could translate in viro to an ability to kill cells at lower antibody concentrations; i.e., conditions that may be most commonly achieved upon irk rive administration. From studies with the 9.2.27 antibody in human melanoma, it was shown that administration of 50 milligrams of antibody resulted in heterogenous antibody localization with higher levels in the most well vascularized regions and lower levels in sites more distant from blood or lymphatic vessels (Oldham et al., 1984; Schroff et al.. 1985). To achieve localization in more avascular sites doses of 200-500 mg were required. Despite these dose limitations, antibody could be delivered to every tumor cell within a lesion with saturation of all antigen sites achieved in about half the patients. Thus it was extrapolated from these studies that for solid tumor therapy, conjugate would need to be administered at dose levels of 3-7 mgikg in order to achieve complete or near saturation of antigenic sites. Thus. it is important to examine parameters which affect not only potency but also toxicity and delivery. One parameter which has a profound impact on these parameters is the type of bond between antibody and the toxin. Ideally, it would be one that would produce the most potent conjugate with high selectivity, and high yields. By comparing different linkages with PE conjugates, we determined that thioether linkages yielded conjugates of at least equal potency to disulfide conjugates by in vitro evaluation in cell-free and cytotoxicity assays. In addition,

thioether linkage produced highly potent conjugates with all four antibodies evaluated. Thioether bonds with A-chains or hemitoxins have usually resulted in conjugates with reduced potency (Lambert et al., 1985; Edwards et al., 1983). SMCC proved to be the optimum reagent for generating thioether bonds, resulting in easily controlled, rapid conjugate formation. As far as we know, our results are the most efficient yet reported with regard to formation of I : I molar ratio conjugate with the lowest offering of toxin (1 : 1). This high efficiency is very important for large scale manufacturing necessary for clinical trials. In addition. we demonstrated that monovalent F(ab)’ conjugates were slightly more potent than those with whole antibody utilizing both disulfide and thioether conjugates. This is similar to results reported for anti-Ig immunotoxins used with “piggyback” B-chain (Fulton et al., 1986) but differs from results of comparison made with A-chain immunotoxins of anti-CALLA antibody (Raso et al., 1982). These distinctions may be due to presence of B-chain and its ability to enhance internalization or intracellular routing. Unexpectedly, we found that thioether conjugates were consistently more selective with long exposure; e.g., less toxic to antigen negative cells, than comparable disulfide bonded conjugates. This difference in z’itro extrapolated to less toxicity in vim When tested for toxicity in both mice and monkeys, thioether conjugates were consistently 2-10 fold less toxic than comparable disulfide conjugates. Indeed, primates tolerate as much thioether bonded holotoxin conjugate as melanoma patients administered ricin A-chain conjugate (Spittler et al., 1987). Thioether conjugates of antibody fragments were less selective with long exposure than those made with intact antibody but better than their disulfide coun-

Linkage

of PE immunotoxins

terpart. This indicates a role of the antibody in steric occlusion of sites in PE responsible for cell binding and toxicity. Many of the positive observations with thioether bonded conjugates can be explained by increased stability. The evidence from both long term (3 days or more) in vitro assays and animal toxicology experiments suggests that significant disruption of disulfide bonds can occur, leading to the release of PE that appears to be more toxic in free than conjugated form. Good evidence exists with A-chain and hemitoxin conjugates that cleavage of disulfide bonds occurs in uivo, resulting in reduced tumor locahzation and more rapid clearance (Bourrie et al., 1986; Letvin et al., 1986). Although disulfide conjugates of the Thy 1.2 antibody and PAP have been reported to be stable (Ramakrishnan and Houston, 1984), we have found that 9.2.27-PAP disulfide conjugates behave similarly to other disulfide bonded A-chains and hemitoxons (Pa.vanasasivam et al., 1987). With PE we found more rapid clearance and reduced tumor localization of both disulfide and thioether linked conjugates compared to unconjugated antibody. This short serum half-life is probably a reflection of rapid removal of conjugate due to putative receptors for PE in the liver. However, thioether bonded conjugates had a significantly longer serum half-life than disulfide conjugates, additional evidence for disulfide bond reduction in Go. It appears that the hepatocyte is the primary target for PE either free or conjugated to antibody (Willingham, personal communication). This contrasts with recent studies with ricin A-chain that demonstrated that conjugate is cleared primarily through reticuloendothelial cells in liver and spleen that express mannose receptors (Letvin et al., 1986; Skilleter et al., 1985). Uptake of ricin or conjugates of A-chain can be at least partially blocked by oxidation of carbohydrate on the A-chain (Skilleter et aI., 1985) or by blockade with mannose containing glycoproteins (Bourrie et al., 1986). Although PE lacks carbohydrate, its uptake by hepatocytes may be receptor mediated. However, in aitro it is difficult to demonstrate specific binding of PE to cells which are readily killed by the toxin (Fitzgerald et al., 1980). PE seems to form more selective conjugates than ricin or abrin probably due to the presence of fewer PE receptors on most human cells (Fitzgerald et al., 1984). The primate toxicology data confirmed that the liver is the major target organ, as the only observed changes occurred in hepatic function. LDH isoenzyme data were performed and only the hepatic form of the enzyme was elevated. The thioether bonded conjugates described in this manuscript are currently undergoing clinical evaluation in ovarian cancer. Understanding the nature of PE binding to hepatocytes will permit studies on further reduction in toxicity of PE conjugates. Other parameters will also have to be examined (Morgan et al., 1985) if this approach is to fulfil1 Paul Erlich’s

vision of antibodies 1960).

281

as magic

bullets

(Himmelweit,

Acknowledgements-We would like to thank Drs Pastan and Fitzgerald for their helpful advice and for providing disulfide linked PE conjugates for comparison. We would also like to thank Jeff Fellows, Stan Gofinch and Lori Lyen for fine technical assistance.

REFERENCES

Allured V. S., Collier R. J., Carroll S. F. and McKay D. B. (1986) Structure of exotoxin A of Pseudomonas aeruginosa at 3.0 angstrom resolution. Proc. natn. Acad. Sci. U.S.A., 1320-1324. Bjorn M. J., Ring D. and Frankel A. (1985) Evaluation of monoclonal antibodies for the development of breast cancer immunotoxins. Cancer Res. 45, 12141221. Bourrie B. J., Casellas P., Blythman H. E and Jansen F. K. (1986) Study of the plasma clearance of antibody-ricin A-chain immunotoxins. Evidence for specific recognition sites on the A-chain that mediate rapid clearance of the immunotoxin. Eur. J. Biochem. 155, l-10. Edwards D. C., Ross W. C. J., Cumber A. J., McIntosh D., Smith A.. Thorpe P. E., Brown A., Williams R. H. and Davies A. J. S. (1983) A comparison of the in vitro and in uivo activities of conjugates of anti-mouse lymphocyte globulin and abrin. Biochem. Biophys. Acia 717,212-211. Fitzgerald D. J. P., Morris R. E. and Salinger C. B. (1980) Receptor mediated internalization of Pseudomonas toxin by mouse fibroblasts. Cell 21, 867-873. Fitzgerald D. J. P.. Waldmann T. A.. Willineham M. C. and P&tan I. (1984) Pseudomonas exotoxi;anti-TAC: cell specific immunotoxin active against cells expressing the human T cell growth factor receptor. J. clin. Invest. 74, 966-97 1. Fulton R. J., Uhr J. W. and Vittetta E. S. (1986) The effect of antibody valency and lysosomotropic amines on the synergy between ricin A-chain and ricin B-chain containing immunotoxins. J. Immun. 136, 3103-3108, Godal A.. Fodstad 0.. Morgan A. C. and Pihl A. (1986) Melanoma cell lines show striking inherent differences in sensitivity to immunotoxins containing holotoxins. J. Natn. Cancer Inst. 77, 1241-1253. Himmelweit F. (1960) In The Collected Papers of Paul Erlich (Edited by Himmelweit F.). Pergamon, New York, Vol. 3. Houston L. L. (1983) Inactivation of ricin using 4 azidophenyl-beta-D-galacto-pyranoside and 4 diazophenylbeta-n galacto-pyranoside. J. biol. Chem. 258, 7209-7212. Hwang K. M., Foon K. A., Cheung P. H., Pearson J. W. and Oldham R. K. (1983) Selective anti-tumor effect of a potent immunoconjugate composed of the A chain of abrin and a monoclonal antibody to a hepatomaassociated antigen. Cancer Res. 44, 45784586. Hwang K. M., Morgan A. C. Jr., Fodstad 0. and Oldham R. K. (1985) Radiolocalization of xenografted human malignant melanoma by a monoclonal antibody (9.2.27) to a melanoma-associated antigen in nude mice. Cancer Res. 45, 4150-4155. Lambert J. M., Senter P. D., Yound A. Yau, Blattler W. A. and Goldmacher V. S. (1985) Purified immunotoxins that are reactive with human lymphoid cells: monoclonal antibodies conjugated to the ribosome-inactivating proteins gelonin and pokeweed antiviral proteins. J. biol. Chem. 260, 12,035-12,041. Leppla S. H. (1976) Large scale purification and characterization of the exotoxin of Pseudomonas aeruginosa. Infect. Immun. 14, 1077-1086. Letvin N. L., Chalifous L. V., Reimann K. A., Ritz J., Schlossman S. F. and Lambat J. M. (1986) In vivo administration of lymphocyte-specific monoclonal antibodies in non-human primates. Delivery of ribosome

282

ALTOS C. MORGANJK et al

inactivating proteins to spleen and lymph node T-cells. J. clin. In&;. 18, 666-673. _ _ McIntosh D. P.. Edwards D. C.. Cumber A. J.. Parnell B. D., Dean C. J., Ross W. C. .I. and Forrester J. A. (1983) Ricin B chain converts a non-cytotoxic antibody-ricin A chain conjugate into a potent and specific cytotoxic agent. F.E.B.S. Let!. 164, 17-20. Morgan A. C. Jr., Schroff R. W., Pearson J. W. and Pavanasasivam G. (1985) Parameters affecting the utility of monoclonal antibodies as carriers of drug, toxins and radioisotopes. In Monoclonal Antibodies and Cancer Therapy (Edited by Reisfeld R. A. and Sell S.). UCLA Symposium on Molecular and Cellular Biology. Alan R. Liss, Inc., New York, NY, pp. 237-241. Morgan A. C. Jr., Galloway D. R. and Reisfeld R. A. (I 98 I) Production and characterization of monoclonal antibody to a melanoma-specific glycoprotein. Hybridoma 1, 27 36. Morgan A. C. Jr., Bordonaro J.. Pearson J. W. and Pavanasasivam G. (1987) Immunotoxins to a human melanoma-associated antigen: resistance to pokeweed antiviral protein conjugates in rirro. J. Cancer InsI. 78, 1101-1106. Okabe T., Kaizu, T. Fujisawa. Watanabe J., Kojima K., Yamashita T. and Takaku F. (1984) Monoclonal antibodies to surface antigens of small cell carcinoma of the lungs. Cancer Res. 44, 5273-5278. Oldham R. K., Foon K. A., Morgan A. C. Jr., Woodhouse C. S.. Schroff R. W.. Abrams P. G.. Fer M.. Schoenbereer C. S., Farrell M., Kimball E. and Sherwin S. A. (1984) Monoclonal antibody therapy of malignant melanoma: in rice localization in cutaneous metastases after intravenous administration. J. C/in. Oncol. 2, 123551244. Pavanasasivam G., Pearson J. W., Bohn W., Oldham R. K., Sadoff J. C. and Morgan A. C. Jr. (1987) Immunotoxins to a human melanoma associated antigen: comparison of gelonin with ricin and other A-chain conjugates. Cancer Res. 47, 3 16993 173. Pirker R., Fitzgerald D. J., Hamilton T. C.. Ozols R., F., Laird W., Frankel A. E., Willingham M. C. and Pastan I. (1985) Characterization of immunotoxins active against ovarian cancer cell lines. J. c[in. Inre.y/. 76, 1261-1267. Quinones R. R., Youle R. J., Kersey J. H., Zanjani E. D.. Azemove S. M., Soderling C. C. B., Lebien T. W., Beverly P. C. L., Neville D. M. and Vallera D. A. (1984) Anti-T cell monoclonal antibodies conjugated to ricin as potential reagents for human GVDH prophylaxis: effect on the generation of cytotoxic T cells in both peripheral blood and bone marrow. J. Immun. 132, 678483. Ramakrishnan S. and Houston L. L. (1984) Comparison of the selective cytotoxic effects of immunotoxins containing ricin A chain of pokeweed antiviral protein and anti-Thy 1.1 monoclonal antibodies. Cancer Res. 44, 2OlL208. Ramakrishnan S. and Houston L. L. (1985) Immunological and biological stability of immunotoxins in vice as studied by the clearance of disulfide-linked pokeweed and antiviral protein conjugates from blood. Cancer Res. 45, 203 l-2036. Raso V., Ritz J., Basala M. and Schlossman S. F. (1982) Monoclonal antibody-ricin A chain conjugate which is selectively cytotoxic for cells bearing the common acute lymphoblastic leukemia antigen. Cancer Res. 42, 457463. Schroff R. W., Woodhouse C. S.. Foon K. A.. Oldham R. K., Farrell M. M., Klein R. A. and Morgan A. C. Jr. (1985) Intratumor localization of monoclonal antibody in patients with malignant melanoma treated with antibody to a 250 K dalton melanoma-associated antigen. J. Narn. Cancer Inst. 14, 299-306. Skilleter D. N., Price R. J. and Thorpe P. E. (1985) Modification of the carbohydrate in ricin with metaperiodate and cyanoborohydride mixtures: effect on binding, uptake and toxicity to parenchymal and non-

parenchymal cells of rat liver. Biochem. Biophys. Acru 852, 12-21. Spittler L. E.. delRio M., Khentigan A., Wedel N. I.. Brophy N. A.. Miller L. L., Harkonen W. S., Rosendorf L. L., Lee H. M.. Mischak R. P., Kawahata R. T.. Stondemire J. B.. Fradkin L. B.. Bautista E. E. and Scannon P. J. (1987) Therapy of patients with malignant melanoma using a monoclonal anti-melanoma antibodyricin A-chain immunotoxin. Cancer Res. 47, 1717-1723. Thorpe P. E.. Brown A. N. F., Ross W. C. J., Cumber A. J., Deter S. I., Edwards D. C., Davies A. J. S. and Stirpe F. (1981) Cytotoxicity acquired by a conjugation of an anti-Thy I.1 monoclonal antibody and the ribosome inactivating protein gelonin. Eur. J. Biochern. 116,447~454. Thorpe P. E.. Mason D. W., Brown A. N. F., Simmons S. J., Ross W. J. C., Cumber A. J. and Forrester J. A. (1982) Selective killing of malignant cells in a leukaemic rat bone marrow using an antibody-ricin conjugate. Nature 297, 594-596. Thorpe P. E., Ross W. C. J., Brown A. N. F., Myers C. D.. Cumber A. .I.. Foxwell B. M. J. and Forrester J. T. (1984) Blockade of the galactose-binding sites of ricin by its linkage to antibody. Specific cytotoxic effects of the conjugates. Eur. J. Biochem. 140, 63-71. Thorpe P. E., Brown A. N. F., Brenner J. A. G. Jr., Foxwell B. M. J. and Stirpe F. (1985) An immunotoxin of rnonoclonal anti-Thy I.1 antibody and a ribosome inactivating protein from Suponaria ojicinalis: potent antitumor effects in ciwo and in aiuo. J.N.C.I. 75, I51 159. Uchiyama T.. Broder S. and Waldmann T. A. (1981 a) A monoclonal antibody (anti-TAC) reactive with activated and functionally mature human T-cells. 1. Production of anti-TAC monoclonal antibody and distribution of TAC positive cells, J. Immun. 126, i 393Sl397. Uchivama T.. Nelson D. L.. Fleisher T. A. and Waldmann T. ‘A. (I 981 b) A monoclonal antibody (anti-Tat) reactive with activated and functionally mature human T-cells. II. Expression of TAC antigen on activated cytotoxic T-cells. suppressor cells and one of two types of helper T-cells. J. Immun. 126, 1398.. 1403. Vallera D. A., Quinones R. R., Azemove S. M. and Soderling C. C. B. (1984) Monoclonal antibodv-toxin conjugatis reactive against human T lymphocytes. A comparison of antibody linked to intact ricin with antibody linked to ricin A-chain. Transplantation 37, 387-392. Vanness B. G., Howard J. B. and Bodley J. W. (1980) ADP-ribosylation of elongation factor 2 by diphtheria toxin, Isolation and properties of the novel ribosyl-amino acid and its hydrolysis products. J. hiol. Chem. 255, 10,717~10,720, Vittetta E. S., Fulton R. J. and Uhr J. W. (1984) Cytotoxicity of a cell-reactive immunotoxin containing ricin A chain is potentiated by an anti-immunotoxin containing ricin B chain. J. e.xp. Med. 160, 341-346. Wilbur D. S., Jones D. S., Fritzberg A. R. and Morgan A. C. Jr. (1986) Radioiodination of monoclonal antibodies. Labeling with para-iodophenyl (PIP) derivatives for in l,ico stability of the radioiodine label. J. nucl. Med. 27,959. Willingham M. C.. Fitzgerald, D. J. and Pastan I. (1987) Pseudomonas exotoxin coupled to a monoclonal antibody against ovarian cancer inhibits the growth of human ovarian cancer cells in a mouse model. Proc. natn .4cad. Sci. U.S.A. 54, 2474-2478. Woodhouse C. S. Bordonaro J. P., Beaumier P. L. and Morgan A. C. (1990) Second generation monoclonal antibodies to a human melanoma-associated proteoglycan. In Human Melanoma: Immunology Diagnostics and Therapy (Edited by Ferrone S.). Springer-Verlag. New York. In press. Youle R. J., and Neville D. M. Jr. (1982) Kinetics of protein synthesis inactivation by ricin-anti-Thy I.1 monoclonal antibody hybrids. Role of the ricin B subunit demonstrated by reconstitution. J. biol. Chem 257, 159881601.