Structural features of the antibody-A chain linkage that influence the activity and stability of ricin A chain immunotoxins.

The importance of the various structural elements constituting a ricin A chain immunotoxin to the stability of the disulfide bond between the antibody and A chain was examined using a panel of immunoconjugates prepared with the mouse monoclonal antibody Fib75. Analogues of the standard ricin A chain immunotoxin prepared with the N-succinimidyl 3-(2-pyridyldithio)propionate disulfide cross-linker included immunoconjugates made with N-succinimidyl 4-[(iodoacetyl)amino]benzoate the thioether cross-linker; with N-succinimidyl 3-(2-pyridyldithio)butyrate, the hindered disulfide cross-linker; with a peptide spacer between the antibody and cross-linker; or with the dodecapeptide corresponding to the C-terminus of ricin A chain. The cytotoxic activities of the immunoconjugates and their susceptibility to reduction by glutathione in vitro were compared. The thioether-linked immunotoxin could not be cleaved by glutathione in vitro and had low cytotoxic potency, consistent with the requirement of a reducible disulfide linkage for activity. The hindered disulfide-linked immunotoxin was 3-fold more stable to reduction than the immunotoxin containing a standard unhindered disulfide linkage, but the cytotoxic activities of the two constructs were indistinguishable. The introduction of a flexible peptide Ala-Ala-Pro-Ala-Ala-Ala-Pro-Ala-Pro-Ala between Fib75 and the disulfide linkage introduced by SPDP had no deleterious effect on cytotoxic activity and no effect on the susceptibility of the disulfide linkage to reduction. This finding suggests that the enforced proximity of the A chain to the antibody caused by the use of a short chemical cross-linker in a conventional immunotoxin has no influence on either of these properties in this system.(ABSTRACT TRUNCATED AT 250 WORDS)

Molecular and biological properties of an abrin A chain immunotoxin designed for therapy of human small cell lung cancer.

An immunotoxin (IT) comprising abrin A chain attached to the mouse monoclonal antibody SWA11, recognising a cell surface antigen highly associated with human small cell lung cancer (SCLC), was synthesised using a hindered disulphide crosslinker, N-succinimidyl 3-(2-pyridyldithio) butyrate (SPDB), and purified by Blue Sepharose CL-6B affinity chromatography. The IT preparation contained monomeric conjugate, composed of one abrin A chain molecule linked to one SWA11 molecule, and was free from unconjugated A chain or antibody. The IT fully retained the cell-binding capacity of the antibody component and the ribosome-inactivating activity of the A chain. In cytotoxicity assays using the SW2 SCLC cell line in tissue culture, SWA11-SPDB-abrin A chain inhibited the incorporation of 3H-leucine by 50% at a concentration of 10 pM and by 99% at a concentration of 1 nM. The anti-tumour efficacy of the IT was tested in nude mice bearing established s.c. solid SW2 tumour xenografts. A single i.v. injection of SWA11-SPDB-abrin A chain at a non-toxic dose induced a significant 7 to 10 day growth delay that could not be matched by administration of equivalent doses of either unconjugated SWA11 or abrin A chain alone. The results of this study indicate that the antigen recognised by SWA11 is an effective target for therapy of SCLC with A chain ITs in vivo.

Comparative stabilities in vitro and in vivo of a recombinant mouse antibody FvCys fragment and a bisFvCys conjugate.

A murine antibody FvCys fragment with a single additional cysteine residue at the C terminus of the VH domain was expressed in Escherichia coli from a modified expression plasmid containing the structural genes for the VH and VL domains derived from the anti-lysozyme hybridoma D1.3. Chemical cross-linking between the introduced sulfhydryl groups of two FvCys fragments by means of bis-maleimidohexane was used to generate a bisFvCys conjugate. The stability of the bisFvCys conjugate and an FvCys analogue that had been reacted with N-ethyl-maleimide to block the free sulfhydryl group, FvCys(BL), were compared after 125I-labeling. The bisFvCys conjugate was completely stable to incubation in solution at 37 degrees C for 24 h whereas only 60% of the FvCys(BL) fragment remained soluble. After i.v. administration to normal Wistar rats, both Fv proteins were rapidly cleared from the circulation with biphasic kinetics that were best fitted to a two-compartment open pharmacokinetic model. The alpha-phase half-life of the bisFvCys conjugate, 0.32 h, was significantly longer than that of the FvCys(BL) fragment, 0.15 h (p less than 0.001) whereas there was no significant difference between the beta-phase half-lives, 1.4 to 1.6h. No chain cleavage or covalent attachment to serum protein was detected by SDS-PAGE analysis of serum samples. However, gel permeation HPLC revealed that both Fv proteins associated with serum proteins in vivo and in vitro.

Blood clearance in the rat of a recombinant mouse monoclonal antibody lacking the N-linked oligosaccharide side chains of the CH2 domains.

The serum half-lives of a wild-type recombinant mouse monoclonal antibody of the IgG2b isotype and a mutant antibody differing from the wild-type antibody by a single amino acid substitution introduced into the CH2 domain, the replacement of Asn 297 by Ala to delete the conserved site of heavy chain glycosylation, were determined in the rat. The biological half-life of the aglycosyl Asn 297-Ala mutant recombinant antibody (4.8 days) was significantly shorter than that of the normally glycosylated wild-type antibody (7.4 days) by enzyme immunoassay. A similar difference between the biological half-lives of 125I-labelled aglycosyl and wild-type antibodies (2.9 and 4.0 days, respectively) was determined by gamma counting. Analysis of serum samples demonstrated that both recombinant antibodies were present in the circulation predominantly as intact monomeric IgG and revealed no differences that could account for the more rapid elimination of the aglycosyl antibody. The results of this investigation indicate that the carbohydrate residues contribute only in part to the survival of IgG in vivo and suggest that the diminished half-life of the aglycosyl antibody is due to increased catabolism in the extravascular tissues.

Recombinant mouse monoclonal antibodies with single amino acid substitutions affecting Clq and high affinity Fc receptor binding have identical serum half-lives in the BALB/c mouse.

The serum half-lives of three recombinant mouse monoclonal antibodies, differing radically in their ability to bind to Clq or FcRI but only minimally in structure, were determined in the BALB/c mouse following intravenous administration. The wild-type antibody, a chimaeric antibody comprising variable domains binding 3-iodo-4-hydroxy-5-nitrophenylacetate and constant domains of the mouse IgG2b isotype, was eliminated from the bloodstream with biphasic kinetics: alpha-phase, 0.5 days; beta-phase, 7.0 days. The alpha- and beta-phase half-lives of mutant recombinant antibodies with single amino acid substitutions, either Glu 235-Leu allowing binding to the mouse FcRI, or Lys 322-Ala reducing Clq binding 30-fold, were indistinguishable from those of the wild-type antibody demonstrating that the biological half-life of intact mouse IgG is independent of the ability to bind Clq or FcRI. The major implication of the present study is that IgG molecules which have been genetically engineered to eliminate interaction with other components of the immune system should retain the long half-life typical of natural antibodies.

The effect of ricin B chain on the intracellular trafficking of an A chain immunotoxin.

Covalent linkage of the A chain of ricin to the LICR-LOND-Fib75 monoclonal antibody produced an immunotoxin, Fib75-SS-ricin A, which demonstrated immunospecific toxicity to human bladder carcinoma cells in tissue culture (Forrester et al., 1984). The present studies have shown that ricin B chain potentiates the toxicity of the immunotoxin by two orders of magnitude and also significantly increases the rate of protein synthesis inhibition. Using immunoelectron microscopy, the receptor-mediated endocytosis and intracellular routing of the immunotoxin was studied with and without ricin B chain treatment after immunolocalisation of the conjugate. Fib75-SS-ricin A was internalised by the EJ cells predominantly in uncoated pits and vesicles and directed to the endosomes. Some degradation of the complex appeared to take place in multivesicular endosomes at early timepoints and 24 h after internalisation, most of the immunotoxin was found in lysosomes. Some ricin A chain epitopes were detected in Golgi vesicles. Cells treated with immunotoxin and ricin B chain endocytosed the complex predominantly in coated pits and coated vesicles. Using pre-embedding immunoperoxidase techniques, ricin chains were found in the whole Golgi complex and most of the conjugate escaped lysosomal degradation. Internalised immunotoxin was recycled back to the plasma membrane in an active form associated with vesicles which appeared to be derived predominantly from multivesicular endosomes. A similar mode of recycling has recently been reported (McIntosh et al., 1990) for ricin holotoxin in the same cell line. These observations may explain the potentiating effect of toxin B chains in the antibody-directed targeting of toxin A chains.

Cytotoxic activity of ricin A chain immunotoxins recognising cluster 1, w4 and 5A antigens associated with human small cell lung cancer.

The potential of mouse monoclonal antibodies raised against the human small cell lung cancer (SCLC) cell line SW2 to form active immunotoxins was evaluated using an indirect assay of immunotoxin cytotoxicity. Monoclonal antibodies recognising different SCLC-associated antigens, designated as cluster w4 and cluster 5A antibodies by the First International Workshop on SCLC Antigens, mediated the cytotoxic action of a screening agent made by linking ricin A chain to sheep anti-mouse IgG Fab' fragment. In contrast, monoclonal antibodies belonging to cluster 1 gave no significant cytotoxic effects in conjunction with the screening agent. Immunotoxins made by the direct chemical conjugation of ricin A chain to SWA11 (cluster w4) and SWA20 (cluster 5A) both exhibited selective toxic effects upon the SW2 cell line in tissue culture, inhibiting the incorporation of 3H-leucine by 50% at a concentration (IC50) of approximately 3 x 10(-10) M. An immunotoxin made with SEN36 (cluster 1) was much less potent with an IC50 greater than 1 x 10(-8) M.

Biochemical, cytotoxic and pharmacokinetic properties of an immunotoxin composed of a mouse monoclonal antibody Fib75 and the ribosome-inactivating protein alpha-sarcin from Aspergillus giganteus.

An immunotoxin was synthesized by the attachment of alpha-sarcin, the ribosome-inactivating protein derived from the mould Aspergillus giganteus, to a monoclonal mouse IgG2 antibody Fib75. The alpha-sarcin immunotoxin exerted toxic effects in tissue culture against the EJ human bladder carcinoma cell line, expressing the antigen recognised by the Fib75 antibody, inhibiting the incorporation of [3H]leucine by 50% at a concentration of 0.46 nM. The cytotoxic effects of the alpha-sarcin immunotoxin were indistinguishable from those of a Fib75 immunotoxin made with ricin A chain. Fib75-alpha-sarcin was cleared from the circulation of the rat with biphasic kinetics following intravenous administration. The alpha- and beta-phase half-lives were 0.8 h and 6 h, respectively, similar to the serum half-lives of analogous Fib75 immunotoxins made with ribosome-inactivating proteins derived from plants. alpha-Sarcin was completely stable in physiological saline buffer at 37 degrees C, whereas the ribosome-inactivating activity of ricin A chain was gradually lost under identical conditions. alpha-Sarcin may be a valuable alternative to ricin A chain for the construction of therapeutic immunotoxins because of its smaller size and greater thermostability.

Blocked and non-blocked ricin immunotoxins against the CD4 antigen exhibit higher cytotoxic potency than a ricin A chain immunotoxin potentiated with ricin B chain or with a ricin B chain immunotoxin.

An immunotoxin consisting of ricin A chain linked to the monoclonal antibody M-T151, recognising the CD4 antigen, was weakly toxic to the human T-lymphoblastoid cell line CEM in tissue culture. The incorporation of [3H]leucine by CEM cells was inhibited by 50% at an M-T151--ricin-A-chain concentration (IC50) of 4.6 nM compared with an IC50 of 1.0 pM for ricin. In contrast, immunotoxins made by linking intact ricin to M-T151 in such a way that the galactose-binding sites of the B chain subunit were either blocked sterically by the antibody component or were left unblocked, were both powerfully cytotoxic with IC50 values of 20-30 pM. The addition of ricin B chain to CEM cells treated with M-T151--ricin-A-chain enhanced cytotoxicity by only eight-fold indicating that isolated B chain potentiated the action of the A chain less effectively than it did as an integral component of an intact ricin immunotoxin. Ricin B chain linked to goat anti-(mouse immunoglobulin) also potentiated weakly. Lactose completely inhibited the ability of isolated ricin B chain to potentiate the cytotoxicity of M-T151--ricin-A-chain and partially (3- to 4-fold) inhibited the cytotoxicity of the blocked and non-blocked ricin immunotoxins. Thus, in this system, the galactose-binding sites of the B chain contributed to cell killing regardless of whether isolated B chain was associated with the A chain immunotoxin or was present in blocked or non-blocked form as part of an intact ricin immunotoxin. The findings suggest that the blocked ricin immunotoxin may become unblocked after binding to the target antigen to re-expose the cryptic galactose-binding sites. However, the unblocking cannot be complete because the maximal inhibition of [3H]leucine incorporation by the blocked immunotoxin was only 80% compared with greater than 99% inhibition by the non-blocked immunotoxin.

Comparative biochemical, cytotoxic and pharmacokinetic properties of immunotoxins made with native ricin A chain, ricin A1 chain and recombinant ricin A chain.

Immunotoxins were constructed by attaching native ricin A chain, ricin A1 chain and recombinant ricin A chain to the mouse monoclonal IgG2a antibody Fib75 by means of a disulphide linkage using the hetero-bifunctional cross-linker SPDP. The Fib75 immunotoxins were of similar composition and exerted identical cytotoxic effects against the EJ human bladder carcinoma cell line in tissue culture. All 3 immunotoxins broke down to the same extent upon incubation with glutathione in vitro. The clearance of the immunotoxins from the circulation of normal Wistar rats was determined following i.v. administration. The concentration of intact immunotoxin in serum samples taken at various intervals up to 48hr after injection was measured by a ricin A chain-specific ELISA. The Fib75 immunotoxin made with native ricin A chain was removed from the circulation most rapidly. Fib75-recombinant ricin A chain persisted in the circulation at a higher level than Fib75-ricin A1 chain. A higher proportion of the ricin A1 chain immunotoxin was lost from the bloodstream during the alpha-phase. The beta-phase half-lives of Fib75-recombinant ricin A chain and Fib75-ricin A1 chain were similar, consistent with the identical susceptibility of the immunotoxins to cleavage by glutathione. The presence of the complex-type oligosaccharide side-chain on the A1 chain may have accelerated the clearance of the A1 chain immunotoxin in relation to that of the immunotoxin made with the aglycosyl recombinant A chain.

Purification of immunotoxins containing the ribosome-inactivating proteins gelonin and momordin using high performance liquid immunoaffinity chromatography compared with blue sepharose CL-6B affinity chromatography.

Comparable amounts of the ribosome-inactivating proteins (RIPs) ricin A chain, gelonin and momordin were allowed to bind to Blue Sepharose CL-6B (immobilised Cibacron Blue F3GA) in phosphate buffer, pH 7.5, and were then eluted quantitatively with buffer containing 0.5 M NaCl. Differences in the elution profiles indicated that the RIPs possessed different affinities for the Cibacron Blue F3GA dye. Conjugation of the RIPs to the monoclonal antibody Fib75 resulted in decreased affinity for Blue Sepharose. Under conditions allowing the complete separation of the Fib75-ricin A chain immunotoxin from unconjugated Fib75, the Fib75 immunotoxins made with gelonin and momordin failed to bind completely to the Blue Sepharose column. The Fib75-gelonin and Fib75-momordin fractions that eluted from the column with 0.5 M NaCl were free of unconjugated Fib75 but were enriched in multiply substituted conjugate molecules. A high performance liquid immunoaffinity chromatography procedure based on the selective binding of conjugated RIP to immobilised affinity-purified anti-RIP antibody permitted the complete separation of the gelonin and momordin immunotoxins from unconjugated Fib75 without altering the composition, molecular integrity or cytotoxic activity of the immunotoxins.

Pharmacokinetics in the rat of a panel of immunotoxins made with abrin A chain, ricin A chain, gelonin, and momordin.

A panel of immunotoxins was constructed by chemically attaching the ribosome-inactivating proteins abrin A chain, ricin A chain, gelonin, and momordin to the monoclonal mouse IgG2a antibody Fib75 by means of a disulfide linkage. All the immunotoxins were toxic in tissue culture to the EJ human bladder carcinoma cell line expressing the antigen recognized by Fib75, inhibiting the incorporation of [3H]leucine by 50% at concentrations between 1 x 10(-10) M and 8 x 10(-10) M. The pharmacokinetics of the immunotoxins in the normal Wistar rat was determined following i.v. administration. The concentrations of intact immunotoxin in serum samples taken at various intervals after injection for up to 24 h were measured by solid-phase enzyme-linked immunosorbent assays specific for each of the four different ribosome-inactivating proteins. The Fib75 immunotoxins were cleared from the circulation with comparable, but not identical, biphasic kinetics best described by a two compartment open pharmacokinetic model. The alpha-phase half-lives of the panel, between 0.35 and 0.71 h, were similar. The beta-phase half-life of Fib75-abrin A chain, 13.3 h, was significantly longer than the beta-phase half-lives of Fib75-ricin A chain, Fib75-gelonin, and Fib75-momordin, between 7.5 and 8.6 h. Fib75-abrin A chain was found to be about 3- to 4-fold more resistant than the other immunotoxins to breakdown by reduction of the disulfide linkage between the A chain and the antibody with glutathione in vitro. This suggests that the longer serum half-life of Fib75-abrin A chain may have been due to greater stability against reduction in vivo. Analysis of serum samples obtained up to 24 h after injection of Fib75-abrin A chain revealed that the chemically intact immunotoxin present in the circulation retained full cytotoxic activity. An abrin A chain immunotoxin made with a different monoclonal mouse IgG2a antibody was also found to be more stable against reduction by glutathione in vitro than an analogous ricin A chain immunotoxin. Thus, abrin A chain may posses unique molecular properties that endow immunotoxins made with this A chain with greater stability in vivo than immunotoxins made with ricin A chain or other ribosome-inactivating proteins.

Selective cytotoxic effects of a ricin A chain immunotoxin made with the monoclonal antibody SWA11 recognising a human small cell lung cancer antigen.

The potential of mouse monoclonal antibodies for recognising different antigens associated with human small cell lung cancer (SCLC) to form active immunotoxins was assessed by an indirect in vitro screening assay. The screening agent used was a conjugate made by linking ricin A chain to a sheep anti-mouse IgG Fab' fragment via a disulphide bond. The monoclonal antibodies SWA11 and SWA20 both mediated the toxic effects of ricin A chain against the HC12 classic SCLC cell line in dose-dependent fashion. The SWA11 antibody was the more effective; in combination with the screening agent at a concentration of 1 x 10(-7) M, it inhibited the incorporation of [3H] leucine into HC12 cells by 94% compared with only 44% inhibition in the case of SWA20. An immunotoxin made by the direct chemical conjugation of ricin A chain to SWA11 exhibited selective toxic effects upon HC12 cells in tissue culture inhibiting the incorporation of [3H] leucine by 50% at a concentration (IC50) of 6.2 x 10(-10) M and by 98% at 1 x 10(-7) M. SWA11-ricin A chain had an IC50 of 4.4 x 10(-10) M against the NCI-H69 classic SCLC cell line but showed no cytotoxic activity against the human lung adenocarcinoma cell line NCI-H23 at a concentration of 1 x 10(-8) M.

The low uptake of an abrin A-chain immunotoxin by rat hepatic cells in vivo and in vitro.

The therapeutic value of antibody-ricin A-chain conjugates (immunotoxins) as antineoplastic agents is limited by their rapid removal from the circulation, primarily by cells of the liver which take up the immunotoxin through receptor mediated recognition of mannose-containing oligosaccharides in the toxin A-chain. We have therefore examined the uptake by rat hepatic cells of a monoclonal antibody (LICR-LOND Fib 75) conjugate assembled with the ricin related, but carbohydrate free, A-chain of the plant toxin abrin. The abrin A-chain immunotoxin was very poorly taken up in vivo and in vitro by both hepatic parenchymal and non parenchymal cells whereas a comparable conjugate assembled with ricin A-chain was actively accumulated by liver cells particularly the hepatic non-parenchymal cells. Furthermore, the abrin A-chain immunotoxin uptake by non-parenchymal cells in vitro was unaffected by mannose and the immunotoxin bound less readily to liver cells than did the ricin A-chain conjugate, consistent with a proposal that its accumulation by hepatic cells is brought about by endocytosis following non-specific binding or by fluid phase pinocytosis. These results suggest abrin A-chain immunotoxins might be further explored as anti-cancer agents since in some cases they could have an improved therapeutic efficacy over immunotoxins constructed with ricin A-chains.

An ineffective monoclonal antibody-ricin A chain conjugate is converted to a tumouricidal agent in vivo by subsequent systemic administration of ricin B chain.

An immunotoxin comprising a tumour-specific monoclonal antibody (11/160) coupled to ricin A chain, although inactive in in vitro cytotoxicity assays against HSNtc sarcoma target cells, was found to be capable of significant tumouricidal activity in syngeneic rats if potentiated by ricin B chain. The 11/160-ricin A, when bound to tumour cells prior to their inoculation, led to a slight inhibition of tumour growth s.c. compared with untreated sarcoma cells or those coated with antibody alone. However, all tumours in these groups developed progressively (69/69), whereas in those rats receiving 15 micrograms or 150 micrograms ricin B chain i.v. 5 min after tumour cell inoculation, the 'take rate' was reduced to 75% and 30% respectively, and significantly longer latent periods were evident for those tumours which did develop. Ricin B chain similarly inhibited, in a dose-dependent manner, the lung colonisation potential of 11/160-ricin A coated HSNtc cells. No effects were obtained if the B chain treatment followed inoculation of untreated or antibody-coated cells, suggesting that systemically administered B chain is capable of gaining access to and activating antibody-ricin A chain conjugates bound to the surface of syngeneic sarcoma cells in lung or subcutaneous sites. Tumour inhibition was obtained in some instances with intervals of up to 24 h between inoculation of conjugate-coated tumour cells and B chain. Experiments are in progress to determine if such potentiation may be feasible in a therapeutic rather than a prophylactic setting using this syngeneic solid tumour system.

Effect of linkage variation on pharmacokinetics of ricin A chain-antibody conjugates in normal rats.

We have investigated the pharmacokinetics of three ricin A chain-antibody conjugates having different bridging structures. Conjugate 1 has a disulphide linkage and was prepared with the N-succinimidyl-3-(2-pyridyldithio)propionate cross-linking reagent. Conjugate 2 has a protected disulphide linkage with a methyl group substituted on the carbon atom of the bridging structure adjacent to the disulphide linkage. Its preparation necessitated the synthesis of a new cross-linking reagent N-succinimidyl 3-(2-pyridyldithio)-butyrate. Conjugate 3 has a sulphide linkage and was prepared with the cross-linking reagent N-succinimidyl 4-(iodoacetylamino)benzoate which was synthesized by a novel route. Conjugate 1 is reducible, conjugate 2 less easily reducible and conjugate 3 nonreducible. On administration to animals all three conjugates displayed biphasic kinetics. The reducibility of the bond had no significant effect on the early disappearance of the conjugate from the circulation. However, at the later time points ease of reduction of the bond was associated with a more rapid disappearance of conjugate.

Fate of an antibody-ricin A chain conjugate administered to normal rats.

The pharmacokinetics and catabolism of ricin A chain, a mouse monoclonal antibody (LICR-LOND-Fib 75) and a disulphide linked conjugate of the two have been studied following their intravenous administration to normal rats. Results indicate that the conjugate was removed from the circulation much more rapidly than the antibody but less quickly than the free ricin A chain. Disappearance of the conjugate from the circulation appeared to be biphasic with an early rapid initial phase followed by a much more rapidly than the antibody but less quickly than the free ricin A chain. Disappearance of the conjugate from the circulation appeared to be biphasic with an early rapid initial phase followed by a much slower phase. The fate of a conjugate with a 125I iodide label in the antibody component was compared with that of a conjugate similarly labelled but in the ricin A chain component. The results indicate that breakdown of the conjugate involves both cleavage of the disulphide linkage and complete catabolism of the whole conjugate molecule with the release of 125I iodide. Rapid cleavage of the disulphide bond in the vasculature does not appear to be responsible for the initial rapid disappearance of the conjugate from the circulation.

Delivery of ricin and abrin A-chains to human carcinoma cells in culture following covalent linkage to monoclonal antibody LICR-LOND-Fib 75.

With the object of generating specific cytotoxic agents, we have prepared covalent conjugates of the A-chains of ricin and of abrin with monoclonal antibody LICR-LOND-Fib 75 and investigated their toxicity toward a human tumor cell line in culture. Both conjugates proved to be potent cytotoxins toward cells carrying the appropriate antigen. The agent containing abrin A-chain was toxic at a significantly lower concentration and exerted its maximum effect more rapidly than the one containing ricin A-chain. Inclusion of chloroquine in the incubation medium significantly enhanced the toxic action of both conjugates without loss of immunospecificity. Because of the widespread occurrence on human tumor cell lines of the antigen recognized by Fib 75, these conjugates, particularly the one containing abrin A-chain, may find application in freeing human bone marrow ex vivo of infiltrated tumor cells prior to reinfusion as an autograft.