A human monoclonal antibody 264RAD targeting αvß6 integrin reduces tumour growth and metastasis, and modulates key biomarkers in vivo.

αvß6 integrin expression is upregulated on a wide range of epithelial tumours, and is thought to play a role in modulating tumour growth. Here we describe a human therapeutic antibody 264RAD, which binds and inhibits αvß6 integrin function. 264RAD cross-reacts with human, mouse and cynomolgus monkey αvß6, and inhibits binding to all ligands including the latency-associated peptide of TGF-ß. Screening across a range of integrins revealed that 264RAD also binds and inhibits the related integrin αvß8, but not the integrins α5ß1, αvß3, αvß5 and α4ß1. In vitro 264RAD inhibited invasion of VB6 and Detroit 562 cells in a Matrigel invasion assay and αvß6 mediated production of matrix metalloproteinase-9 in Calu-3 cells. It inhibited TGF-ß-mediated activation of dermal skin fibroblasts by preventing local activation of TGF-ß by NCI-H358 tumour cells in a tumour cell-fibroblast co-culture assay. In vivo 264RAD showed dose-dependent inhibition of Detroit 562 tumour growth, regressing established tumours when dosed at 20 mg/kg once weekly. The reduction in growth associated with 264RAD was related to a dose-dependent inhibition of Ki67 and phospho-ERK and a reduction of αvß6 expression in the tumour cells, coupled to a reduction in fibronectin and alpha smooth muscle actin expression in stromal fibroblasts. 264RAD also reduced the growth and metastasis of orthotopic 4T1 tumours. At 20 mg/kg growth of both the primary tumour and the number of metastatic deposits in lung were reduced. The data support the conclusion that 264RAD is a potent inhibitor of αvß6 integrin, with some activity against αvß8 integrin, that reduces both tumour growth and metastasis.

Liposomal encapsulation enhances the antitumour efficacy of the vascular disrupting agent ZD6126 in murine B16.F10 melanoma.

Vascular disrupting agents (VDAs) are able to affect selectively tumour endothelial cell morphology resulting in vessel occlusion and widespread tumour cell necrosis. However, single-agent antitumour activity of VDAs is typically limited, as tumour regrowth occurs rapidly following drug treatment. To improve the therapeutic effectiveness of VDAs, we investigated liposomal targeting using ZD6126 as a model VDA. ZD6126 is a phosphate-prodrug of the tubulin-binding vascular disrupting agent ZD6126 phenol. ZD6126 was encapsulated into long circulating PEG-liposomes for passive targeting and PEG-liposomes conjugated with peptide ligands containing the RGD-motif for active targeting to alpha(v)-integrins on tumour endothelial cells. ZD6126 could be stably encapsulated, and liposomes displayed minimal leakage in vitro (<10% in 3 weeks). In vivo, upon intravenous injection, free ZD6126 was rapidly converted into ZD6126 phenol, which was cleared from the circulation within minutes. In contrast, ZD6126 encapsulated into either RGD-targeted or PEG liposomes showed prolonged blood circulation times (t(1/2)=10 h), and ZD6126 phenol exposure was also prolonged (t(1/2)=8 h). Both liposomal formulations displayed tumour accumulation plus hepatosplenic uptake by local macrophages. The altered pharmacokinetics and tissue distribution profiles of both liposomal ZD6126 formulations resulted both in single-dose and multiple-dose regimes, in improved therapeutic efficacy in established murine B16.F10 melanomas compared with free ZD6126. The passively and actively targeted liposomes showed equal antitumour efficacy, indicating that delivery of ZD6126 to the tumour tissue may suffice to disrupt tumour blood vessels without the need for specific targeting to the tumour endothelium.

Tumour dose response to the antivascular agent ZD6126 assessed by magnetic resonance imaging.

ZD6126 is a vascular targeting agent that disrupts the tubulin cytoskeleton of proliferating neo-endothelial cells. This leads to the selective destruction and congestion of tumour blood vessels in experimental tumours, resulting in extensive haemorrhagic necrosis. In this study, the dose-dependent activity of ZD6126 in rat GH3 prolactinomas and murine RIF-1 fibrosarcomas was assessed using two magnetic resonance imaging (MRI) methods. Dynamic contrast-enhanced (DCE) MRI, quantified by an initial area under the time-concentration product curve (IAUC) method, gives values related to tumour perfusion and vascular permeability. Multigradient recalled echo MRI measures the transverse relaxation rate T(2)*, which is sensitive to tissue (deoxyhaemoglobin). Tumour IAUC and R(2)* (=1/T(2)*) decreased post-treatment with ZD6126 in a dose-dependent manner. In the rat model, lower doses of ZD6126 reduced the IAUC close to zero within restricted areas of the tumour, typically in the centre, while the highest dose reduced the IAUC to zero over the majority of the tumour. A decrease in both MRI end points was associated with the induction of massive central tumour necrosis measured histologically, which increased in a dose-dependent manner. Magnetic resonance imaging may be of value in evaluation of the acute clinical effects of ZD6126 in solid tumours. In particular, measurement of IAUC by DCE MRI should provide an unambiguous measure of biological activity of antivascular therapies for clinical trial.

Induction of apoptosis by the ADEPT agent ZD2767: comparison with the classical nitrogen mustard chlorambucil and a monofunctional ZD2767 analogue.

ZD2767P is a phenol mustard glutamate prodrug which is currently being developed for Antibody Directed Enzyme Prodrug Therapy (ADEPT). In ZD2767 ADEPT an active bi-functional alkylating drug, ZD2767D (4-[N,N-bis(2-iodoethyl)amino]phenol), is generated following cleavage of ZD2767P by the bacterial enzyme carboxypeptidase G2 (CPG2) which is targeted to the tumour by conjugation to the F(ab')(2)fragment of the anti-CEA antibody A5B7. The aim of the studies described here was to identify the mode of cell death induced by ZD2767P + CPG2 in comparison to the established nitrogen mustard chlorambucil. The contribution of bifunctional and monofunctional ZD2767 DNA lesions to cell death induction was investigated using a monofunctional ZD2767D analogue. Apoptosis in LoVo tumour cells was studied by three different methods (nuclear morphology, annexin V staining and TUNEL). Levels of apoptosis detected using the three assays were similar, and each drug treatment produced apoptosis at levels above those in control cells at concentrations which resulted in tumour cell growth inhibition. The bi-functional compounds, ZD2767P + CPG2 and chlorambucil, induced apoptosis in a concentration and time dependent manner, with equitoxic concentrations producing equivalent levels of apoptosis. In contrast, the mono-functional ZD2767D analogue at 100 microM resulted in the maximal level of apoptosis at 25 h with no further increase over the following 72 h. These studies have demonstrated that apoptosis is the mechanism of cell death induced by the ZD2767 ADEPT system, and that levels of apoptosis produced by ZD2767 are similar to those observed at equitoxic concentrations of the classical nitrogen mustard chlorambucil. The mono-functional ZD2767 analogue also induced apoptosis, but with a different time course and characteristics. In conjunction with previous data, these studies have shown that the potent activity of ZD2767 can be attributed to the ability of the compound to induce bifunctional DNA lesions and engage apoptosis.

Cellular glutathione as a determinant of the sensitivity of colorectal tumour cell-lines to ZD2767 antibody-directed enzyme prodrug therapy (ADEPT).

ZD2767P, a nitrogen mustard glutamate prodrug, is currently being evaluated in Phase 1 clinical trials of antibody directed enzyme prodrug therapy (ADEPT). There was no significant relationship between basal glutathione (GSH) concentration and sensitivity to ZD2767P + carboxpeptidase G2 (CPG2) in colorectal tumour cell-lines. Depletion of intracellular GSH using buthionine sulfoximine (BSO) resulted in only a modest potentiation of ZD2767P + CPG2 activity and hence BSO is unlikely to markedly enhance the activity of this ADEPT treatment.

ZD2767, an improved system for antibody-directed enzyme prodrug therapy that results in tumor regressions in colorectal tumor xenografts.

ZD2767 represents an improved version of antibody-directed enzyme prodrug therapy. It consists of a conjugate of the F(ab')2 A5B7 antibody fragment and carboxypeptidase G2 (CPG2) and a prodrug, 4-[N,N-bis(2-iodoethyl)amino]phenoxycarbonyl L-glutamic acid. The IC50 of the prodrug against LoVo colorectal tumor cells was 47 microM, and cleavage by CPG2 released the potent bis-iodo phenol mustard drug (IC50 = 0.34 microM). The drug killed both proliferating and quiescent LoVo cells. Administration of the ZD2767 conjugate to nude mice bearing LoVo colorectal xenografts resulted in approximately 1% of injected ZD2767 conjugate localizing/g of tumor after 72 h, and blood and normal tissue levels of the conjugate were 10-50-fold lower. A single round of therapy involving the administration of the prodrug 72 h after the conjugate to athymic mice bearing established LoVo xenografts resulted in approximately 50% of the tumors undergoing complete regressions, tumor growth delays greater than 30 days, and little toxicity (as judged by body-weight loss). Similar studies using a control antibody-CPG2 conjugate that does not bind to LoVo tumor cells resulted in a growth delay of less than 5 days, confirming the tumor specificity of this approach. These studies demonstrate the potential of ZD2767 for the treatment of colorectal cancer.

New mustard prodrugs for antibody-directed enzyme prodrug therapy: alternatives to the amide link.

Antibody-directed enzyme prodrug therapy (ADEPT) is a two-step approach for the treatment of cancer which seeks to generate a potent cytotoxic agent selectively at a tumor site. In this work described the cytotoxic agent is generated by the action of an enzyme CPG2 on a relatively nontoxic prodrug. The prodrug 1 currently on clinical trial is a benzamide and is cleaved by CPG2 to a benzoic acid mustard drug 1a. We have synthesized a series of new prodrugs 3-8 where the benzamide link has been replaced by, for example, carbamate or ureido. Some of these alternative links have been shown to be good substrates for CPG2 and therefore new candidates for ADEPT. The active drugs 3a and 4a derived from the best of these prodrugs are potent cytotoxic agents (1-2 microM) some 100 times more than 1a. The prodrugs 3 and 4 are some 100-200-fold less cytotoxic, in a proliferating cell assay, than their corresponding active drugs 3a and 4a.

Optimization of alkylating agent prodrugs derived from phenol and aniline mustards: a new clinical candidate prodrug (ZD2767) for antibody-directed enzyme prodrug therapy (ADEPT).

Sixteen novel potential prodrugs derived from phenol or aniline mustards and their 16 corresponding drugs with ring substitution and/or different alkylating functionalities were designed. The [[[4-]bis(2-bromoethyl)-(1a), [[[4-[bis(2-iodoethyl)-(1b), and [[[4-[(2-chloroethyl)-[2-(mesyloxy)ethyl]amino]phenyl]oxy] carbonyl]-L-glutamic acids (1c), their [[[2- and 3-substituted-4-[bis(2-chloroethyl)amino]phenyl]oxy]carbonyl]-L- glutamic acids (1e-1), and the [[3-substituted-4-[bis(2-chloroethyl)amino]phenyl]carbamoyl]-L- glutamic acids (1o-r) were synthesized. They are bifunctional alkylating agents in which the activating effect of the phenolic hydroxyl or amino function is masked through an oxycarbonyl or a carbamoyl bond to a glutamic acid. These prodrugs were designed to be activated to their corresponding phenol and aniline nitrogen mustard drugs at a tumor site by prior administration of a monoclonal antibody conjugated to the bacterial enzyme carboxypeptidase G2 (CPG2) in antibody-directed enzyme prodrug therapy (ADEPT). The synthesis of the analogous novel parent drugs (2a-r) is also described. The viability of a colorectal cell line (LoVo) was monitored with the potential prodrugs and the parent drugs. The differential in the cytotoxicity between the potential prodrugs and their corresponding active drugs ranged between 12 and > 195 fold. Compounds 1b-d,f,o exhibited substantial prodrug activity, since a cytotoxicity differential of > 100 was achieved compared to 2b-d,f,o respectively. The ability of the potential prodrugs to act as substrates for CPG2 was determined (kinetic parameters KM and kcat), and the chemical stability was measured for all the compounds. The unsubstituted phenols with different alkylating functionalities (1a-c) proved to have the highest ratio of the substrates kcat:KM. From these studies [[[4-[bis(2-iodoethyl)amino]phenyl]oxy]carbonyl]-L-glutamic acid (1b) emerges as a new ADEPT clinical trial candidate due to its physicochemical and biological characteristics.

Anti-tumour effects of an antibody-carboxypeptidase G2 conjugate in combination with phenol mustard prodrugs.

ADEPT is an antibody-based targeting strategy for the treatment of cancer. We have developed two new prodrugs, 4-[N,N-bis(2-chloroethyl)amino]-phenoxycarbonyl-L- glutamic acid (PGP) and (S)-2-[N-[4-[N,N-bis(2-chloroethyl)amino]- phenoxycarbonyl]amino]-4-(5-tetrazoyl)butyric acid (PTP), which are cleaved by the bacterial enzyme CPG2 to release the 4-[N,N-bis(2-chloroethyl)amino] phenol drug. In vitro, both prodrugs are approximately 100- to 200-fold less potent than the parent drug (1 h IC50 = 1.4 microM) in LoVo colorectal tumour cells. These prodrugs have been evaluated for utility in ADEPT when used in combination with a conjugate of CPG2 and the F(ab')2 fragment of the anti-CEA monoclonal antibody, A5B7. The conjugate was shown to localise specifically to established LoVo tumour xenografts growing in nude mice and optimal tumour-normal tissue ratios were achieved after 72 h. Administration of either prodrug, at doses which cause 6-8% body weight loss, 72 h after administration of the A5B7-CPG2 conjugate to the LoVo tumour-bearing mice resulted in tumour regressions and growth delays of 14-28 days. The PTP prodrug in combination with a high dose of conjugate (10 mg kg-1) gave the best anti-tumour activity despite being a 10-fold worse substrate for CPG2 than PGP. Prodrug alone, active drug alone or prodrug in combination with a non-specific conjugate had minimal anti-tumour activity in this tumour model.

In vitro and in vivo comparison of a randomly coupled antibody fragment-enzyme conjugate with a site-specific conjugate.

Two antibody fragment-enzyme conjugates, one obtained by random coupling of the two protein component, the other by site-specific ligation of the same component, were compared in vitro and in vivo for their usefulness in antibody directed enzyme prodrug therapy (ADEPT). The in vitro studies have shown that the site-specific conjugate has a higher antigen binding capacity, while both conjugates had similar specific enzymic activities. In vivo, the site-specific conjugate was cleared more rapidly. When correction was made for this faster clearance, both conjugates showed similar antitumor efficacy in a mouse xenograft system upon administration of a prodrug.

Comparison of the cellular internalization of antibodies used either as immunotoxins or in ADEPT.

The internalization into tumor cells of two antibodies (C242 and 454A12), which make potent immunotoxins when linked to ricin A-chain, and an antibody (A5B7), which does not make a potent immunotoxin but has proven useful in ADEPT, was evaluated. The 454A12 antibody was rapidly taken into the cells, 50% of the antibody being internalized after 2 h. The C242 antibody was internalized more slowly, approx 50% being taken up by the cells in 24 h. With A5B7, less than 10% of the antibody was internalized after 24 h. Internalization of the C242 antibody was accompanied by the appearance of antibody degradation products in the cell medium after 2 h, and this degradation could be inhibited by addition of a metabolic inhibitor that prevented cell internalization. In contrast, minimal degradation of the A5B7 antibody could be detected up to 24 h after binding to the cells. In conclusion, both 454A12 and C242 antibodies, which make potent immunotoxins, were internalized into tumor cells. The A5B7 antibody, which does not make a potent immunotoxin, was not internalized, and this property may be one reason why A5B7 has proved useful for delivery of enzymes in ADEPT.

The interactions of protein inhibitors with tumour proteases studied in solution and immobilised on cell surfaces in frozen sections.

The cell surface protease guanidinobenzoatase (GB) has been purified from human colonic and lung carcinoma tissue by an affinity step involving the binding of the enzyme either onto fibrin fibrils or onto agmatine-sepharose. The inhibitor protein (I) was extracted from the cytoplasm of tumour cells and isolated by an affinity step involving the binding of I to GB on the surface of cultured carcinoma cells. The interaction of GB and I in solution was followed by kinetic studies employing the release of the fluorescent 4-methylumbelliferone (MU) from the synthetic substrate 4-methylumbelliferyl-p-guanidinobenzoate (MUGB). The interaction of soluble I with membrane bound GB was followed by using the yellow fluorescent probe 9-aminoacridine (9AA) which binds to active GB but not to GB-I. The results of these studies demonstrated the presence of isoenzymic froms of GB which were recognized specifically by their appropriate isoinhibitor, isolated from the appropriate cell type. This high degree of selectivity suggests a cell specific regulatory role for the inhibitors and the possibility that they might be used for the delivery of cytotoxic molecules to the surface of specific types of tumour cells.

Antitumor effects of an antibody-carboxypeptidase G2 conjugate in combination with a benzoic acid mustard prodrug.

The F(ab')2 fragment of the antitumor monoclonal antibody, A5B7, was covalently linked to the bacterial enzyme carboxypeptidase G2 (CPG2). The resulting conjugate was used in combination with a prodrug of a benzoic acid mustard alkylating agent to treat human colon tumor xenografts in a two-step targeting strategy, antibody-directed enzyme prodrug therapy (ADEPT). The prodrug, 4-[(2-chloroethyl) (2-mesyloxyethyl)amino]-benzoyl-L-glutamic acid is rapidly converted by CPG2 to a drug that is at least 15x more toxic in vitro against LS174T colorectal tumor cells than the prodrug. Optimal tumor/blood ratios of the A5B7-CPG2 were achieved 72 h after administration of the conjugate to athymic mice bearing established LS174T tumor xenografts. Significant antitumor activity was seen in LS174T tumor-bearing mice treated with the conjugate followed 3 d later by the prodrug. In contrast, prodrug, conjugate, or active drug alone did not result in any antitumor activity in this tumor model. These studies demonstrate the advantage of a two-step ADEPT system for the treatment of colorectal cancer.

Further evidence for different isoenzymic forms of a cell surface protease, guanidinobenzoatase, associated with tumours.

Normal colonic epithelial cells possess a cell surface protease referred to as guanidinobenzoatase (GB) and a corresponding cytoplasmic protein inhibitor of GB. Colonic carcinoma cells possess two isoenzymic forms of GB, the normal and the carcinoma specific form, each of which is recognised by the corresponding inhibitors present in the cytoplasm of colonic carcinoma cells. An affinity-purified inhibitor preparation obtained from the cytoplasm of cultured colonic carcinoma cells inhibited these two colonic carcinoma isoenzymic forms of GB but not the GB associated with other forms of tumour. The data suggest that each cell type possessing isoenzymic forms of cell surface GB also possesses the corresponding cell-specific inhibitors of GB.

The role of fibrin fibrils in the dissociation of a cell surface protease-inhibitor complex and evidence for the recapture of the inhibitor protein.

Colonic epithelial cells possess a cell surface protease referred to as guanidinobenzoatase (GB). Active GB can be located by the fluorescent active site directed competitive inhibitor 9-amino acridine (9AA) followed by fluorescence microscopy. The cell surface GB can be transferred to fibrin fibrils, which have a higher affinity for GB than the cell surface. The cytoplasm of colonic epithelial cells contains a protein which inhibits membrane bound GB, forming a latent form of GB or GB-inhibitor complex. This complex can also be dislodged from the epithelial cell surface due to the high affinity of fibrin for GB, with the consequent dissociation of the enzyme-inhibitor complex and solubilisation of the inhibitor. This use of fibrin has led to the demonstration of the transfer of a selective inhibitor protein from one cell surface (the donor) to a second cell surface (the target).

Drug targeting with monoclonal antibodies. A review.

Monoclonal antibodies have been widely used in attempts to target anti-neoplastic agents selectively to tumours. Problems associated with the use of monoclonal antibodies as the targeting moiety include lack of complete tumour selectivity, antigenic heterogeneity, tumour access and immunogenicity. Considerable effort in the targeting field is being expended in an attempt to reduce or overcome these problems. Attachment of monoclonal antibodies to low molecular weight cytotoxic drugs, protein toxins, radionuclides or enzymes capable of conversion of inactive prodrugs to cytotoxic drugs, has, despite these problems, resulted in conjugates which do have selective anti-tumour effects in animal models. The advantages and limitations of these different approaches are reviewed. It remains to be established in man if any of these approaches will result in significant therapeutic benefit in major solid tumours.

Comparison of the pharmacokinetics and hepatotoxic effects of saporin and ricin A-chain immunotoxins on murine liver parenchymal cells.

Immunotoxins containing the ribosome-inactivating protein, saporin, are very effective antitumor agents but are highly toxic to mice. They induce severe necrotic lesions in the liver parenchyma of the recipients. Such extensive damage to the liver parenchyma is not observed with ricin A-chain immunotoxins even at 5-fold higher dosage. The hepatotoxicity of the saporin immunotoxins was found in the present study to arise from a combination of two effects. First, saporin and saporin immunotoxins were 30- and 6-fold more toxic to primary cultures of mouse liver parenchymal cells than were ricin A-chain and ricin A-chain immunotoxins, respectively. This was despite the fact that the cells bound 4- to 5-fold less saporin or saporin immunotoxins than ricin A-chain or ricin A-chain immunotoxins. The binding of ricin A-chain and its immunotoxin to the cells was mediated through the carbohydrate residues present on the A-chain whereas saporin is not glycosylated and thus must bind to other sites on the cell surface which result in transport of saporin relatively efficiently to the cytosol. The second reason for the hepatotoxic action of the saporin immunotoxin was that it had a longer blood half-life (t 1/2 alpha = 1.1 h; t 1/2 beta = 17.1 h) than the ricin A-chain immunotoxin (t 1/2 = 0.52 h; t 1/2 beta = 9.7 h). Analyses using a two-compartment pharmacokinetic model showed that the two immunotoxins broke down in vivo to give free antibody at a similar rate (t 1/2 = 10-12 h) but that the ricin A-chain immunotoxin was eliminated 11 times more rapidly than the saporin immunotoxin by routes other than breakdown. It was calculated that, in mice given a median lethal dose of saporin immunotoxin, the blood levels of immunotoxin remained above the concentration that killed 50% of parenchymal cells in vitro for more than 48 h. In mice given a median lethal dose of ricin A-chain immunotoxin, the blood levels fell below the concentration that was toxic to parenchymal cells in vitro within 4 h. The longer blood half-life of the saporin immunotoxin may also explain our previous finding that it had antitumor activity superior to that of a ricin A-chain immunotoxin in mice.

Improved antitumor effects of immunotoxins prepared with deglycosylated ricin A-chain and hindered disulfide linkages.

A monoclonal anti-Thy-1.1 antibody (OX7) was coupled to either native or chemically deglycosylated ricin A-chain (dgA) using one of two different cross-linking agents. One cross-linker, N-succinimidyloxycarbonyl-alpha-methyl-alpha-(2-pyridyldithio)tolu ene (SMPT), generates a sterically hindered disulfide bond which is relatively resistant to reduction, whereas the other, 2-iminothiolane hydrochloride, generates an unhindered disulfide bond with greater lability. A two-compartment pharmacokinetic model was used to analyze the blood levels of each immunotoxin and its breakdown product (free antibody) after i.v. injection into mice. Immunotoxins prepared with SMPT broke down in vivo 6.3-fold more slowly than those prepared with 2-iminothiolane hydrochloride, and immunotoxins containing native A-chain were cleared 2- to 3-fold more rapidly from the bloodstream than those containing dgA. As a result, 24 h after injection, 16% of the OX7-SMPT-dgA remained in the blood as compared with 0.4 to 2.5% of the other immunotoxins. Immunotoxins prepared with dgA were about 3-fold more toxic to mice than those prepared with native A-chain, whereas immunotoxins prepared with SMPT were only slightly more toxic than those prepared with 2-iminothiolane hydrochloride. When equivalent toxic doses of the immunotoxins were administered i.v. to mice which had been given injections of Thy-1.1+ AKR-A/2 lymphoma cells, the OX7-SMPT-dgA gave the best antitumor effect. A dose equivalent to one-seventh of the median lethal dose extended the survival time of the animals by the extent expected if 99.999% of the tumor cells had been eradicated. Furthermore, the tumors that did develop in the mice treated with OX7-SMPT-dgA were mutants which were resistant to all the immunotoxins. Some of the mutants were deficient in Thy-1.1 whereas others were not. In conclusion, both the use of the SMPT cross-linker and deglycosylation of the A-chain significantly improve the therapeutic index of the immunotoxins in AKR-A/2 tumor-bearing mice.

Uptake of native and deglycosylated ricin A-chain immunotoxins by mouse liver parenchymal and non-parenchymal cells in vitro and in vivo.

The therapeutic activity of ricin A-chain immunotoxins is undermined by their rapid clearance from the bloodstream of animals by the liver. This uptake has generally been attributed to recognition of the mannose-terminating oligosaccharides present on ricin A-chain by receptors present on the non-parenchymal (Kupffer and sinusoidal) cells of the liver. However, we demonstrate here that, in the mouse, the liver uptake of a ricin A-chain immunotoxin occurs in both parenchymal and non-parenchymal cells in equal amounts. This is in contrast to the situation in the rat, where uptake of the immunotoxin is predominantly by the non-parenchymal cells. Recognition of sugar residues on the A-chain portion of the immunotoxin plays an important role in the liver uptake by both cell types in both species. However it is not the only mechanism since, firstly, an immunotoxin containing ricin A-chain which had been effectively deglycosylated with metaperiodate and cyanoborohydride was still trapped to a significant extent by hepatic non-parenchymal cells after it was injected into mice. Secondly, deglycosylation, while eliminating uptake of the free A-chain by parenchymal and non-parenchymal cells in vitro, only reduced the uptake of an immunotoxin by either cell type by about half. Thirdly, the addition of excess D-mannose or L-fucose inhibited the uptake of free A-chain by mouse liver cell cultures by more than 80% but only inhibited the uptake of the native A-chain immunotoxin by about half and had little effect on the uptake of the deglycosylated ricin A-chain immunotoxin. Recognition of the antibody portion of the immunotoxin by liver cells seems improbable, since antibody alone or an antibody-bovine serum albumin conjugate were not taken up in appreciable amounts by the cultures. Possibly attachment of the A-chain to the antibody exposes sites on the A-chain that are recognised by liver cells in vitro and in vivo.