Solid-phase synthesis of heterocyclic aromates: applicability towards combinatorial chemistry; a review.

Because of their biological activity, stability in vivo, the rigid spatial positioning of their substituents, and their synthetic challenges, heterocyclic aromates continue to be of interest to both academic and industrial medicinal chemists. Currently, many drug-like heterocyclic aromates are prepared via solid-phase organic chemistry methods. This review examines the applicability of those methods towards combinatorial chemistry with respect to the basic demands of such an approach: 1) synthesis, work-up and subsequent purification should be easily automated enabling the efficient simultaneous synthesis of large numbers of highly pure compounds in a minimum amount of time, 2) large diversity among the ligands to be synthesized, 3) high conversion rates of the individual reaction steps, and 4) the use of commercially available starting materials. Although many methods have been developed for the synthesis of heterocyclic aromates, very few of the available methods enable the synthesis of highly diverse heteroaromatic libraries.

Non-depolarizing neuromuscular blocking activity of bisquaternary amino di- and tripeptide derivatives.

Herein we describe the synthesis of novel di- and tripeptide derivatives with two quaternary nitrogen groups attached and the biological testing of these compounds for neuromuscular blocking (NMB) activity in vitro and in vivo. The short peptide scaffold was selected because it offers potential for desired distance between the two pharmacophoric quaternary nitrogen groups, short duration of action, straightforward synthesis, and compatibility with an injectable formulation. From a small series of compounds 20c,e are identified as effective non-depolarizing NMB agents in vitro and in vivo in anesthetized cats and Rhesus monkeys with potencies similar to those of the clinical reference compounds rocuronium (4) and suxamethonium (2) (monkey ED(90) = 0.68, 0.23, 0.16, 5.04 micromol/kg, respectively). These new peptide derivatives 20c,e have similar potency and onset time but longer duration and slower recovery than the clinically used reference compounds. The structure-activity relationships described for this chemical series lead to the conclusion that the di- or tripeptide fragment can be regarded as an alternative template to the steroid or aliphatic ester of previously reported NMBs and within this tripeptide-derived series clog P correlates well with in vitro NMB activity.

Solid-phase synthesis of 4-substituted imidazoles using a scaffold approach.

Immobilized 4-iodoimidazole 2 was used in a metal/halogen exchange reaction followed by treatment with electrophiles and subsequent cleavage from the resin to yield 4-substituted imidazoles 8-11. Grignard reaction with the resin-bound ketones 5 yielded the corresponding alcohols 11. This approach was used for a library synthesis of 35 imidazoles.

Direct crosslinking of the antitumor antibiotic sparsomycin, and its derivatives, to A2602 in the peptidyl transferase center of 23S-like rRNA within ribosome-tRNA complexes.

The antitumor antibiotic sparsomycin is a universal and potent inhibitor of peptide bond formation and selectively acts on several human tumors. It binds to the ribosome strongly, at an unknown site, in the presence of an N-blocked donor tRNA substrate, which it stabilizes on the ribosome. Its site of action was investigated by inducing a crosslink between sparsomycin and bacterial, archaeal, and eukaryotic ribosomes complexed with P-site-bound tRNA, on irradiating with low energy ultraviolet light (at 365 nm). The crosslink was localized exclusively to the universally conserved nucleotide A2602 within the peptidyl transferase loop region of 23S-like rRNA by using a combination of a primer extension approach, RNase H fragment analysis, and crosslinking with radioactive [(125)I]phenol-alanine-sparsomycin. Crosslinking of several sparsomycin derivatives, modified near the sulfoxy group, implicated the modified uracil residue in the rRNA crosslink. The yield of the antibiotic crosslink was weak in the presence of deacylated tRNA and strong in the presence of an N-blocked P-site-bound tRNA, which, as was shown earlier, increases the accessibility of A2602 on the ribosome. We infer that both A2602 and its induced conformational switch are critically important both for the peptidyl transfer reaction and for antibiotic inhibition. This supposition is reinforced by the observation that other antibiotics that can prevent peptide bond formation in vitro inhibit, to different degrees, formation of the crosslink.

The effect of ethyldeshydroxy-sparsomycin and cisplatin on the intracellular glutathione level and glutathione S-transferase activity.

Ethyldeshydroxy-sparsomycin (EdSm) is a ribosomal protein synthesis inhibitor which synergistically enhances the antitumor activity of cisplatin against L1210 leukemia in vivo. Because cellular glutathione (GSH) and glutathione S-transferases (GST) are reported to interfere with the antitumor activity of cisplatin, we analyzed the effect of EdSm and cisplatin on GSH and GST activity in selected tumor cells. For this purpose we used three murine leukemia tumors with different sensitivities towards EdSm and cisplatin: L1210-WT, sensitive to both drugs, L1210-Sm, resistant to EdSm, and L1210-CDDP, resistant to cisplatin. No significant differences were detectable between these three cell lines regarding the population doubling time, the cell size, and the cellular level of protein and glutathione. Neither of the resistant L1210 subclones showed P-glycoprotein expression. Drug exposure, however, changed the intracellular dynamics. Exposure to EdSm strongly decreased the amount of cellular protein, decreased the overall GST activity and led to GSH depletion, whereas exposure to cisplatin induced a rise in the amount of protein, in GSH, and in the total GST activity. These effects are dose-dependent and correlate well with the sensitivity of the tumor cells for EdSm or cisplatin. In addition, exposure to EdSm lowered the V(max) of GST in L1210-WT and L1210-Sm; however, in L1210-CDDP both the V(max) and the K(m) were increased. That this was not a direct effect of EdSm on GST was shown in a cell-free system, where EdSm did not influence the GST activity nor could it act as a substrate for GST. Our results suggest that the synergistic combination of EdSm and cisplatin might be explained by EdSm switching off the cellular detoxification mechanism for cisplatin, i.e. by inhibition of de novo synthesis and subsequent depletion of GSH and GST.

Sparsomycin and its analogues: a new approach for evaluating their potency as inhibitors of peptide bond formation.

The ability of several sparsomycin analogues to inhibit peptide bond formation was studied in vitro. Peptide bonds are formed between puromycin (S) and the acetylPhe-tRNA of acetylPhe-tRNA/70 S ribosome/poly(U) complex (complex C), according to the puromycin reaction: [formula: see text] It was shown that the sparsomycin analogues, like sparsomycin itself, inhibit peptide bond formation in a time-dependent manner; they react with complex C according to the equation [formula: see text] where C*I is a conformationally altered species in which I is bound more tightly than in CI. The determination of the rate constant k(7) for the regeneration of complex C from the C*I complex allows evaluation of these analogues as inhibitors of peptide bond formation. According to their k7 values, these analogues are classified in order of descending potency as follows: n-pentyl-sparsomycin (4) > n-butyl-sparsomycin (3) approximately n-butyl-deshydroxy-sparsomycin (6) > benzyl-sparsomycin (2) > deshydroxy-sparsomycin (5) approximately sparsomycin (1) > n-propyl-desthio-deshydroxy-sparsomycin (7). The analogues with an aromatic or a larger hydrophobic side chain are stronger inhibitors of the puromycin reaction than are those with a smaller side chain or those lacking the bivalent sulfur atoms; replacement of the hydroxymethyl group with a methyl group does not affect the position of the compound in this ranking; compare the positions of compounds 1 and 3 with those of 5 and 6. In the case of compound 7, C*I adsorbed on cellulose nitrate disks was not sufficiently stable to allow examination by the method applied to the other analogues, probably due to a relatively large value of k7. This analogue showed also time-dependent inhibition, but after the isomerization of CI to C*I, the kinetics of inhibition become complex, and C*I interacted further with puromycin, either as C*I or after its dissociation to C*.

Peptide-derived transition state analogue inhibitors of thrombin; synthesis, activity and selectivity.

In a study to combine the transition state analogue concept with the principle of catalytic site spanning, a series of peptide-derived transition state analogue (TSA) inhibitors of thrombin has been synthesized and tested. In the sequence H-D-Phe-Pro-Arg-Gly-OH (2) the Arg-Gly amide bond has been replaced by three classes of transition state analogues, being the ketomethylene, the hydroxyethylene and the hydroxymethylene amide bond replacements. Compound 12a, in which the amide bond has been replaced by the ketomethylene group, was found to be the most potent thrombin inhibitor of the series studied. Subsequently, penta- and hexapeptide sequences with good affinity for thrombin were developed, i.e. H-D-Phe-Pro-Arg-Gly-Phe-OH (16) and H-D-Phe-Pro-Arg-Gly-Phe-Lys-OH (26). In these sequences the Arg-Gly amide bond was then replaced by the ketomethylene group. The resulting compounds 43a and 47a, respectively, were evaluated in vitro as inhibitors of thrombin and factor Xa. Compound 47a was found to be the most potent thrombin inhibitor of the series studied (Ki = 29 nM). The combination of the transition state analogue concept and the principle of peptide elongation (tetrapeptide-->hexapeptide) yields thrombin inhibitors of high potency and selectivity. The effects of these two alterations reinforce each other indicating a synergistic effect. This might be rationalized by entropy factors.

Schedule-dependent enhancement of antitumor activity of ethyldeshydroxy-sparsomycin in combination with classical antineoplastic agents.

The efficacy of the protein synthesis inhibitor ethyldeshydroxy-sparsomycin (EDSM) as a biochemical response modifier of several antitumor agents against L1210 leukemia and B16 melanoma is described. Seven drugs with different intracellular targets were selected for this combination study. Tumor implantation and drug treatment were both i.p., and the time interval between the administration of EDSM and the cytostatic agent was varied. Our results show that in the B16 tumor model EDSM is not able to potentiate any of these drugs, whereas antagonism is seen in combination with doxo-rubicin (DX). In the L1210 tumor model, however, no loss of activity is seen for this specific combination. The effect of the combination of cytosar (Ara-C), 5-fluorouracil (5-FU) or vincristine (VCR) with EDSM in the L1210 model is strongly time interval dependent. Loss of 5-FU antitumor activity is seen when EDSM is given 3 or 24 h after 5-FU; however, no effect is observed when EDSM is given 6 h after 5-FU. Enhancement of the 5-FU activity is not noticed. The VCR activity is potentiated when EDSM is given at least 6 h after VCR administration, which increases the antitumor response from 32 to > 60 days and the percentage survivors from 33 to 83% (p = 0.04). In combination with Ara-C, potentiation of antitumor activity is seen only when EDSM is given 24 h after Ara-C, which increases the antitumor response from 32 to > 55 days and the percentage survivors from 11 to 50% (p = 0.008). No modulatory effects are found when EDSM is combined with carmustine or DX.(ABSTRACT TRUNCATED AT 250 WORDS)

Antitumour activity and retinotoxicity of ethyldeshydroxy-sparsomycin in mice.

The colony formation in agar of human tumour xenografts was used as a test system to study the cytostatic activity of ethyldeshydroxy-sparsomycin (EdSm) at the cellular level. EdSm was additionally studied in vivo in human tumour xenografts and murine tumour models. EdSm showed a clear dose-response effect in vitro. At continuous exposure with 0.01 micrograms/ml, 2 out of 11 of the tumours responded (a gastric and a small cell lung carcinoma). At 0.1 mu/ml EdSm, the tumour response was 5/11 tumours and at 1 microgram/ml the compound was active in all tumours. The maximal tolerable doses of EdSm in vivo have been determined in non-tumour bearing CDF1 mice. In the intraperitoneally (i.p.) given multiple dose schedules the respective LD10 doses indicated that the tolerable cumulative dose increases when lower doses are given more frequently. This also enhances the antitumour activity in L1210 leukaemia to 172% T/C. On the other hand, continuous infusion strongly diminished the tolerable dose as well as the antitumour activity. EdSm was also active against i.p. inoculated P388 leukaemia (150% T/C), B16 melanoma (156% T/C), and RC carcinoma (197% T/C), and the subcutaneously (s.c.) inoculated L1210 (139% T/C) and RC (138% T/C). Absence of tumour responses was found in the following s.c. implanted murine tumours: M5076 sarcoma, osteosarcomas C22LR and CP369, and the LL carcinoma, as well as in the human tumour xenografts: LXFG 529, a non-small cell lung carcinoma; GXF 251, a gastric carcinoma; and FMa, an ovary carcinoma. Possible long-range retinotoxic effects of EdSm were investigated in tumour-bearing mice, cured after surviving treatment with LD50 doses of EdSm, by assaying the protein biosynthetic capacity of the retinal by assaying the ocular rhodopsin and opsin levels as parameters. In none of these cases could a significant reduction in either opsin or rhodopsin levels be measured and no changes were seen histologically.

Preclinical antitumor activity of ethyldeshydroxysparsomycin in combination with cisplatin.

The efficacy of cisplatin (CDDP) in combination with the protein synthesis inhibitor ethyldeshydroxysparsomycin (EDSM) has been tested in two tumor models at various schedules. Mice with L1210 leukemia or B16 melanoma were treated with CDDP alone or in combination with EDSM. Against L1210 leukemia, which is sensitive to CDDP, combinations elicited increases in life-span for all treatment schedules compared to those achieved with the corresponding dose of CDDP. Moreover, the combination of EDSM with this platinum compound yielded a cure rate > 80%, compared to < 35% for single CDDP treatment. Although the B16 melanoma is rather resistant to both CDDP and EDSM, combinations of these agents against B16 melanoma showed schedule dependent efficacy and in certain schedules significant therapeutic advantage over individual drug treatment, but cures were not observed. Our results suggest that EDSM has significant synergistic capabilities in both animal tumor models, but strong therapeutic enhancement of cisplatin efficacy is only seen when the tumor is sensitive to CDDP.

Concentration and sequence dependent synergism of ethyldeshydroxy-sparsomycin in combination with antitumor agents.

The modulating effect of ethyldeshydroxy-sparsomycin (EDSM), an inhibitor of ribosomal protein synthesis, on cytostatic agents was studied on cultured B16 melanoma cells using the microculture tetrazolium test (MTT). The data were analyzed for true synergism using the combination index and the median effect principle. The extent of cytotoxic drug interaction was influenced by the duration of drug exposure, the dose ratio, as well as the treatment schedule. When drug ratios were used, synergism was observed upon pre- and post-treatment in combination with cisplatin, cytosine arabinoside (Ara-C), methotrexate (MTX) and 5-fluorouracil (5-FU). The combination of a fixed dose of EDSM was synergistic with cisplatin, Ara-C, vincristine (VCR) and MTX, in the order MTX > Ara-C > VCR > cisplatin, while the combinations with doxorubicin, 5-FU and etoposide (VP-16) were shown to be antagonistic. These results suggest that only certain drugs and treatment schedules might be worthwhile for combination studies with EDSM in vivo and indicate a role for EDSM as modifier of antitumor responses in cancer chemotherapy.

Correlation of the in vitro cytotoxicity of ethyldeshydroxysparsomycin and cisplatin with the in vivo antitumour activity in murine L1210 leukaemia and two resistant L1210 subclones.

The cultured murine leukaemia L1210 cell populations used in the present study were derived from L1210 cells that had been grown in vivo. Subclones resistant to sparsomycin (L1210/Sm) or cisplatin (L1210/CDDP) were also developed in vivo. The doubling times of the cultured cell populations were identical. Fractions surviving after drug treatment in vitro were determined by colony formation in soft agar. The results, based on the differential sensitivity of the cell populations to ethyldeshydroxysparsomycin (EdSm) and CDDP, indicated that after a short exposure, cultured L1210/CDDP cells were cross-resistant to EdSm. L1210/Sm cells, however, were not cross-resistant to CDDP. The results obtained in cultured cell populations were confirmed in vivo. CD2f1 mice bearing i.p. implants of 1 x 10(5) tumour cells were given EdSm or CDDP and a combination of the two agents. Drugs were given once daily every 4 days for 3 doses starting at 24 h after tumour implantation. Treatment of mice bearing L1210/wt leukaemia with combined EdSm and CDDP caused strongly synergistic antitumour activity. In animals bearing the two resistant subclones, however, combined drug treatment did not improve the antitumour activity. The corresponding median survival of mice receiving combined drug treatment was 60 days in each group containing 6 mice bearing L1210/wt, with 4-6 cures being noted; 19 days in animals harbouring L1210/Sm, with 2 cures being recorded among 6 mice; and 11 days in mice bearing L1210/CDDP, with no cure being obtained. The results of this study indicate that the synergism resulting from combined treatment with CDDP and EdSm is a function of the cellular properties of the target tumour-cell populations and is independent of host factors.

Potentiation of cisplatin antitumour activity by Ethyldeshydroxy-Sparsomycin in L1210 leukemia.

The combination of Ethyldeshydroxy-Sparsomycin (EdSm) with cisdiamminedichloroplatinum(II) (CDDP) caused significant antitumour activity against murine L1210 leukemia. Although single drug treatment by cisplatin generated some cures, all schedules of combined treatment, using nontoxic doses of EdSm (5mg/kg) and cisplatin (3 mg/kg), resulted in the cure of 4 to 6 mice in each group consisting of 6 mice. No differences in antitumour activity were observed between pretreatment, simultaneous treatment or posttreatment of cisplatin with EdSm. Increasing the number of tumour cells implanted i.p. diminished the antitumour effect of both EdSm as well as CDDP, but not for the drug combination. Changing the route of administration from i.p. to i.v. for one of the drugs out of the combination resulted in loss of antitumour activity.

Biochemical and kinetic characteristics of the interaction of the antitumor antibiotic sparsomycin with prokaryotic and eukaryotic ribosomes.

Using 125I-labeled phenol-alanine sparsomycin, an analogue of sparsomycin having higher biological activity than the unmodified antibiotic, we studied the requirements and the characteristics of its interaction with the ribosome. The drug does not bind to either isolated ribosomal subunits or reconstituted whole ribosomes. For sparsomycin binding to 70S and 80S ribosomes, the occupation of the peptidyltransferase P-site by an N-blocked aminoacyl-tRNA is a definitive requirement. The sparsomycin analogue binds to bacterial and yeast ribosomes with Ka values of around 10(6) M-1 and 0.6 x 10(6) M-1, respectively, but its affinity is probably affected by the character of the peptidyl-tRNA bound to the P-site. Chloramphenicol, lincomycin, and 16-atom ring macrolides compete with sparsomycin for binding to bacterial ribosomes, but streptogramins and 14-atom ring macrolides do not. Considering the reported low affinity of puromycin for bacterial ribosomes, this antibiotic is also a surprisingly good competitor of sparsomycin binding to these particles. In the case of yeast ribosomes, blasticidin is a relatively good competitor of sparsomycin interaction, but anisomycin, trichodermin, and narciclasin are not. As expected, puromycin is a poor competitor of the binding in this case. The results from competition studies carried out with different sparsomycin analogues reveal, in some cases, a discrepancy between the drug ribosomal affinity and its biological effects. This suggests that some intermediate step, perhaps a ribosomal conformational change, is required for the inhibition to take place.

Chemical, biochemical and genetic endeavours characterizing the interaction of sparsomycin with the ribosome.

Sparsomycin interaction with the ribosome and characteristics of the drug binding site in the particle were studied using chemical modification of the drug, affinity labeling methods and isolation of drug resistant mutants. The structure-function relationship studies, performed with a large number of drug derivatives, indicate that the drug interacts with the ribosome by its western and eastern moieties. The uracil ring, in the western end of the drug molecule, probably forms hydrogen bonds with the rRNA, while the apolar CH3-S-CH3 group in the eastern end interacts with a hydrophobic ribosomal domain that affinity labeling results seem to indicate is formed by protein. An increase in lipophilicity in this part of the antibiotic results in a dramatic increase in the inhibitory activity of the drug. The sparsomycin binding site is not accessible in free ribosomes, but the presence of an N-blocked amino acyl-tRNA at the P-site turns the particles capable of reversible interaction with the drug. After failure using Escherichia coli, a sparsomycin-resistant mutant was obtained by direct mutagenesis on Halobacterium halobium, a species with a unique copy of rRNA genes, stressing the role of rRNA on the drug interaction site.

Interaction of the antibiotic sparsomycin with the ribosome.

Phenol-alanine-sparsomycin, a derivative of sparsomycin carrying a p-hydroxy-benzyl function easily labeled by iodination, has been used to study the interaction of this drug with the ribosome. Our study indicated that the binding of the drug to the ribosome is sensitive to trichloracetic acid and is equally affected by disintegration of the particle after RNase and protease treatments. The ribosome is not irreversibly inactivated, and the chemical structure of the drug is not affected by interaction with the particle. These data are not compatible with the proposed covalent association of sparsomycin with the ribosome by G. A. Flynn and R. J. Ash (Biochem. Biophys. Res. Commun. 114:1-7, 1983); therefore, the antibiotic must inhibit protein synthesis through a reversible interaction with the ribosome.

Pharmacokinetics of the antitumor antibiotic n-pentyl-sparsomycin in beagle dogs.

N-pentyl-sparsomycin (PSm) is a lipophilic analogue of sparsomycin (Sm), which is a well known inhibitor of protein synthesis. This compound was selected for preclinical pharmacokinetic studies because of its high in vitro and in vivo antitumor activity. In this study in which the drug was evaluated in beagle dogs under anaesthesia, the drug concentrations in plasma, urine and bile samples were determined using high performance liquid chromatography (HPLC). Plasma protein binding was approximately 54%. The mean t1/2 beta was 0.2 hours (12 minutes) and t1/2 tau was 0.75 +/- 0.1 hours (45 +/- 6 minutes). During continuous infusions up to 5.25 hours, the steady state was reached in 3 out of 6 experiments, suggesting that in some cases the real t1/2 tau was longer than measured. PSm was actively reabsorbed from the renal tubuli. This process was saturable at the higher doses. Tubular reabsorption played only a minor role in pharmacokinetics as most of the drug (67%) was eliminated by the non-renal clearance. The non-renal clearance was saturable at higher doses of PSm and was the reason for non-linearity of pharmacokinetics.

Novel sulfated polysaccharides: dissociation of anti-human immunodeficiency virus activity from antithrombin activity.

Novel sulfated polysaccharides, sulfated bacterial glycosaminoglycan (Org 31581) and chemically degraded heparin (Org 31733), have proved to be potent and selective inhibitors of human immunodeficiency virus (HIV) in vitro. Their 50% inhibitory concentrations for HIV type 1 (HIV-1) in MT-4 cells were 0.67 and 0.52 micrograms/ml, respectively. These values are comparable to those obtained for dextran sulfate and standard heparin (0.39 and 0.89 micrograms/ml, respectively). Org 31581 and Org 31733 showed much less antithrombin activity than did dextran sulfate and standard heparin. These results indicate that the anti-HIV activity of sulfated polysaccharides can be dissociated from their antithrombin activity. Org 31581 and Org 31733 were equally inhibitory to HIV-2 and HIV-1 and were also inhibitory to the replication of human cytomegalovirus. Syncytium formation, induced by cocultivation of MOLT-4 (clone 8) cells with chronically HIV-1-infected HuT 78 cells, was also inhibited by Org 31581. As previously demonstrated with dextran sulfate and heparin, both Org 31581 and Org 31733 blocked virus adsorption to the host cells. These compounds offer great promise as candidate drugs for the chemotherapy of HIV infections.

In vitro and in vivo anticancer activity of mitozolomide and sparsomycin in human tumor xenografts, murine tumors and human bone marrow.

The colony formation in agar of human tumor xenografts, of murine tumors and of human bone marrow was used as a test system to determine the in vitro activity of the two novel cytostatic agents, mitozolamide and sparsomycin. Mitozolomide was additionally studied in vivo in nine human tumor xenografts. The comparison of in vitro/in vivo activity allows an assessment of the relevant in vitro dose based on in vivo pharmacological behavior of a compound. Both compounds showed clear dose/response effects in vitro. A dose of 3 micrograms/ml mitozolomide, given by continuous exposure, was active (colony number of test less than 30% of the control group) in 12/42 (29%) human tumor xenografts as well as in the four murine tumors, P388, L1210, B16 melanoma and colon carcinoma 38, whereas the two human bone marrows showed no significant suppression of the ability to form colonies in culture. The comparison of in vitro with in vivo activity suggests that the in vitro dose of 3 micrograms/ml corresponds best to the activity observed in animal experiments. The highest activity was observed in small-cell cancer of the lung (4/5), followed by melanomas (2/7) and non-small-cell cancer of the lung (2/9). Furthermore, activity was found in a cancer of the large bowel, stomach, breast and in one sarcoma. In the treatment of nine human tumor xenografts growing subcutaneously in nude mice, mitozolomide effected a complete or partial remission in 6 out of 9 tumors. In comparison to standard drugs mitozolomide is one of the most effective compounds in these tumors. These data indicate that mitozolomide possesses potent broad-spectrum activity in human tumor xenografts. Sparsomycin (0.1 micrograms/ml, continuous exposure) was active in 11/46 (24%) human tumor xenografts and in 4/5 of the murine tumors, whereas the colony-forming capacity of four human bone-marrows showed no inhibition, suggesting that this dose level may be the relevant in vitro dose. However, the high in vitro activity in murine tumors is incompatible with the in vivo activity. In mice the only responsive tumor was leukemia P388, whereas the L1210, B16 melanoma and colon carcinoma 38 were resistant. At the dose level of 0.03 microgram/ml only 3/30 (10%) of the human tumor xenografts were sensitive. In an earlier clinical phase I study the dose-limiting adverse effect was eye toxicity and not bone-marrow suppression.(ABSTRACT TRUNCATED AT 400 WORDS)