Search

Human acute non-lymphocytic leukemia (ANLL) cell lines have been valuable models for studying leukemic transformation and mechanisms regulating blood cell growth and differentiation. Until recently, all human leukemia lines were grown in fetal bovine serum and were not dependent on exogenous hemopoietic growth factors. Since most primary human ANLL cells require growth factors for in vitro cell growth, the behavior of growth factor-independent cells may not be representative of most acute leukemias. Recently, factor-dependent cell lines were established from patients with acute leukemias which are absolutely dependent on growth factors for in vitro proliferation and survival. These cells were derived from different primary leukemia phenotypes, and exhibit heterogeneous growth and differentiation responses. Like the murine growth factor-dependent cell lines which preceded them, these new human lines are important models for growth factor signal transduction as well as proliferative responses and differentiation. Factor-dependent human cell lines have already provided important insights into potential mechanisms of human blood cell transformation, and identification of cellular and genetic abnormalities responsible for the leukemia phenotype.

Four second-generation Illudin analogues were synthesized and tested for antitumor activity using a metastatic lung carcinoma xenograft model resistant to conventional antitumor agents. One analogue, the parent illudofulvene-derivative called Acylfulvene, inhibited xenograft primary tumor growth and prolonged life span of tumor-bearing animals when administered i.p. or i.v. The efficacy of Acylfulvene exceeded that of mitomycin C, cisplatin, paclitaxol, the parent compound Illudin S, and an earlier analogue, dehydroilludin M. Promising features of this new analogue are: (a) the retention of in vitro activity against a variety of mdr tumor phenotypes including gp170+, gp150+, GSHTR-Pi, topoisomerase I, and topoisomerase II mutants; and (b) an apparent selective cytotoxicity toward cells deficient in either ERCC2 or ERCC3 DNA helicase activity.

In preclinical in vitro and in vivo systems, mAbs to human transferrin (Tf) receptors blocked iron uptake from Tf and showed antitumor activity. However, Tf receptors are also displayed by normal tissues, and a large, soluble pool of circulating serum Tf receptors has been detected. We report results of a Phase Ia trial of IgA monoclonal anti-Tf receptor antibody 42/6. Twenty-seven patients with advanced refractory cancer received 33 treatments with 42/6 administered as a 24-h infusion at doses ranging from 2.5 to 300 mg/m2. 42/6 was generally well tolerated, although one patient receiving a second treatment experienced an allergic-type response associated with a human antimouse antibody response. Three patients with hematological cancers showed mixed tumor responses; there were no partial or complete remissions. Peak serum levels of antibody were obtained at the termination of the 24-h infusion. At doses/=25 mg/m2, there was a linear relationship between the 42/6 dose and average peak serum 42/6 levels ranging from1 to 36 microgram/ml. Serum Tf receptors showed a dose-dependent decrease during 42/6 infusion to 20-30% of baseline, and remained depressed for at least 48 h after terminating the infusion. Serum 42/6 levels rose in an inverse relationship to the drop in Tf receptors. 42/6 induced an increase in serum iron and Tf saturation consistent with blockade of peripheral iron uptake, and reduced Tf receptor display by bone marrow cells. Human antimouse antibody was detected in nine patients. Anti-Tf receptor antibody was well tolerated and mediated in vivo effects on iron uptake and Tf display. Antibody concentrations capable of inhibiting malignant blood cell growth were obtained without toxicity. This represents the first clinical trial of an IgA mouse mAb, and one of only a few trials in which an antibody reacting with a broad range of normal tissues has been administered. Additional clinical trials of anti-Tf receptor antibodies in blood cell cancers are indicated.

Six first-generation illudin analogs were tested for antitumor activity using in vitro cytotoxicity and in vivo xenograft models. One analog, dehydroilludin M, inhibited xenograft growth and prolonged life span of tumor bearing animals, when administered IP or IV, whereas the parent Illudin S compound was ineffective. The efficacy of dehydroilludin M in the MV522 lung carcinoma model exceeded that of 9 known anticancer drugs, and equaled that of mitomycin C. Dehydroilludin M retained the in vitro relative selective cytotoxicity for carcinomas and myeloid leukemia cell lines noted with the parent illudin compounds. In vitro cytotoxicity data predicted response of xenografts. Dehydroilludin M also retained the in vitro activity of the parent compounds against different multidrug resistant mdr cell lines.

The human lung carcinoma cell line MV522 was previously noted to produce extensive metastasis to the lungs, spleen and lymph nodes after subcutaneous transplantation into athymic nude mice. Animals eventually succumb to these metastases, and not primary tumor growth. The ability to produce extensive metastasis after a simple subcutaneous injection in 100% of animals (100 tested to date) would be an advantage when screening compounds for anticancer activity. To validate the utility of this xenograft model for testing anticancer agents, we tested the ability of 10 anticancer drugs to either inhibit primary tumor growth and/or prolong life span of MV522-bearing animals. Among the 10 antitumor conventional agents, only mitomycin C and taxol demonstrated primary tumor growth inhibition. Mitomycin C produced a mild increase in median life span of 41% to 63%, while taxol had inconsistent effects. The metastatic MV522 carcinoma model appears to reflect clinical resistance of primary non-small cell lung cancer to conventional chemotherapeutic agents and should be useful for testing new anti-cancer drugs.

Illudins, novel natural products with a structure unrelated to any other known chemical, display potent in vitro and in vivo anti-cancer activity against even multi-drug resistant tumors, and are metabolically activated to an unstable intermediate that binds to DNA. The DNA damage produced by illudins, however, appears to differ from that of other known DNA damaging toxins. The sensitivity pattern of the various UV-sensitive cell lines differs from previously studied DNA cross-linking agents. Normally, the ERCC1- (excision repair cross complementing) and ERCC4-deficient cell lines are most sensitive to DNA cross-linking agents, with ERCC2-, ERCC3- and ERCC5-deficient cell lines having minimal sensitivity. With illudins the pattern is reversed, with ERCC2 and ERCC3 being the most sensitive. The sensitivity to illudins in complementation groups 1 through 3 is due to a deficiency of the ERCC1-3 gene products, as cellular drug accumulation studies revealed no differences in transport capacity or total drug accumulation. Also, a transgenic cell line in which ERCC2 activity was expressed through an expression vector regained its relative resistance to the illudins. The EM9 cell line, which displays sensitivity to monoadduct producing chemicals, was not sensitive. Thus, excision repair is involved in repair of illudin-induced damage and, unlike other anti-cancer agents, the involvement of ERCC2 and ERCC3 helicases is critical for repair to occur. The requirement for ERCC2 and ERCC3, combined with the finding that ERCC1 but not ERCC2 is upregulated in drug-resistant tumors, may explain the efficacy of illudins against drug-resistant tumors. The inhibition of DNA synthesis in cells within minutes after exposure to illudins at nanomolar concentrations may be related to the finding that the ERCC3 gene product is actually the p89 helicase component of the BTF2 (TFII) basic transcription factor and the high sensitivity of ERCC3-deficient cells to illudins.

The human factor-dependent leukemia cell line UCSD/AML1 contains the t(3;3) (q21;q26) characteristic of the syndrome of acute leukemia with high platelets. The human homologue of the murine leukemia oncogene evi-1 was recently localized to chromosome 3q24-3q28 and transcription of evi-1 is a frequent event in mouse-retrovirus-induced leukemias (17). To determine whether translocations near human 3q24 might induce similar genetic changes, we examined and compared evi-1 and c-myc expression and regulation in UCSD/AML1 cells. Steady-state evi-1 transcripts were detected in UCSD/AML1 and murine leukemia M1 cells, but were not present in HL60 or Namalwa human leukemia cells. Transcription assays showed the evi-1 gene was actively transcribed in UCSD/AML1, but not HL60 nuclei. Evi-1 transcript sizes and half-life were similar in UCSD/AML1 and human HEC-1B carcinoma cells which express evi-1 transcripts, but do not have abnormalities involving chromosome 3. An alternative splice site detected by polymerase chain reaction was present in transcripts from both cell lines. Regulation of evi-1 RNA in UCSD/AML1 cells was similar to that of actin transcripts in response to cycloheximide or phorbol-ester-induced macrophage differentiation. After withdrawal of granulocyte/macrophage colony-stimulating factor (GM-CSF), evi-1, actin, and histone H3 transcripts declined in concert with exit from the cell cycle. Minor differences in rates of recovery were noted for these three genes after GM-CSF restimulation. In contrast, c-myc was expressed at high levels in UCSD/AML1 cells and showed evidence for specific regulation in response to cycloheximide, phorbol ester, and GM-CSF withdrawal and restimulation. These studies suggest the 3q translocation in UCSD/AML1 cells is associated with evi-1 transcription and expression of a potential transforming gene. In contrast to c-myc, evi-1 expression is minimally altered by biologically active chemicals or growth factor stimulation.

Intracellular ricin and immunotoxin trafficking has been difficult to study as only one to two cytosolic ricin A chain (RTA) molecules are sufficient to cause cell death. Previous studies (R.J. Youle and M. Colombatti, J. Biol. Chem., 262: 4676-4882, 1987) using anti-ricin hybridomas identified the secretory pre-Golgi as a critical site for RTA intoxication. We used ricin and RTA immunotoxins constructed with transferrin (TF) or anti-murine TF receptor antibody (RI7/217) to compare patterns of cytotoxicity and intracellular trafficking in anti-ricin hybridomas. Anti-RTA and anti-ricin B chain (RTB) hybridomas bound similar amounts of ricin and secreted comparable amounts of anti-ricin immunoglobulin. Anti-RTA hybridomas were 50- to 500-fold more resistant to ricin than nonsecretory and anti-RTB hybridomas, defining a ricin-resistant phenotype. All hybridomas expressed similar levels of surface TF receptors. RTA immunotoxins were constructed using human TF or RI7/217 and a disulfide linker. In protein synthesis inhibition assays, ricin-resistant hybridomas were manyfold more resistant to RI7/217-RTA than were ricin-sensitive hybridomas. In contrast, all hybridomas were equally sensitive to TF-RTA. Monensin increased ricin cytotoxicity minimally against all hybridomas, but dramatically increased RI7/217-RTA cytotoxicity in ricin-resistant and ricin-sensitive hybridomas in a way that abrogated the ricin-resistant phenotype. In contrast, monensin increased TF-RTA cytotoxicity equally in all hybridomas. Ammonium chloride had little effect on ricin or RI7/217-RTA cytotoxicity, but increased TF-RTA cytotoxicity against all hybridomas. Taken together, these results suggest that RTA molecules mediating cytotoxicity pass through an anti-RTA antibody-containing pre-Golgi compartment when bound to RTB or RI7/217, but not when bound to TF. Monensin abrogates the ricin-resistant phenotype when RTA is linked to RI7/217, but not RTB. This suggests that monensin alters RI7/217-RTA processing proximal to the pre-Golgi and that passage through the pre-Golgi may not be necessary for translocation of RTA to the cytoplasm. Ammonium chloride alters toxin cytotoxicity only when RTA is linked to TF, suggesting that only TF trafficks RTA through an acid-sensitive compartment prior to cytoplasmic translocation. With the addition of potentiating agents, each toxin studied showed a unique cytotoxicity profile against the anti-ricin hybridomas, demonstrating a dominant role of the cell binding ligand in intracellular toxin trafficking.

To test the feasibility of using the human epidermal growth factor receptor (EGFR) as a model for growth factor receptor action in human hematopoietic cells, we infected Burkitt lymphoma cells (Namalwa) with a recombinant amphotrophic retrovirus containing a thymidine kinase promoter-driven human EGFR complementary DNA and the neomycin resistance gene. Neomycin-resistant cells expressing surface EGFR were selected by cell sorting using anti-EGFR monoclonal antibody 225. The selected cells expressed a Mr 170,000 protein immunoprecipitated by monoclonal antibody 225 and apparently identical to EGFR from A431 carcinoma cells. Infected Namalwa cells expressed 42,000 epidermal growth factor (EGF) binding sites/cell, and Scatchard analysis showed two affinities (Kd approximately 5 nM and approximately 0.5 nM). EGFR autophosphorylation was detected using antiphosphotyrosine antibodies after 5 min exposure to EGF. EGF binding induced rapid EGFR internalization (t1/2 = 9 min) and mobilization of transferrin receptors to the cell surface within 1 min. In fetal bovine serum-containing and serum-free cultures, EGF did not stimulate Namalwa cell proliferation. However, in the presence of 1.25% dimethyl sulfoxide (DMSO), EGF caused a dose-dependent increase in DNA synthesis. DMSO induced a cell cycle block in G1, which was partially reversed by EGF. DMSO induced changes in some B-cell markers suggesting cellular differentiation and increased surface EGF receptor number. Cells grown in DMSO and EGF were established as an EGF-dependent cell line for greater than 12 weeks, whereas cells grown in DMSO without EGF died within 1-2 weeks. Namalwa cells expressing EGFR demonstrated more rapid tumor growth in athymic mice. These studies demonstrate expression of functional EGFR mediating early biochemical and growth responses in a human hematopoietic cell, and indicate that EGFR can be used as an effective model in human hematopoietic cells. Results using DMSO are consistent with studies in other human leukemia cells indicating that agents inducing differentiation can restore growth factor dependence in previously factor-independent leukemia cells.

Thirty-two murine monoclonal antibodies (MAbs) against the external domain of the human transferrin (Tf) receptor have been obtained using human Tf receptor glycoprotein produced in a baculovirus expression system as immunogen. The MAbs were tested separately, and in combination, for their ability to inhibit growth of CCRF-CEM leukemic T-cells in tissue culture. One MAb, D65.30, an IgG1, inhibited growth of CCRF-CEM cells as effectively as the anti-Tf receptor MAb 42/6, an IgA, previously found to have the highest antiproliferative activity in this assay. Some combinations of two or more MAbs inhibited the in vitro growth of CCRF-CEM much more effectively than single MAbs. Eleven IgG1 anti-Tf receptor MAbs, when combined individually with D65.30, increased the inhibition of CCRF-CEM growth from approximately 65% to greater than 90% in a 7-day growth assay. Similarly, many IgG MAbs in combination with 42/6 also inhibited CCRF-CEM growth by greater than 90%. The growth-inhibitory effects of certain combinations of MAbs were clearly synergistic, because either one or both MAbs tested separately were inactive. These pairs of MAbs also inhibited the growth of HL-60 and KG-1 leukemic cells by greater than 90% and partially inhibited the growth of K562 erythroleukemia and M21 melanoma cells, which are resistant to MAb 42/6. However, not all combinations of anti-Tf receptor MAbs were more effective; eight MAbs markedly antagonized the antiproliferative effects of D65.30, whereas 12 others had little or no effect. Preincubation of HL-60 cells with three different pairs of MAbs, D65.30 and A27.15, B3/25 and 42/6, and B3/25 and TR3A, inhibited subsequent colony formation by greater than 95%, demonstrating that their action is cytotoxic, not cytostatic. The antiproliferative activity of these pairs of MAbs correlates with their ability to block Tf-mediated 59Fe uptake and perturb Tf receptor expression. Treatment of nude mice bearing established s.c. CCRF-CEM tumors with a combination of MAbs D65.30 and A27.15 inhibited tumor growth and in some animals led to complete tumor regression. Each MAb administered separately was much less effective. We conclude that combinations of two or more anti-Tf receptor MAbs can interact synergistically to inhibit cell growth in vitro and tumor growth in vivo.

The Tf receptor is a prototypical cell surface nutrient receptor critical for growth of both normal and malignant blood cells. Because of the apparently universal requirement of growing blood cells for Fe, control of Tf receptor expression and function will continue to be a topic of considerable interest in experimental hematology and tumor cell biology. Studies of Tf receptor gene expression have already revealed unique patterns of gene regulation, and understanding control of Tf receptor gene expression during hemopoietic cell growth and differentiation may provide insights into factors controlling expression of other proliferation-related genes. Although much further study is required, early studies suggest that measuring serum Tf receptor and Tf-Tf receptor complexes may be a major advance in clinical assessment of erythropoiesis. Whether altering Tf receptor protein function or gene expression can be exploited to treat hematologic tumors remains unclear, but further studies are in progress.

Previous studies suggested tumor colony-forming cells (CFC) grown from xenografts might be useful as a preclinical, in vitro drug screen. To further evaluate this possibility, eight melanoma and six ovarian carcinoma xenografts were established from untreated patients and tested for in vitro CFC growth. For each tumor, linear relationships between cells plated and colony (30 cells or greater than 75 micron diameter) and cluster (10-30 cells or 50-75 micron) growth were observed. All eight melanomas grew sufficient colonies (greater than or equal to 30) for in vitro drug assessment, although four required hypoxic (pO2 = 40) incubation to reliably attain this level of growth. Only one in six of the ovary xenografts consistently grew enough colonies, and growth was not significantly improved by hypoxic incubation, or addition of luteinizing hormone, follicle-stimulating hormone, or steroid hormones. Cloning efficiencies (colonies + clusters/cells plated) for tumors demonstrating adequate growth ranged from 0.01% to 0.3%. For most tumors, no direct relationship was observed between characteristics of xenograft tumors (size) or their resulting cell suspensions (viabilities, cell yield) and CFC growth. Cell suspensions were incubated with a 3 log concentration of nine established chemotherapeutic agents. Resulting dose-effect curves were linear and showed no plateaus of drug effect. Analyzing 447 in vitro drug trials on six melanomas and one ovarian carcinoma, interexperiment variability was high. Cell lines were established from three xenografts using a low concentration of fetal bovine serum (1%), and also examined for in vitro drug sensitivity. Using both liquid culture isotope incorporation and a colony-forming assay, drug sensitivity profiles for the cell lines were nearly identical to those for parent xenograft CFC. However, assays performed using the cell lines were more reproducible than those using xenograft tissue. The authors conclude that tumor CFC can be reliably grown from melanoma xenografts, but in vitro drug assays using these xenografts are poorly reproducible. The xenografts are a resource for establishing cell lines, and drug assays performed using these lines are highly reproducible. Similarities in drug sensitivity profiles for parent xenograft CFC and derived cell lines suggest that, despite poor reproducibility, repetitive assays using melanoma CFC accurately reflect some properties of cells which sustain tumor cell growth.

Concurrent administration of sodium thiosulfate reduced the toxicity of cis-dichlorodiammineplatinum(II) (DDP) in a dose-related manner in mice. Sodium thiosulfate protected mice against an otherwise lethal dose of DDP (20 mg/kg), and reduced DDP-induced weight loss. Sodium thiosulfate (800 mg/kg) injected within 1 hour before or 1/2 hour after DDP blocked nephrotoxicity as measured by a rise in BUN, an increase in kidney weight, and medullary hemorrhage. In culture, sodium thiosulfate markedly reduced the toxicity of DDP to mouse colony-forming units. Concurrent injection of sodium thiosulfate partially reduced the antitumor activity of DDP. The therapeutic dose range of DDP was expanded, but the maximum increase in lifespan of mice bearing L1210 leukemia was reduced by 40%. Sodium thiosulfate offers the possibility of systemic protection against the cytotoxicity of regionally administered DDP in man.

The potential use of nude mouse xenografts as a source of human tumor tissue for preclinical assessment of drug activity was examined by a comparison of in vitro sensitivity to four anthracycline derivatives of eight xenografts maintained in nude mice and tumor colony-forming units (CFU) from the xenografts grown in agar. Results obtained in the two systems with doxorubicin (Dx) and a new, closely related derivative (epi-doxorubicin, epi-Dx) correlated well. In vivo tumor growth delay and maximum in vitro tumor CFU kill for these two drugs showed a significant correlation (r = 0.64, P less than .01). Separation of tumors into "sensitive" and "insensitive" tumor populations on the basis of maximum in vitro CFU kill also predicted the in vivo response to these two drugs in 88% of cases. Similar analyses performed on in vitro and in vivo results with two other new anthracyclines (4'-deoxy-doxorubicin, deoxy-Dx and 4'-O-methyl-doxorubicin, O-Me-Dx) showed a significant negative correlation between in vivo and in vitro results; the in vitro system failed to predict in vivo activity of these two drugs. No significant differences in in vitro activity against normal, granulocyte/macrophage progenitors (CFU-GM) or against various tumor CFUs were detected. Thus, the selectivity (activity against normal tissue compared with that against tumor) of the four drugs appeared equal. These data suggest that in vitro screening of drugs using tumor CFUs from nude mouse xenografts may predict the in vivo activity of drugs for which pharmacologic data are available, but illustrate the difficulties in attempting to predict the in vivo activity of new drugs for which no such data are available.

To evaluate the utility of nude mouse xenografts as preclinical drug screens, the activity of ten established chemotherapeutic agents was evaluated against seven melanoma and three ovarian carcinoma xenografts. Xenografts were established using primary explants from patients who had not received chemotherapy and serially passaged as sc tumors in nude mice. In vivo drug activities for dactinomycin, carmustine, vinblastine, melphalan, amsacrine, cisplatin, bleomycin, mitomycin, doxorubicin, and etoposide were evaluated by 4 weekly ip injections of 10% less than LD10 doses. Plots of relative tumor growth versus time were nearly log-linear. Analysis of in vivo activity was performed using percent control growth (treated/control tumor volume) and by calculation of a novel growth delay index obtained by fitting growth curves to a quadratic regression model. Both modes of data analysis identified alkylating agents (melphalan, carmustine, and mitomycin) as the most active drugs against human melanomas. Melphalan, mitomycin, and cisplatin showed the greatest activity against ovarian xenografts. However, complete tumor regressions were noted only with melphalan, mitomycin, and cisplatin against a single ovarian tumor xenograft. Correlation analysis suggested xenograft tumor growth rate was an important determinant of drug response. These results suggest that preclinical, new-drug screening with melanoma xenografts would identify drugs such as alkylating agents as active, and may not provide an advantage over murine leukemia screens. However, screening with ovarian xenografts may more closely reflect clinical drug activity. Criteria for detecting active drugs in such systems are discussed.

PALA is thought to inhibit an early step in de novo pyrimidine synthesis, causing depletion of intracellular pyrimidine nucleotides. Dipyridamole, a nucleoside transport inhibitor which can block restoration of nucleotide levels via the salvage pathway, was tested for its ability to augment the cytotoxicity of PALA against normal and malignant human cells in vitro. At the clinically relevant concentration of 1 microM, dipyridamole increased the cytotoxicity of PALA against a melanoma, a colon carcinoma, a promyelocytic leukemia (HL-60), and normal marrow (CFU-GM) in clonogenic assays. Dipyridamole produced 50% inhibition of uridine uptake in these cells at concentrations of less than 0.1 microM and reduced the LD50 of PALA by approximately 50% in mice. These results indicate that dipyridamole can markedly potentiate the activity of PALA in vitro and in vivo.

Benzaldehyde (BA), an agent extracted from figs, is reported to have antitumor activity in vitro against a variety of experimental tumors and in vivo against refractory human neoplasms. We employed recently developed in vitro techniques to examine the effect of BA on growth of malignant human cells, and examined its effects in vivo against two human tumor xenografts established from primary specimens. BA showed dose-dependent inhibition of HL60 promyelocytic leukemia cells and normal human granulocyte/macrophage colonies. It showed no in vitro activity against either KG-1 myeloid leukemia cells or chronic lymphocytic leukemia cells grown in colony culture. BA failed to inhibit growth of either HL60 or KG-1 cells in liquid culture, and did not induce differentiation of HL60 cells. When tested against human tumor colony-forming cells from 30 patients with solid tumors, inhibition of colony growth greater than 70% was seen in six patients (20%). BA failed to inhibit the in vivo growth of either the T222 epidermoid carcinoma xenograft or the T380 ovarian carcinoma xenograft. We conclude that BA lacks significant activity against most human tumors tested in these experimental systems.

To investigate the feasibility of using tissue obtained from human tumor xenografts for in vitro screening of antineoplastic agents, we grew human tumor colony-forming units (CFU) in semisold agar from xenografts serially passaged in nude mice. Growth of human tumor CFU was accomplished from nine xenografts representing five different histological tumor types (ovarian carcinoma, adenocarcinoma of the colon, malignant melanoma, epidermoid carcinoma of the lung, and malignant astrocytoma). Cloning efficiency ranged from 0.04 to 0.1% and showed significant variability both between tumor types and between individual animals bearing the same type of xenograft. A high percentage of tumor CFU was in S phase [47 +/- 20% (S.D.)] as determined by the thymidine "suicide" technique. The number of tumor CFU observed increased linearly with increasing numbers of cells plated. In vitro drug sensitivity of the tumor CFU was assessed to Adriamycin, cis-platinum, and melphalan. The patterns of drug sensitivity were found to be reproducible and stable over a period of 9 months. Drug sensitivity curves to Adriamycin for five xenografts representing four tumor types showed complex patterns with plateau portions similar to those described for tumor CFU from primary tumors. The rank order of sensitivity of the tumors was compared to that of normal granulocyte-macrophage progenitors and, with the exception of the melanomas, was found to correlate well with clinical experience (order of sensitivity = colon less than ovary less than bone marrow). Growth of human tumor CFU from xenografts represents a reproducible and stable means for the study of the biology of tumor CFU and has potential applications as a means for screening new anticancer agents.