In vitro cytotoxicity patterns of standard and investigational agents on human bone marrow granulocyte-macrophage progenitor cells.

Inhibitory concentrations (ICs) against human bone marrow granulocyte-macrophage colony forming cells (GM-CFC) were established for 26 cancer chemotherapy agents, including seven investigational agents by ten day exposure. Each drug was tested at four or more concentrations to generate reliable survival curves. The analysis of the survival curves produced three patterns according to which drugs were classified: class A drugs had a shouldered curve with terminal exponential kill of GM-CFC, class B drugs produced initial exponential component followed by a plateau, and class C drugs produced linear curves. These categories provide the relationship between drug concentration and cytotoxicity, e.g., the cytotoxicity of class B drugs, after initial kill, did not increase in spite of serial doubling of concentrations whereas the class C drugs had proportional killing with two-fold concentration increment. A number of drugs were active at in vitro concentrations of less than or equal to 0.01 microgram ml-1 and caused log reduction of GM-CFC with an approximate concentration of 0.0001 microgram ml-1. Drugs known to require in vivo bioactivation, namely dacarbazine, procarbazine, and ifosfamide were active at high concentrations (greater than 10.0 micrograms ml-1). We propose that for myelosuppressive agents the GM-CFC provides a useful biologic reference to determine in vitro cut off concentrations to be utilized for drug screening. For nonmyelosuppressive agents, however, it may be suboptimal.

The human tumor stem cell assay revisited.

The human tumor stem cell assay (HTSCA) is a bilayer soft agar system for growing fresh human tumor specimens in vitro to determine drug sensitivity and improve our understanding of tumor biology. Recent clinical correlations of 60% accuracy for predicting a positive clinical response and a 90% accuracy for predicting a lack of response to therapeutic agents suggest promising clinical usefulness. However, the clinician should be aware of the assay's inherent pitfalls, such as heterogeneity of the tumor specimen, inability to obtain pure single-cell suspensions, low cloning efficiency, unusual drug dose-dependent survival curves, uncertain validity of in vitro pharmacology, non-standardized criteria for in vitro sensitivity, and the variability of in vitro results. A brief summary of the concepts, potential, and limitations of this assay are discussed.

Normalization of in vitro sensitivity testing of human tumor clonogenic cells.

The use of normal bone marrow (granulocyte-macrophage colony-forming units) as a point of reference to normalize the in vitro activities of anticancer agents has been investigated. The cytotoxic effects of four substituted anthraquinone derivatives, and of vinblastine on myeloid progenitors of different donors were reproducible up to a cell kill of approximately 60%. Equitoxic in vitro concentrations for normal bone marrows did not correlate with in vivo pharmacokinetic concentrations of these drugs. Breast tumor progenitor cells of 46 specimens were more sensitive than were bone marrow progenitors to the anthraquinone derivatives in 26 to 39% of instances, ratios which are similar to the clinically observed response rates of patients with breast carcinoma to these agents. Tumors were either sensitive or resistant to all four drugs in 68% (10 tumors were more sensitive, and 21 tumors were less sensitive than normal bone marrow); but in 32% of instances there were differences in tumor sensitivity for the four drugs, and the assay could select one to three drugs for which the tumor sensitivity was greater than that of bone marrow. Correlations of in vitro sensitivity and of clinical response to single agent treatments were determined in 21 patients, and the concordance was 71%. The value of the assay in predicting clinical response ranked best for sensitivity determinations within the normalized dose ranges, when testing within three different dose ranges was compared in a group of six patients. The concordance was higher in the small (1 or 2 metastatic sites) than in the large (greater than or equal to 3 metastatic sites) tumors (85 versus 50%), indicating a confounding influence of tumor load on the ability of the assay to predict efficacy of treatment. A rule of thumb is proposed for altering the in vitro sensitivity test results for large tumors that improves the overall concordance to 90%.

Treatment of cultured human colon carcinoma cells with fluorinated pyrimidines.

The shape of the initial part of the dose-dependent response curve of LoVo cells, an established human colon carcinoma cell line, exposed for 1 hr to graded concentrations of 5-FU depended on the medium supplement, i.e., fetal calf serum (FCS), in which the cells were treated and subsequently incubated for colony-formation. At concentrations of 50--100 micrograms/ml (equivalent to peak plasma levels following an in vivo bolus dose of 15 mg/kg) cell kill was completely prevented by FCS. The serum did not contain thymidine (TdR) but had significant amounts of uridine (UR). When 5-FU was delivered in dialyzed FCS, concentrations of 50--100 micrograms/ml achieved only a modest 15% cell kill after 1 hour treatment. Regardless of medium supplement, the killing effect of 5-FU did not increase beyond concentrations greater than 2,000 micrograms/ml. Increasing the exposure interval dramatically increased the killing of LoVo cells by 5-FU, although the effects of medium supplement on the degree of cell survival persisted for about 12 hours. Virtually all of the incorporated 5-FU was transformed into 5-FUR, and a very small proportion eventually was incorporated into nucleic acids, suggesting that the killing effect of 5-FU on LoVo cells is mediated mostly by ribosidation and not by conversion into the deoxyribonucleoside. This conclusion is supported by the failure of 5-FUdR to kill LoVo cells after a treatment interval of one hour, even at concentrations of 5000 micrograms/ml; yet after the same exposure interval, 5-FUR effectively killed cells at concentrations of 50--100 micrograms/ml. TdR afforded no protection from cell kill by 5-FU. In contrast, UR was capable of protecting LoVo cells from the lethal effects of both 5-FU and 5-FUR even at concentrations as low as 10 micrograms/ml. Ftorafur exposed to LoVo cells for 1 hour had a slight killing effect (about 20--25%) at concentrations ranging up to 2000 micrograms/ml. Although the lethal effect of ftorafur was slightly increased after longer periods of incubation, it failed to reach 90% even after intervals of 48 hours. The results on cellular sensitivity that we obtained for LoVo cells treated with various fluorinated pyrimidines differ substantially from those of other investigators who used different methods to assess cell killing on nonhuman and noncolonic cell systems. The predictive relevance of these data as compared to those obtained in other systems is justified by the suboptimal results with these agents in clinical practice.

Cellular response to treatment with 4-(3-(2-chloroethyl)-3-nitrosoureido)-cis-cyclohexanecarboxylic acid, a water-soluble nitrosourea derivative.

The lethal effects of 4-(3-(2-chloroethyl)-3-nitrosoureido)-cis-cyclohexanecarboxylic acid (cis-acid), a water-soluble nitrosourea derivative, were investigated on a human lymphoma cell line. The survival of asynchronous cells exposed to increasing concentrations of the drug was characterized by a threshold exponential curve (Do = 20 microgram/ml; Dq = 20 microgram/ml, 1 hour) similar to that of other nitrosourea derivatives. cis-Acid exerted its main killing effect on cells in early S and in late G2 phase. Cells in mid S and early G1 phase were tenfold more resistant. Changes in survival response as a function of cell cycle stage were reflected primarily by changes in the extent of the shoulder region of the survival curve. In contrast to other nitrosoureas, the lethal effectiveness of cis-acid in solution was stable and the drug could sterilize large numbers of cells in short periods of time. Another important major difference observed for cis-acid with respect to classic nitrosourea derivatives was the capacity of treated cells to recover from sublethal and potentially lethal damage. Our studies have shown that cis-acid is as effective in killing cultured human lymphoma cells as other nitrosoureas, but possibly with a mechanism different from that of these compounds. The major shortcoming noted for cis-acid, namely the capacity of treated cells to recover from drug-induced damage, is offset by the relatively long stability of its killing effect. This, and the fact that cis-acid can be administered in an aqueous solution, make this agent an appealing compound for clinical trials.

Survival and cycle-progression delay of cultured human lymphoma cells treated with 1-propanol, 3,3'-iminodi-, dimethanesulfonate (ester), hydrochloride (Yoshi 864).

Asynchronous human lymphoma cells treated for 1 hour with increasing concentrations of 1-propanol, 3,3'-iminodi-, dimethanesulfonate (ester), hydrochloride (Yoshi 864) revealed a shouldered survival curve typical of the effects of alkylating agents and of ionizing radiation on this cell line. Yoshi 864 was unstable under the conditions of treatment, its killing effect being reduced by 50% after only 4 hours. Synchronized cells showed stage-dependent sensitivity: early-S, late-G2, and late-G1 phases were the most sensitive while mid- and late-S and early-G2 phases were relatively insensitive. Yoshi 864 induced a concentration- and incubation time-dependent delay in the transit of asynchronous cells through G2 phase, with maximum accumulation values obtained after 12 hours of incubation with 100 mug/ml. This effect was largely reversible and no further kinetic changes were noted in the progeny of treated cells. Incubation of synchronized cells for 1 hour with 100 mug/ml demonstrated a block in G2, the manifestation of which during the lifespan of the treated cell or in its immediate progeny was cell-cycle dependent. Thus, cells treated in G1, early-, and mid-S phases showed a delay in the subsequent G2 phase while cells treated in late S and in G2 manifested this effect in the G2 phase of the immediate progeny. There was no correlation between this blocking effect in G2 with cell survival assessed by colony formation. Yoshi 864, although a rather inefficient killing drug, may represent a useful chemical synchronizing agent.

Combination chemotherapy in vitro with cis-dichlorodiammineplatinum(II).

Cis-dichlorodiammineplatinum(II) (DDP) was paired with 11 other chemotherapeutic drugs in a search for possible synergistic pairs. Cell lethality was investigated by the colony-formation technique utilizing a long-term culture of human lymphoma cells. Synergistic effects were noted for pair combinations with all alkylating agents, bleomycin, adriamycin, camptothecin, and cytosine arabinoside. Additive effects were recorded for combinations with prednisolone and VP-16-213, while DDP reduced the capacity of cells to absorb sublethal damage induced by alkylating agents.

Comparison of in vitro methods to determine drug-induced cell lethality.

In proliferating cell populations, the inability to reproduce indefinitely is the only relevant criterion to assess cell lethality. The in vitro colony formation technique (CF) used to determine reproductive death is, however, too slow and has several technical limitations. For finding suitable, more rapid techniques that assessed drug-induced cell killing, a human lymphoma cell line was exposed in vitro to increasing concentrations of adriamycin, bleomycin, and 1,3-bis(2-chloroethyl)-1-nitrosourea for 1 hr. Survival was assayed immediately after treatment and at regular intervals thereafter. Data from CF were compared to those resulting from the following tests: doubling time, labeling index, dye exclusion, 51Cr release, and rate of [3H]thymidine uptake (scintillation index). Dye exclusion and 51Cr release failed to demonstrate any killing effect for the 3 drugs. The percentage of killing calculated from doubling time determinations, although dose dependent, failed to correlate with CF. Scintillation and labeling index values displayed similar temporal fluctuations but were not clearly dose dependent and did not correlate with CF. Thus, CF appears as the most reliable, dose-dependent index of cell lethality. Tests that measure metabolic death grossly overestimate or underestimate killing activity induced by 3 of the most effective antitumor drugs.