Origin of multidrug resistance in cells with and without multidrug resistance genes: Chromosome reassortments catalyzed by aneuploidy

Abstract

Cancer cells and aneuploid cell lines can acquire resistance against multiple unrelated chemotherapeutic drugs that are over 3,000-fold those of normal levels and display spontaneous resistances up to 20-fold of normal levels. Two different mechanisms were proposed for this phenotype: (i) classical mutation of drug metabolizing genes or (ii) chromosome reassortments, catalyzed by cancer- and cell line-specific aneuploidy, which generate, via new gene dosage combinations, a plethora of cancer phenotypes, including drug resistance. To distinguish between these mechanisms, we have asked whether three mouse cell lines can become drug resistant, from which two or three genes have been deleted, and on which multidrug resistance is thought to depend: Mdr1a, Mdr1b, and Mrp1. Because all three lines could acquire multidrug resistance and were aneuploid, whereas diploid mouse cells could not, we conclude that aneuploid cells become drug resistant via specific chromosome assortments, independent of putative resistance genes. We have asked further whether aneuploid drug-resistant Chinese hamster cells revert spontaneously to drug sensitivity in the absence of cytotoxic drugs at the high rates that are typical of chromosome reassortments catalyzed by aneuploidy or at the very low or zero rates (i.e., deletion) of gene mutation. We found that four drug-resistant hamster cell lines reverted to drug sensitivity at rates of about 2–3% per generation, whereas two closely related lines remained resistant under our conditions. Thus, the karyotypic instability generated by aneuploidy emerges as the common source of the various levels of drug resistance of cancer cells: minor spontaneous resistances reflect accidental chromosome assortments, the high selected resistances reflect complex specific assortments, and multidrug resistance reflects new combinations of unselected genes located on the same chromosomes as selected genes.

Soon after the introduction of chemotherapy over 50 years ago (1–3), it was discovered that cancer cells and aneuploid cell lines, but not normal diploid cells, could become resistant to one, or even multiple, cytotoxic drugs at concentrations that exceed normally lethal doses by over 3,000-fold (1, 4–12). Subsequently, spontaneous drug resistance exceeding normally lethal doses up to 20-fold was found in some cancer cells (14–16). Besides this abnormal ability to become drug resistant, cancer cells and cell lines also share immortality, i.e., indefinite growth in vivo and in vitro, tumorigenicity, and aneuploidy (17–20).

In view of the clinical relevance of the resistance of cancer to chemotherapy, the genetic basis of drug resistance has been studied extensively both in cancers in vivo and in cell lines in vitro (2, 9, 12, 21, 22). However, despite these efforts, the ability of cancer cells to become drug resistant in terms of conventional gene mutation is still paradoxical on several counts: (i) acquisition of drug resistance by cancer cells occurs at rates of 10∧-3 to 10∧-6 per mitosis (1, 2, 5), whereas mutation of both alleles of drug-metabolizing enzymes should occur at rates of 10∧-12 to 10∧-14, because the spontaneous mutation rates of mammalian genes are 10∧-6 to 10∧-7 (5, 18, 23, 24); (ii) the rates at which cancer cells acquire drug resistance are independent of the chromosomal ploidy of the respective cell, whereas conventional gene mutation depends exponentially on the natural or artificially altered ploidy of the cell (7, 8, 25); (iii) selection for resistance against one cytotoxic drug also generates resistance to entirely unrelated drugs, a phenomenon that led to the suggestion of multidrug resistance genes (MRGs) (11, 12, 22).

Several hypotheses have been advanced to resolve these paradoxes. The most studied of these hypotheses holds that the drug and multidrug resistance of cancer cells is due to either mutational inactivation of drug-metabolizing genes (9, 18) or mutational activation of multidrug transport or resistance genes that simultaneously protect against multiple unrelated cytotoxic drugs (12, 16, 26, 27). However, the gene mutation–drug resistance hypothesis is severely constrained, as the following unanswered questions demonstrate: (i) What kind of mutation would exclusively protect cancer cells against chemotherapy from which normal cells are dying (3, 15)? (ii) Why would drug resistance genes mutate only at exceptionally high rates in cancer cells? (iii) Why would resistance via mutation of a drug-metabolizing gene not depend exponentially on the ploidy of the respective chromosome? (iv) Why would resistance against the same drug not correlate with the expression of the same MRGs in different cancers (14, 16, 22, 26, 28–35)? Indeed, a burgeoning “family” (22) or “superfamily” (14) of MRGs has now been discovered as a result of ever-new noncorrelations between drug resistance of different cancers and previously known putative drug resistance genes (14, 16, 21, 22, 26, 29, 31). (v) Why would the same expression of the same drug resistance gene confer very different degrees of protection to different cells (14)? (vi) How can selection with nonmutagenic drugs, such as puromycin and methotrexate, generate in vivo and in vitro within a few cell generations a spectrum of cell types differing over 3,000-fold in their degrees of resistance against the same drug? (vii) Why is there no, or only moderate, (>1- to 20-fold) protection to aneuploid cancer cell lines expressing spontaneous or transfected MRGs in vitro, unless these cells are selected for survival in the presence of cytotoxic drugs (13, 16, 22, 28, 33, 34, 82)? (viii) Why would spontaneous MRG expression confer only little, or even no, protection to cancers, except to those that appear after chemotherapy (15, 21, 33, 35)? (ix) Why are white blood cells of MRG-transgenic mice protected only against 2-fold normal lethal doses of cytotoxic drugs (36)? And why are mice with MRG deletions only 2-fold more sensitive than undeleted progenitors (82)? (x) How can one MRG product protect against multiple entirely unrelated drugs?

Because the drug resistance–gene mutation hypothesis fails to answer these questions, we have recently proposed an alternative mechanism that can answer all of them. According to this mechanism, drug resistance is generated de novo in cancer cells by chromosome reassortments that are catalyzed by cancer-specific aneuploidy (37). This mechanism predicts that a drug-resistant phenotype can evolve stepwise from a simple accidental chromosome combination, which provides low drug resistance to a complex specific chromosome combination that confers high drug resistance. Conversely, the hypothesis predicts reversion of the drug-resistant phenotype by the same mechanism. The hypothesis that drug resistance of aneuploid cells is achieved by selection of specific assortments of chromosomes also predicts multidrug resistance, because any chromosome combination that is specific for a selected function is also specific for many unselected functions encoded by syntenic genes of the reassorted chromosomes. Thus, cells selected for resistance against one specific drug can also be resistant against unselected drugs and can have variant cellular morphologies (37). The hypothesis further predicts that normal diploid cells cannot become drug resistant by this mechanism, because aneuploidy is not compatible with normal function, development, and germinal inheritance (38–40). Our hypothesis is based on the following sets of data:

(i) Aneuploidy is the most common genomic abnormality of cancer cells (18, 41, 42).

(ii) Aneuploidy catalyzes chromosome reassortments, because it unbalances the gene networks that encode the dosage-sensitive components of the spindle apparatus, including even the number of centrosomes (43). As a result, the risk of a given chromosome of a highly aneuploid cell being lost or doubled per mitosis is about 2% (44, 45). In contrast, the risk of chromosomes of normal diploid cells being lost or doubled during mitosis is very low, ranging from 0% in human embryos (46) and adolescents (47) to 0.4% in adults, on the basis of finding trisomic chromosomes in 0.2% (48). Thus, the characteristic instability and heterogeneity of cancer-specific karyotypes and resulting phenotypes even among the cells of the same cancer (41, 49–51) are a direct consequence of aneuploidy-catalyzed chromosome reassortments.

(iii) Different assortments of normal chromosomes generate different normal and, above all, abnormal phenotypes by altering simultaneously the dosages of thousands of their constituent genes. This mechanism is analogous to a car factory changing its output by altering the number and balance of assembly lines. For example, various increases of normal cellular performance are achieved in the cells of the human liver, heart, and megakaryocytes by polyploidies in which the normal balance of chromosomes is maintained (52). By contrast, changing the normal balance of chromosomes just a little (resulting in aneuploidy) generates abnormal phenotypes such as Down's and Klinefelter's syndromes (3), and changing it above a critical threshold generates the many unique phenotypes of cancer (19, 45, 53, 54). Cancer-specific phenotypes generated de novo by such abnormal chromosome combinations include metastasis, immortality, dedifferentiation, cancer-specific DNA indices (19), abnormal nuclear and cellular morphologies (37, 55), antigenic variation (56), the ability of human cancer cells to grow even in animal hosts (56, 57), resistance to polio and other human viruses (55, 58), and probably resistance to cytotoxic drugs (37). Because the expression of some of these functions is controlled by chromosomal constellations that are not necessary to maintain cancer, for example drug resistance, the same kind of cancers may differ widely with regard to such “incidental” functions (49).

To distinguish between the chromosome reassortment and gene mutation–drug resistance hypotheses, we have tested here two critical predictions of the chromosome reassortment hypothesis: (i) aneuploid cells, even without drug resistance genes, should be able to acquire drug resistance; (ii) drug-resistant cells generated by chromosomal reassortments should be able to revert to drug-sensitive cells in the absence of selective drugs at the same high rates with which they are generated. By contrast, drug-resistant cells generated by gene mutation should be as stable as conventional gene mutation of diploid cells or be irreversible (!) if drug resistance depends on genes that have been deleted. In what follows, we demonstrate that both these critical predictions of the chromosome reassortment–drug resistance hypothesis are confirmed.

Materials and Methods

Cells.

The origin and physiology of mouse cell lines with deletions of two genes thought to confer multidrug resistance, denoted 771 and 886, and of another with deletions of three such genes, denoted 4B, have been described (16). Mouse primary embryo cells (one or a few generations in cell culture) were obtained from L. Evans (National Institutes of Health) and G. Yerganian (Brandeis University, Boston, MA). Both mouse embryonic cells and cell lines were propagated in DMEM supplemented with 10% FCS and antibiotics, following published procedures (59).

Cytogenetic Analysis.

Metaphase chromosomes were prepared and analyzed as described previously (37, 54).

Treatments of Cultured Cells with Cytotoxic Drugs.

Cells at either 2–4 million per 10-cm Falcon culture dish (Becton Dickinson) in 7- to 10-ml medium, or at 0.5–1.5 million per 6-cm dish in 3- to 4-ml medium were treated after they had attached to the culture dish and typically after incubation over night. Puromycin (Sigma), cytosine arabinoside (araC) (Mack-Pfizer, Illertissen, Germany), colcemid (GIBCO/BRL), and methotrexate (Lederle, Sigma) were applied as described previously (37). Actinomycin D (Sigma) was dissolved in DMSO and added to the surface of the culture medium after the cells had attached to the culture dish. As in a previous publication, the concentration of drugs is reported in micrograms per Petri dish (37).

Results

Selection of Drug-Resistant Variants from Mouse Cell Lines Without Multidrug Resistance Genes.

To test whether mammalian cells depend on MRGs for drug resistance, we examined whether mouse cell lines can become drug resistant from which both alleles of two or three of those genes have been deleted on which multidrug resistance is thought to depend: Mdr1a, Mdr1b, and Mrp1. Two of these lines, denoted 771 and 886, are Mdr1a^−/− and Mdr1b^−/−, and the third, denoted 4B, is Mdr1a^−/−, Mdr1b^−/−, and Mrp1^−/− (Materials and Methods) (16). As a control for the effects of these MRGs on drug resistance, two cell lines with intact MRGs, WT-2ac1 and WT-12, were studied in parallel. These lines had been derived from the same mouse strains from which the MRG deletion mutants were originally derived (Materials and Methods) (16).

Because the chromosome reassortment–drug resistance hypothesis postulates that the origin of drug-resistant chromosome combinations is catalyzed by aneuploidy, we first determined whether the mouse cell lines were aneuploid by analyzing the chromosome numbers of metaphases (Materials and Methods). As shown in Table 1 and, as expected from the literature (18–20, 60, 61), each of the five cell lines is indeed aneuploid. Cell line 771 has a modal chromosome number of 70 and a distribution that ranged from 68 to 144 chromosomes per cell. Line 886 has a modal chromosome number of 63 and a distribution that ranged from 59 to 128 chromosomes per cell. Line 4B has a modal chromosome number of 68–69 and a distribution that ranged from 65 to 73 chromosomes per cell. The WT-2ac1 line has a modal chromosome number of 82 and a chromosome distribution that ranged from 78 to 123, and the WT-12 line has a modal chromosome number of 113 and a chromosome distribution that ranges from 106 to 113. Because normal diploid mouse cells have 40 chromosomes, it follows that each of these cell lines is aneuploid.

Table 1.

Chromosome numbers of individual cells of mouse cell lines without and with multidrug resistance genes

Cell line	mn*	n†	Chromosomes per cell
771, Mdr1a,b−/−	70	15	68, 68, 69, 70, 70, 70, 71, 71, 72, 74, 74, 75, 76, 76 + double minute chromosomes, 144
886, Mdr1a,b−/−	63	15	59, 61, 62, 63, 63, 63, 63, 63, 64, 65, 67, 72, 72, 115, 128
4B, Mdr1a,b−/− and Mrp1−/−	68–69	15	65, 67, 67, 68, 68, 68, 69, 69, 69, 70, 71, 71, 71, 73, 73
WT-2ac1	82	15	78, 78, 78, 80, 82, 82, 82, 82, 82, 82, 82, 83, 83, 84, 123
WT-12	113	10	106, 106, 108, 109, 110, 110, 112, 113, 113, 113

^*mn is the modal chromosome number. 

^† n is the number of metaphases analyzed. 

Selection of drug-resistant variants of the mouse cell lines was started with puromycin, a nonfunctional analog of alanine tRNA, as described previously (37). Following the “classic stepwise selection protocol” (12), selection of drug-resistant cells was initiated at low drug concentration. Over 50% of all cells of each mouse line survived an initial selection at 1 μg of puromycin per 10-cm culture dish. Under these conditions, all treated cultures remained confluent monolayers, although many more dead cells were floating in the culture medium than in untreated controls. After 2 weeks, the 1-μg puromycin-resistant cells were subjected to further rounds of selection with stepwise increases of puromycin concentrations up to 20 μg per 10-cm culture dish (Table 2). Resistance was recorded between 9 and 21 days after the initiation of treatment at a given drug concentration, either as single microcolonies per dish or as percentages of cells growing in the presence of drugs compared with those growing without drugs. It can be seen from Table 2 that eventually each of the three cell lines, 771, 886, and 4B, was able to generate drug-resistant sublines that were viable in the presence of 20 μg of puromycin per 10-cm culture dish. Each of the two control cell lines with intact MRGs, WT-2ac1, and WT-12 also generated puromycin-resistant sublines (Table 2).

Table 2.

Generation of puromycin-resistant variants from mouse cell lines without drug-resistance genes

Cells	Percent resistance to puromycin*
	2.5 μg of puromycin second selection	5 μg of puromycin second selection	20 μg of puromycin after multiple selections
771, Mdr1a,b−/−	50†	<2 (few colonies)	>90
886, Mdr1a,b−/−	50	<2 (few colonies)	>90
4B, Mdr1a,b−/−, Mrp1−/−	<10 (colonies)	0	>90
WT-2ac1, wild type	>90	>90	>90 (3 selections)
WT-12, wild type	>90	75	>90 (3 selections)

^*2–4 million cells were seeded per 10-cm dish and treated the next day. 

^† Resistances, reported as percent confluence of a drug-treated culture by the time an untreated control had become confluent, were recorded 9–16 days after initiation of drug treatments. Low percentages of resistance manifest as isolated colonies. 

Compared with the two cell lines with intact MRGs, the MRG deletion mutants were more sensitive to puromycin at low concentrations. Therefore, resistant variants of the deletion mutants were obtained only after several more rounds of selections with stepwise increases of puromycin than were necessary to obtain resistant variants of the cells with MRGs (Table 2). Indeed, among the MRG deletions, the triple deletion, 4B, was more sensitive to early selection steps at 2.5 and 5 μg of puromycin per 10-cm culture dish than the two double deletions, 771 and 886 (Table 2). It follows that the MRGs control the sensitivity of unselected cells to puromycin but do not control the ability of these cells to generate drug-resistant sublines.

No Drug-Resistant Diploid Mouse Cells.

In contrast to the five aneuploid mouse cell lines, diploid mouse embryo cells (Materials and Methods) did not generate puromycin-resistant colonies under our conditions. Instead, the diploid mouse cells slowly deteriorated over 4 weeks at the same low concentrations of puromycin (up to 2 μg per 10-cm dish) that generated resistant colonies from each of the five aneuploid lines. This result confirms previous failures to isolate drug-resistant variants from diploid Chinese hamster embryo cells (37).

Multidrug Resistance.

Because the chromosome reassortment–drug resistance hypothesis predicts multidrug resistance for cells selected to resist one specific drug, each of the three puromycin-resistant MRG deletion lines was tested for resistance against araC, colcemid, and Actinomycin D, as described previously (11, 37).

For the selection of multidrug resistance, confluent cultures of the 771, 886, and 4B cell lines, which had been rendered resistant to 20 μg of puromycin per 10-cm dish (Table 2), were split five ways and were incubated in 10-cm dishes either with puromycin alone (20 μg), with the same amount of puromycin plus araC (3 μg), with puromycin plus colcemid (0.1 μg), or with puromycin plus Actinomycin D (0.3 μg) (left, Table 3). Resistance against these drug combinations was recorded as the relative confluency a multidrug-treated culture had reached by the time a puromycin-only-treated culture was 100% confluent (left, Table 3). In parallel, unselected samples of each of the three cell lines were also treated with each of these drugs alone to determine their backgrounds of spontaneous drug or multidrug resistance (right, Table 3). Note that the initial puromycin resistances of the three cell lines in right, Table 3 are the first selections and are thus lower than those of the second selections reported in Table 2. Because it was observed here and previously (37) that puromycin and, to a lesser degree, Actinomycin D are fast-acting toxins, killing most susceptible cells within 2–4 days after initiation of treatment, and that araC and colcemid are slower, killing most susceptible cells only 1–3 weeks after initiation of treatment, the toxicity of these slow drugs was recorded only 2–3 weeks after initiation of treatment (Table 3).

Table 3.

Multidrug resistance of puromycin-resistant and unselected mouse cell lines without drug-resistance genes

Relative resistance of puromycin-resistant cells in percent*					Resistance of unselected cells in percent
Cell	Puromycin (20 μg)	Puromycin (20 μg) + araC (3 μg)	Puromycin (20 μg) + colcemid (0.1 μg)	Puromycin (20 μg) + Actinomycin (0.3 μg)	Cell	Puromycin (2.5 μg, 5 μg)	AraC (0.6 μg, 3 μg)	Colcemid (0.1 μg)	Actinomycin D (0.15 μg)
771-P20	100	50	70	<1†	771	7.5, 0	20, <1	20	<1
886-P20	100	65	50	<1†	886	2, 0	75, <1	50	<1
4B-P20	100	90	60	<1†	4b	3, 0	90, <1	10	0

^*Final resistances against araC and colcemid were recorded 2–3 weeks after initiation of treatment (see text) and otherwise as in Table 2. 

^† A few colonies were obtained if treatment was interrupted after 2–4 days. 

It can be seen in Table 3 that the puromycin-resistant cell lines were 50- to 90-fold more resistant against araC and 1- to 6-fold more resistant against colcemid than their unselected progenitors. However, no enhanced resistance against Actionomycin D was detected under our conditions (Table 3). We conclude that puromycin resistance of the three MRG deletion mutants also confers multidrug resistance, as for example against such puromycin-unrelated drugs as araC and colcemid.

Because it may be argued that MRG-deletion mutants cannot become resistant to Actinomycin D due to a lack of resistance genes, we have subjected each of the three puromycin-resistant cell lines without MRGs to selection for Actinomycin D resistance, following the stepwise protocol described above and previously (37). Starting with colonies that had survived 0.3 μg of Actinomycin D per 10-cm dish for 2–4 days (left, Table 3), resistant sublines of 771 and 886 have since been obtained that can grow in 0.15 μg of Actinomycin D and a subline of 4B that can grow in 0.06 μg of Actinomycin D. Thus mouse cells without MRGs can generate Actinomycin D resistance.

Spontaneous Reversion of Drug Resistance of Aneuploid Chinese Hamster Cell Lines.

In a further effort to distinguish between chromosome reassortments and gene mutation as the mechanisms of drug resistance, we have analyzed the spontaneous reversion rates of the drug-resistant phenotypes of several aneuploid Chinese hamster cell lines described recently (37). On the basis of the low spontaneous mutation rates of 10∧-6 to 10∧-7 per gene per mitosis and the inability of spontaneous deletions to revert (see above), the gene mutation hypothesis predicts very stable, even irreversible, phenotypes. By contrast, phenotypic reversion via chromosome reassortments would occur at high rates, considering that the risk of a given chromosome of highly aneuploid cells to be lost or doubled per mitosis is about 2% (44, 45).

In view of these considerations, we have propagated six drug- and multidrug-resistant Chinese hamster cell lines for about 30 generations, i.e., five passages at 50- to 100-fold dilutions, in the absence of selective drugs. Drug-sensitive precursors of these cell lines have been originally derived from embryo cells transformed in vitro with benzpyrene and dimethylbenzanthracene and designated B 644, D 313, and D 3 (37, 45). Sublines resistant to either colcemid, araC plus colcemid, araC plus puromycin plus colcemid, methotrexate, or colcemid plus puromycin were then prepared as described recently (Table 4) (37).

Table 4.

Spontaneous reversion of drug resistance of aneuploid Chinese hamster cell lines propagated for 30 generations without selective drugs

Cell line	Number of colonies* without drugs	Number of colonies* with drugs	Relative resistance, percent
B 644-Col02‡	15	4	27
B 644-A5 + Col01‡	107	1	<1
B 644-A5 + P5 + Col01‡	400	34	12
D 3-MTX5‡	100†	40†	40
D 313-P5 + Col01‡	40	42	>100
D 313-P5 + Col02‡	204	184	90

^*Aliquots of 5,000 and 50,000 cells were plated per 10-cm Petri dish and an area of 560 mm² was counted. In the presence of drugs, colonies were typically smaller and included more granulated cells, than in the absence. 

^† Percent confluence. 

^‡ Col02 is resistant to 0.2 μg and Col01 to 0.1 μg colcemid per 10-cm dish, A5 is resistant to 5 μg araC, P5 is resistant to 5 μg puromycin, and MTX5 is resistant to 5 μg methotrexate. 

After about 30 generations in the absence of drugs, the percentage of drug-resistant cells of each line was determined. For this purpose, the numbers of colonies formed by the same number of input cells grown in the presence and absence of the respective selective drugs were compared (Table 4). Alternatively, the percentage of confluency of a culture in the presence of drugs was determined at the time when the drug-free culture had reached 100% confluency (Table 4).

It can be seen from Table 4 that four of the six cell lines tested had lost between 60 and 99% of resistant cells after 30 unselected generations. Moreover, partial loss of drug resistance was also apparent in the drug-resistant fraction of these lines, because the colonies growing in the presence of drugs were smaller and included cells with morphological defects, such as granular inclusions. The 60–99% reversion rates of these four cell lines during 30 unselected generations correspond to an approximate reversion rate of 2–3% per generation. This rate is directly compatible with the known risk of a chromosome of a highly aneuploid cell to be lost or doubled per mitosis (see above) (44, 45) and thus supports the chromosome reassortment–drug resistance hypothesis.

Nevertheless, two closely related multidrug-resistant cell lines, D 313-P5 + Col01 and its derivative D 313-P5 + Col02, did not significantly revert to drug sensitivity during 30 unselected generations under our conditions (Table 4). However, all things considered, even this result is compatible with the chromosome reassortment–drug resistance hypothesis, because (i) the two lines have originated from drug-sensitive precursors at the same high rates as the other drug-resistant lines studied here and previously (37), and, importantly, (ii) selected phenotypes generated by specific chromosome combinations may include other unselected phenotypes that confer a selective advantage for growth under our conditions. Indeed, some aneuploid combinations must generate genomic cul-de-sacs from which it is not possible to return to a given phenotype, for example those from which the chromosome with a wild-type allele from an originally heterozygous precursor has been deleted. Further work, e.g., determining whether the karyotype remains unchanged so long as the phenotype does not change, and perhaps growing these cells under different conditions, is necessary to answer this question.

We conclude that the high spontaneous reversion rates of drug resistance of four of six lines tested directly support the chromosome reassortment–drug resistance hypothesis, and that the nonreversion or minimal reversion of two closely related lines is also more parsimoniously explained by our hypothesis. These conclusions confirm and extend those of others who have observed high reversion rates of drug resistance in most cancer cells (9, 10, 12, 62–65) and also of those who have observed slow or no reversions in some cancer cells (9, 12).

Discussion

De Novo Generation of Drug Resistance by Chromosome Reassortments.

Our results demonstrate that generation of drug- and multidrug-resistant variants from aneuploid cancer cells and cell lines (i) does not depend on specific previously described drug resistance genes, and (ii) is spontaneously reversible at rates that are orders of magnitude in excess of those compatible with gene mutation. These results support the hypothesis that chromosome reassortments catalyzed by aneuploidy are sufficient to generate de novo high levels of drug resistance, independent of gene mutation.

Recent literature provides an ideally controlled example in support of the chromosome reassortment hypothesis. In an attempt to control the diploid and chronically hyperplastic phases of chronic myeloid leukemia as well as the aneuploid and malignant phase, or blast crisis (19), the same cytotoxic drug, STI-571, was developed (66, 67). According to a News and Views article in Nature, the “new-age drug” was “rationally designed” to inhibit the putative common cause of the chronic and malignant phases, a tyrosine kinase encoded by the BCR-ABL hybrid gene that “drives the cells … to become cancerous” (66). But, contrary to expectation, only patients suffering from the diploid chronic phase showed lasting responses, whereas “most” patients suffering from the aneuploid blast crisis “relapsed within a few months, despite continued treatment” (66). In searching for an answer, some investigators have quickly offered several mutations of the BCR-ABL gene that would render the encoded kinase resistant against STI-571 without affecting its putative oncogenic kinase function (67). But, as the News and Views article points out, this answer generates at least one new question: “Why do drug-resistant cells emerge from blast crisis and not from the earlier phase?” In addition, one wonders why mutations of the active site of the kinase, which prevent the competitive inhibitor from binding, would not also prevent the kinase from maintaining transforming function. Our hypothesis suggests simple answers to both questions. The blast crisis is caused by aneuploidy rather that by the kinase (19), and aneuploidy also generates drug-resistant variants by chromosome reassortments. Indeed, drug-resistant blast cells without mutations of the active site have already been observed (67). By contrast, the diploid hyperplastic leukemia cells cannot generate drug-resistant variants by chromosome reassortments.

Chromosome Reassortments vs. Gene Mutation as Causes of Drug Resistance.

Here we try to discriminate between the two competing drug resistance hypotheses on the basis of their abilities to explain: (i) High resistance levels that allow variants to grow at concentrations of cytotoxic drugs that can exceed normally lethal levels over 3,000-fold (1, 4, 11, 68, 69). Variants with such high levels of resistance typically arise only after several cell generations from cancers and aneuploid cell lines that have been subjected to cytotoxic drugs, regardless of whether the drugs are mutagenic or nonmutagenic (1–3, 18, 69–71). (ii) Low spontaneous levels of drug resistance that protect cells from untreated cancers and unselected cell lines against cytotoxic drugs at concentrations that are up to 20-fold higher than are normally lethal (13, 16, 22, 34). (iii) The high reversion rates of drug resistance in some cells. (iv) The noncorrelations between overexpression of any one putative drug resistance gene and drug resistance in any kind of cancer (see above). (v) The failure of diploid cells to acquire drug resistance at high rates. In the following, we argue that only the chromosome reassortment hypothesis provides a coherent explanation.

According to the chromosome reassortment–drug resistance hypothesis, the relatively small, i.e., >1- to 20-fold, differences in drug resistance that are observed among previously untreated cancers and unselected cell lines reflect accidental chromosome assortments that generate de novo various drug-resistant phenotypes. By contrast, the high resistances obtained after chemotherapy of cancer or selection of cell lines in vitro with either mutagenic or nonmutagenic drugs reflect complex specific chromosome assortments producing high dosages of drug-resisting genes. This mechanism also directly explains the high reversion rates of drug resistance described here by us and previously by others (9, 10, 12, 62–65). Moreover, this mechanism predicts that the far-ranging levels of homologous drug resistance acquired by aneuploid cells are generated by gene amplifications, alias aneuploidy, and indirectly by product modification involving enzymes that are themselves up- or down-regulated by aneuploidy, rather than by a bewildering range of gene mutations with identical specificities but activities that differ over 3,000-fold. Indeed, all studies that address the “activation” of drug resistance genes in cancer cells and cell lines cite “gene amplification” and product phosphorylation rather than mutation (15, 26, 27).

According to our hypothesis, the noncorrelations between drug resistance and expression of putative MRGs reflect chromosome reassortments that generate either drug resistance or sensitivity without affecting the expression of genes currently known as MRGs. In addition, some noncorrelations may reflect misidentifications of the functions of genes thought to be MRGs. This is particularly true for investigations of MRG expression that do not measure MRG function, as for example a recent study on breast cancer (81).

Indeed, the relevance of known drug resistance genes to cancer chemotherapy is now called into question as a result of the many noncorrelations between the expression of drug resistance genes and drug resistance of cancers, even by leading proponents of the drug resistance gene hypothesis (14, 15, 21). For example, Borst et al. recently wrote, “P-gp (the multidrug resistance protein first identified) was discovered in 1976. Today, 24 years later, there is still no consensus on its contribution to drug resistance in cancer patients” (22). By contrast, the relevance to cancer chemotherapy of the enormous levels of drug resistance conferred by specific chromosome assortments that are acquired during multiple steps of therapeutic or experimental drug selection has been demonstrated experimentally as early as 1952 (1) and has been the nemesis of chemotherapy since it is used to treat cancer (2, 3, 15).

Finally, the chromosome reassortment hypothesis explains why only aneuploid, but not normal diploid, cells can become drug resistant.

Drug Resistance-Specific Karyotypes?

The chromosome reassortment–drug resistance hypothesis predicts that there should be resistance-specific karyotypes. However, it also predicts three reasons why such karyotypes will be difficult to identify: (i) The karyotypes of aneuploid cells are moving targets that change autocatalytically at rates that are directly proportional to the degree of aneuploidy, i.e., 2% per generation in highly aneuploid cells (see above) (45); (ii) due to the relative large numbers of chromosomes of animal cells and the randomness of aneuploidy-catalyzed chromosome reassortments, a drug-specific chromosome constellation will be masked by a great variety of nonspecific collateral variations; and (iii) because there are often several alternative pathways to generate the same phenotype, e.g., resistance to a specific drug (6, 9, 72) or loss of differentiated function, there are probably also several different phenotype-specific karyotypes. In agreement with both of these predictions, all cytogenetic studies of drug resistance have observed nonparental, but as yet no drug-specific, karyotypes (10, 55, 68–70, 73–76).

Likewise, all cytogenetic studies of solid cancer have found aneuploidy but as yet no cancer-specific karyotype (19). The failure to find a “cancer-specific karyotype” is indeed the primary reason why aneuploidy is not accepted as the cause of cancer (19, 41, 77–79). By contrast, we have argued that aneuploidy above a certain threshold is sufficient to cause cancer, and that cancer-specific karyotypes are inevitably masked by collateral nonspecific chromosome assortments due to the inherent instability of aneuploid karyotypes (19, 53, 54).

In the meantime, a drug resistance-specific karyotype should be easier to find than the elusive cancer-specific karyotype, because drug resistance is a much less complex biochemical target than the many additive phenotypes that define cancer (19, 49). In view of these considerations further work is now planned to find a drug-specific karyotype—which, according to an early observer, “would be the most direct evidence of a causal relationship between karyotype and phenotype” (68).

Relevance of Drug Resistance from Chromosome Reassortments to Testing the Toxicity of Chemotherapeutic Drugs in Cell Culture.

Testing the toxicity of chemotherapeutic drugs in human or animal cells in culture typically does not consider the role of aneuploidy in drug resistance. For example, we noted in 1995 that 9 different studies analyzing the toxicity of azidothymidine, which is currently widely used as an antiviral drug in AIDS patients, reported lethal doses ranging from 1 to >1,000 micromolar, depending on the cell types used (80). The lethal doses in presumably diploid primary cells were all low, ranging from 1 to 5 micromolar in 6 cell types and 25 micromolar in one, whereas the lethal doses determined in cell lines ranged from 4 to >1,000 micromolar. Moreover, we observed that initially sensitive cell lines developed drug resistance over a period of 2 months (80). Although both of these results appeared paradoxical in 1995, they can now be explained exactly by the chromosome reassortment–drug resistance hypothesis. In view of these data, we propose that the relevance of toxicity tests of chemotherapeutic drugs in vitro for clinical use could be much improved, if they were based on diploid human or animal cells rather than on aneuploid cell lines.

Abbreviations

MRG

multidrug resistance gene

araC

cytosine arabinoside