Regulatable killing of eukaryotic cells by the prokaryotic proteins Kid and Kis

Guillermo de la Cueva‐Méndez, Anthony D. Mills, Lorena Clay‐Farrace, Ramón Díaz‐Orejas, Ronald A. Laskey

Author Affiliations

  1. Guillermo de la Cueva‐Méndez*,1,
  2. Anthony D. Mills1,
  3. Lorena Clay‐Farrace1,
  4. Ramón Díaz‐Orejas2 and
  5. Ronald A. Laskey1
  1. 1 MRC Cancer Cell Unit, Hutchison/MRC Research Centre, Hills Road, Cambridge CB2 2XZ, Wellcome Trust Cancer Research UK Institute, Tennis Court Road, Cambridge CB2 1QR and Department of Zoology, University of Cambridge, Downing Street, Cambridge, CB2 3EJ, UK
  2. 2 Centro de Investigaciones Biológicas (C.S.I.C.), Department of Molecular Microbiology, Velázquez 144, E‐28006, Madrid, Spain
  1. *Corresponding author. E-mail: gd247{at}
View Full Text


Plasmid R1 inhibits growth of bacteria by synthesizing an inhibitor of cell proliferation, Kid, and a neutralizing antidote, Kis, which binds tightly to the toxin. Here we report that this toxin and antidote, which have evolved to function in bacteria, also function efficiently in a wide range of eukaryotes. Kid inhibits cell proliferation in yeast, Xenopus laevis and human cells, whilst Kis protects. Moreover, we show that Kid triggers apoptosis in human cells. These effects can be regulated in vivo by modulating the relative amounts of antidote and toxin using inducible eukaryotic promoters for independent transcriptional control of their genes. These findings allow highly regulatable, selective killing of eukaryotic cells, and could be applied to eliminate cancer cells or specific cell lineages in development.


Prokaryotic plasmids have developed genetic systems that increase their stable maintenance in bacterial hosts. One group of them, called killer systems, eliminate bacteria that have lost the plasmid during cell division (Jensen and Gerdes, 1995). ParD is a protein killer system of Gram‐negative plasmid R1, which is composed of two genes: kis (for killing supressor) and kid (for killing determinant) that encode the antidote (Kis; 10 kDa) and the toxin (Kid; 12 kDa), respectively. Both genes are organized in a bicistronic operon in which the toxin is located downstream of the antidote, and their expression is controlled in several ways: by coupled transcription; by post‐transcriptional processing of some of its bicistronic mRNAs to produce kis+/kid− and kis+/kid+ messengers; by overlapping translation of both genes; and by a very tight interaction between Kis and Kid to generate a non‐toxic complex that represses its own transcription (Bravo et al., 1987, 1988; Ruiz‐Echevarría et al., 1991, 1995). These controls avoid the synthesis of the toxic component if its antidote has not been translated previously and ensure a balanced production of the antidote relative to the toxin (Figure 1A). Under normal circumstances, both components of this killer system are synthesized at a basal level from their plasmid, allowing the bacterial host to survive. However, the stability of the antidote Kis is lower than that of the toxin Kid (Tsuchimoto et al., 1992). In bacteria that lose the plasmid and thus lack continuous synthesis of the unstable antidote, its more rapid degradation leads to an excess of non‐neutralized Kid protein, which is toxic to the host and inhibits its proliferation (Figure 1B).

Figure 1.

Mechanism of action of protein prokaryotic killer systems and genetic organization of the parD killer system of plasmid R1. (A) The parD killer system of R1 is organized in a bicistronic operon that ensures that transcription of the antidote kis occurs before that of the toxin kid. After transcription, some bicistronic mRNAs are degraded by exonucleolytic processing from their 3′ end up to an inverted repeat that folds in a stem–loop structure (white arrowheads). This ensures an excess of monocistronic mRNAs that encode only for the antidote. Translation of kid starts from a Shine–Dalgarno sequence (indicated by asterisks) that overlaps with the 3′ end of the kis gene. Thus, kid translation can start only if the antidote has been translated previously. All these strategies ensure that, under normal circumstances, the antidote is always in excess over the toxin and they allow the formation of a non‐toxic protein complex that interacts with an inverted repeat located in the parD promoter (black arrows) to repress its own transcription. (B) Toxicity of protein prokaryotic killer systems is neutralized in plasmid‐containing cells by continuous synthesis of an unstable antidote (light grey spheres). Loss of the plasmid during bacterial division prevents further production of the antidote and its preferential degradation allows the toxin to exert its lethal effect on bacterial segregants (dark grey spheres).

Pathogenic bacteria have evolved many toxins to attack eukaryotic cells and many of these have had valuable practical applications (Fitzgerald, 1996; Culver, 1997). However, plasmid stability systems differ from these. They have evolved in bacteria to kill bacteria, but under the tight control of effective antidotes. If a similar system could be developed for eukaryotic cells, it would have many applications. For example, gene therapy approaches for selectively killing cancer cells depend on highly selective targeting or expression of toxins to cause maximum damage to cancer cells whilst minimizing damage to normal cells. A higher degree of selectivity could be achieved if the toxin is targeted to tumour cells and non‐tumour cells are protected from the action of the toxin by a specific antidote. In this work, we show that Kid inhibits cell proliferation in eukaryotes and kills human cells by apoptosis. Furthermore, we demonstrate that the antidote Kis overcomes the toxic effect of Kid in yeast, Xenopus laevis and human cells. We also establish that it is possible to regulate these effects in eukaryotes by means of independent transcriptional regulation of kis and kid, and we discuss strategies for exploiting these results for selectively killing tumour cells and for selective cell‐lineage knock‐outs in development.


Expression of Kid inhibits yeast cell growth and co‐expression of Kis rescues

The first purpose of this work was to test whether the components of the parD system of plasmid R1 could function in eukaryotes. As mentioned above, this killer system is kept silent in Escherichia coli by means of a complex genetic and molecular regulatory circuit, based mainly on the bicistronic nature of the parD operon. Although some bicistronic operons exist in eukaryotes (McBratney et al., 1993; Cornelis et al., 2000), it would be technically difficult to use them effectively for this purpose. Therefore we decided to study the effect of Kis and Kid in Saccharomyces cerevisiae using independent transcriptional control, rather than the native bicistronic nature of the parD operon. Budding yeast was transformed with the integrative plasmid p303MKCKd, in which kis expression is repressed in the presence of methionine and kid expression is activated in the presence of Cu2+ (Figure 2A). Yeast growth was severely inhibited in this transformant in the presence of methionine and Cu2+, but not in their absence or in the presence of Cu2+ only (Figure 2B). These results indicate that expression of Kid inhibits cell proliferation in S.cerevisiae and that co‐expression of its antidote, Kis, protects against inhibition. Importantly, they also indicate that antidote alone has no apparent side effects on yeast cell viability.

Figure 2.

Independent transcriptional control of kis and kid allows activation of the parD system in S.cerevisiae inhibiting cell proliferation conditionally. (A) Scheme depicting plasmid p303MKCKd. (B) Analysis of growth rates of budding yeast transformed with p303MKCKd or with pRS303 (control) in solid media supplemented with the indicated amounts (μM) of methionine and/or Cu2+. Samples were diluted in sterile water at the cell densities indicated and 10 μl of each dilution were used for each dot. Abbreviations: Met25, methionine repressible promoter; CUP1, copper inducible promoter; CYC1, CYC1 terminator.

Microinjected Kid inhibits cell proliferation in frog embryos and kills human cells; Kis protects

Next, we injected purified proteins to see whether Kid inhibits cell proliferation in X.laevis embryos and whether Kis protects from that effect. Two‐cell embryos of X.laevis were microinjected near the animal pole of one of the blastomeres with Kid protein, or an active fusion of the Kis protein (maltose‐binding protein fused to Kis, MBPKis), or both proteins or buffer alone. The effects of these injections on subsequent cell divisions were followed with time (Figure 3A). Kid‐injected blastomeres failed to develop normally unlike the non‐injected half of the embryo. On the other hand, blastomeres injected with MBPKis, MBPKis and Kid, or buffer alone progressed normally in all cases until at least mid‐blastula (Figure 3B). Nuclear staining of sections of the different embryos showed that cells injected with Kid underwent a limited number of divisions (Figure 3C).

Figure 3.

Kid inhibits cell proliferation in X.laevis embryos; Kis neutralizes this effect. (A) Scheme depicting microinjection experiments of Xenopus embryos. One blastomere of a two‐cell embryo was injected with either Kid, MBPKis, Kid plus MBPKis or buffer. Development was allowed to proceed until the control uninjected embryos reached stage 9 (late blastula). (B) Representative embryos from the experiment shown in (A) when buffer, MBPKis, Kid, or Kid plus Kis proteins were microinjected. Almost identical results were obtained with all the embryos injected with Kid (43 in total), MBPKis (15 in total) and Kid plus MBPKis (19 in total). (C) A section of one embryo microinjected into one blastomere with Kid protein. Embryos were fixed, paraffin‐embedded, sectioned and stained for DNA. The uninjected half embryo developed normally whereas the Kid‐injected half showed very few cells, most of which were anucleate. The yellow arrowheads indicate the only two nuclei present in the Kid‐injected half embryo.

These results encouraged us to test the effects of Kis and Kid microinjection in human cells. Thus, we performed similar experiments using HeLa and SW480 cells. Microinjection of Kid into these cells dramatically decreased their survival and eventually led to the death of all Kid‐injected cells. This effect was completely abolished when Kis was pre‐incubated with Kid before injection (Figure 4) and absent when only buffer was injected instead (data not shown).

Figure 4.

In the absence of its antidote Kis, toxin Kid kills human cells. (A) Plot showing that only 9 out of 154 HeLa cells (5.8%) survived 24 h after Kid microinjection, whereas 155 out of 168 HeLa cells (92.5%) did so after Kid plus MBPKis microinjection. (B) Plot showing that only 9 out of 72 SW480 cells (12.5%) survived 48 h after Kid microinjection, whilst 65 out of 65 (100%) of these cells did so when they were microinjected with Kid plus MBPKis.

Kid‐mediated toxicity and Kis‐mediated survival can be regulated in human cells

The results so far demonstrate that the consequences of Kid toxicity are similar in both eukaryotes and prokaryotes. Kid inhibits cell proliferation whilst Kis neutralizes the toxicity of Kid. They also demonstrate that, in yeast, it is possible to substitute the complex prokaryotic regulatory circuits that regulate parD in E.coli by separate control of transcription of the antidote and the toxin. We asked whether independent transcriptional control of kis and kid would regulate cell killing or cell survival in human cells. For that purpose two plasmids, pNATHA1i and pNATHA2i, were constructed. Their design was based on previous observations of relative transcription from a constitutive cytomegalovirus (CMV) early promoter compared with a tetracycline repressible promoter in the presence or the absence of tetracycline (or its analogue doxycycline) in HeLa Tet Off cells (Yin et al., 1996) (Figure 5A).

Figure 5.

Independent transcriptional control of kis and kid allows regulated inhibition of cell proliferation and cell death in HeLa cells. (A) Scheme depicting plasmids pNATHA1i and pNATHA2i, in which kid (if present) is constitutively expressed from a CMV promoter whilst kis expression is repressed in Tet Off cell lines by doxycycline addition to the culture media. (B) Time course comparing the relative growth of HeLa Tet Off (control), kis+ (pNATHA1i+) and kis+/kid+ (pNATHA2i+) stably transfected cells in the presence (Pr kis Off) and the absence (Pr kis On) of doxycycline. Each line represents a comparison of the cell proliferation rate in the cultures without doxycycline (Pr kis On) with the rate for cells grown in the presence of doxycycline (Pr kis Off). Control HeLa cells and pNATHA1i (kis+) cells grow identically with or without kis transcription. In contrast, relative growth decreases over time when kis transcription is repressed by doxycycline in pNATHA2i (kis+/kid+) cells. (C) Cell death rates for the experiment shown in (B). At least 250 cells were assessed at each time point.

HeLa Tet Off cell line was stably transfected with plasmids kis+ (pNATHA1i+) and kis+/kid+ (pNATHA2i+) and clones were selected in the absence of doxycycline so that the promoter for the antidote kis is On (Pr kis On). Figure 5B shows that conditions that switch Kis synthesis Off (Pr kis Off) inhibited proliferation only in cells that contain kid (pNATHA2i+). Furthermore, as expected from the results shown in Figure 4, this inhibition of proliferation was associated with widespread cell death when Kid is expressed without Kis beyond three days (Figure 5C and see below) and with total cell death after 15 days (data not shown). These results demonstrate that, as in the case of S.cerevisiae, inhibition of cell proliferation can be modulated in human cells by independent transcriptional control of kis and kid. Importantly, and as seen previously in the case of yeast and X.laevis, expression of kis alone has no apparent phenotypic effects in HeLa cells.

Kid triggers premature cell death by apoptosis in HeLa cells

Figures 4 and 5C show that excess Kid promotes widespread death in HeLa cells. This observation raised the interesting question of whether the lethal effect of Kid in these cells was due to activation of apoptosis. This pathway, which exists in human cells (Evan and Littlewood, 1998), is absent from S.cerevisiae (Shaham et al., 1998) and the first stages of Xenopus embryonic development (Hensey and Gautier, 1997). To address this question, samples studied in Figure 5 were stained with propidium iodide and fluorescein‐linked Annexin‐V, an early marker for apoptosis, and analysed by confocal microscopy. After 10 days of growth, HeLa Tet Off cell line and kis+ (pNATHA1i+) stable transfectant cells showed a similar small percentage of apoptotic cells in both the absence or presence of doxycycline. In contrast, the percentage of Annexin‐V positive kis+/kid+ (pNATHA2i+) cells increased almost four‐fold upon addition of doxycycline to the growth medium for 10 days (Figure 6A and B). A similar analysis upon longer exposures to doxycycline was not possible as, consistently, all pNATHA2i (kis+/kid+) cells were dead beyond day 10. A magnified view of one of these cells further confirmed apoptosis by morphological criteria (Figure 6C). These results clearly show that widespread cell death observed upon prolonged exposure to Kid in HeLa cells is due to activation of the apoptotic program.

Figure 6.

Expression of kid without kis causes widespread and premature cell death by apoptosis in HeLa cells. (A) Low power fields of control HeLa Tet Off, kis+ (pNATHA1i+) and kis+/kid+ (pNATHA2i+) cell cultures grown in the absence (Pr kis On) and the presence (Pr kis Off) of doxycycline for 10 days and stained with propidium iodide (red) and FITC‐linked Annexin‐V (green). (B) Percentage of Annexin‐V positive cells in the different cultures analysed in (A). (C) Magnified image of one kis+/kid+ (pNATHA2i+) cell grown in the presence of doxycycline (Pr kis Off) for 10 days and stained as in (A).


The parD system of plasmid R1 has evolved to provide a finely balanced mechanism for regulated bacterial cell killing. This system has evolved to ensure the selfish maintenance of the R1 plasmid (Bravo et al., 1987, 1988; Ruiz‐Echevarría et al., 1991, 1995; Tsuchimoto et al., 1992). When the plasmid is present, the antidote exceeds the toxin and the cell survives. If the plasmid is lost during cell division, the antidote decays and the residual toxin kills the plasmid‐free daughter cell; thus, the possession of the plasmid can be likened to holding a hand grenade with the pin pulled; drop it and it kills you. Here we report the surprising finding that the protein components of this prokaryotic system can be used to manipulate cell proliferation or cell survival in a wide range of eukaryotic organisms. Kid inhibits their proliferation and equally important, Kis antagonizes this inhibition. To achieve these effects, it was necessary to place kis and kid under control of eukaryotic promoters. We show in this paper that independent control of both genes using eukaryotic promoters still allows regulatable arrest of cell proliferation and, in some cases, cell death.

Regulated expression of Kid and Kis for selective killing of cancer cells and for targeted cell ablation

Gene therapy approaches for selectively killing cancer cells depend on highly selective targeting or expression of toxins to cause maximum damage to cancer cells, whilst minimizing damage to normal cells. Prokaryotic toxins have been used for this purpose previously (Fitzgerald, 1996), but Kid has the advantage that it can be antagonized by a specific protein antidote in eukaryotic organisms. Unlike simple toxins (Culver, 1997), which require highly specific gene delivery or expression in cancer cells (Zullo et al., 1998), the approach suggested by this work offers the opportunity to both regulate the expression and antagonize the effect of Kid by activating the expression of Kis in non‐tumour cells. Transcriptional regulators known to be inactivated in many human cancers, such as p53 (Hainaut and Hollstein, 2000), could be used to induce kis expression. If so, combining this approach with efficient delivery vehicles might allow a regulated and selective strategy for gene therapy of cancer (Figure 7). Our results show that Kid‐induced cell death does not depend on the presence of functional p53, as SW480 cells, which express a mutant P53 protein, die 48 h after injection of Kid (Figure 4B). The recent determination of the structure of Kid (Hargreaves et al., 2002) opens the possibility of designing small drugs that mimic its effects on human cells.

Figure 7.

Regulated expression of Kid and Kis for selective killing of cancer cells. Scheme depicting a possible approach for anti‐cancer gene therapy based on the transcriptional activation of the kis gene by wild‐type p53 and reinforced by activation of the kid toxin gene by p53 bearing a common mutation. Blue spheres, Kis; red spheres, Kid.

We have shown that Kid inhibits cell proliferation not only in somatic cells, but also in embryonic cells. Thus, an additional use for this approach to regulated inhibition of cell proliferation is the opportunity to perform highly regulated knockouts of cell lineages during development. Targeted ablation has been successfully used in developmental studies (Booth et al., 2000; Lee et al., 2000). Once again, tissue‐specific promoters and conditional promoters could be used to tune the relative expression of kis and kid to allow selected cell knockouts at specific stages of development. This approach could have value in studies of development as well as in studies of differentiation, organogenesis or degenerative disorders.

Materials and methods

Plasmids construction

p303MKCKd was constructed in two steps. First, kis was amplified by PCR and cloned into the SmaI site of p424Met25 (Mumberg et al., 1994). The PvuI fragment of this plasmid containing kis was exchanged by the equivalent fragment of pRS303 (Sikorski and Hieter, 1989) to generate p303MK. Second, kid was subcloned from pCIKid (see below) into pSAL1 (Mascorro‐Gallardo et al., 1996) by exchanging their small ScaI–XhoI fragments. A blunt‐ended BamHI fragment from this plasmid was subcloned in blunt‐ended AatII‐digested p303MK to generate p303MKCKd. For pMALKis (MBPKis overproducer), kis flanked by an EcoRI site at 3′ was obtained by PCR using the oligonucleotides 5′‐ATGCATACCACCCGACTG‐3′ and 5′‐TCGGAATTCAGATTTCCTCCTG‐3′. This PCR product was cloned into XmnI and EcoRI sites of pMAL‐c2 (BioLabs). For pGCHisKisKid (Kid overproducer), kiskid DNA flanked by NdeI and BamHI was obtained by PCR using the oligonucleotides 5′‐GGAATTCCATATGCATACCACCCGACT‐3′ and 5′‐CGGGATCCTCAAGTCAGAATAGTGGACAGG‐3′. This PCR product was cloned into NdeI and BamHI sites of pET15b (Invitrogen) before subcloning its NcoI–BamHI small fragment into the same sites of pRG80‐recA‐Nhis (Giraldo et al., 1998). For pNATHA plasmids, kis was flanked by EcoRI and XbaI by PCR using the oligonucleotides 5′‐CGGAATTCATGCATACTACCACCCGACTG‐3′ and 5′‐CTCTAGATCAGATTTCCTCCTGACC‐3′. This PCR product was cloned into pTRE (Clontech) digested with EcoRI and XbaI. For pNATHA2i plasmid, kid flanked by XhoI and EcoRI sites was amplified by PCR using the oligonucleotides 5′‐CCGCTCGAGATGGAAAGAGGGGAAATCT‐3′ and 5′‐CGGAATTCTCAAGTCAGAATAGTGGACAGG‐3′. This PCR product was cloned in pCI‐neo (Promega) digested with XhoI and EcoRI. Neomycin resistance was deleted from pCI‐neoKid (and from parental pCI‐neo) by BstXI and SmaI digestion and religation. Resultant plasmids lacking the neomycin resistance gene were digested with BglII and BamHI. The CMV promoter‐containing fragments resulting from these digestions were blunt ended and subcloned into a blunt‐ended HindIII pTREKis to obtain pNATHA1i and pNATHA2i plasmids. Tail‐to‐tail orientations between CMV‐ and tetracycline‐dependent transcriptional units were chosen for pNATHA1i and pNATHA2i.

Yeast growth determination

Several fresh colonies of W303a strain (MATa, ade2‐1, trp1‐1, can1‐100, leu2‐3, 112, his3‐11, ura3, psi+) transformed with pRS303 (control) or p303MKCKd were diluted in sterile water to 108 cells/ml. Ten microlitres of this and four serial 1/10 dilutions were placed on SD plates without histidine, supplemented or not with 500 μM methionine and/or 200 μM CuSO4, as indicated in Figure 2, and grown at 23°C for 48 h.

Protein purification

MBPKis was expressed in DH5α and purified by affinity chromatography through an amylose resin (BioLabs) in 20 mM HEPES pH 7.5, 100 mM KCl, 1 mM DTT and 10% ethyleneglycol, following manufacturer's instructions. Kid was expressed in TG1 upon induction of the culture with Nalidixic acid (25 μg/ml; Merck). The soluble fraction was precipitated with 60% ammonium sulphate and resuspended in 20 mM Tris–HCl pH 7.5, 500 mM KCl and dialysed against the same buffer to eliminate the ammonium sulphate. The dialysed fraction was loaded in Fast‐Flow chelating Sepharose column (Pharmacia) activated with Ni2+. A gradient of 0–6 M guanidinium chloride (GnCl) in 20 mM Tris–HCl pH 7.5 was applied to the column until elution of denatured Kid. Kid was diluted in 20 mM HEPES pH 7.5, 6 M GnCl, 150 mM KCl, 20 mM β‐mercaptoethanol, 0.2 mM EDTA and 1.2% CHAPS and refolded by dialysis against 150 mM KCl, 20 mM HEPES pH 8.0, 10 mM β‐mercaptoethanol, 0.1 mM DTT and 10% ethyleneglycol. Soluble protein was concentrated in 3K Centricon tubes.

Protein microinjections

For embryo microinjections, 160 and 720 ng of Kid and MBPKis, respectively, dialysed in 20 mM Tris–HCl pH 8.0 and 50 mM KCl were mixed with each other or with dialysis buffer in 4 μl and incubated on ice for 10 min. Fifty nanolitres of each mix were microinjected into one of the cells of dejellied two‐cell embryos of X.laevis. Embryos were incubated in 4% ficoll 400 in MBS buffer at 18°C until mid‐blastula was reached in the non‐injected controls. Embryos where photographed, fixed and embedded in paraffin wax for sectioning as described previously (Butler et al., 2001). Sections were mounted and stained with Hoechst 33258 or with propidium iodide and analysed by microscopy. For HeLa and SW480 microinjections, proliferating cells were asynchronously grown in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal calf serum and cultured in small chambers. Medium was replaced with warmed CO2‐independent medium (Gibco/BRL) for microinjection. MBPKis and Kid were diluted in water to 2.7 and 2.3 pmol/μl, respectively, and supplemented with 1 mg/ml Texas Red Dextran (70 000 MW; Molecular Probes) and injected into cells. For the Kis and Kid sample, proteins were pre‐mixed at those concentrations prior to injection. After injection, cells were returned to normal DMEM for continuing growth. Images were collected on a Bio‐Rad 1024 confocal microscope for determination of survival rates.

Stable transfections

HeLa Tet Off cell line (Clontech) was co‐transfected with pNATHA plasmids and pTKHyg (Clontech) by the lipofectamine method (Gibco). Stable clones were selected following manufacturers' recommendations.

Calculation of cell growth and cell death rates for HeLa cells

For cell proliferation and cell death experiments, control HeLa Tet Off, kis+ (pNATHA1i+) and kis+/kid+ (pNATHA2i+) cells were each cultured in six plates at equally low density and in selective medium. Doxycycline (0.1 μg/ml) was added to five of the plates of each sample 10, 7, 5, 3 or 1 days before the end of the experiment. Dead and mitotic cells were added back to the plates when medium was changed. All cells were resuspended in PBS, and stained with Trypan Blue to count total and dead cells. At least 250 cells for each sample were counted in a haemocytometer.

Annexin‐V staining

Cells were plated at identical cell densities in the absence or the presence of 0.1 μg/ml doxycycline and grown for 10 days on coverslips. Cells were stained with fluorescein‐linked Annexin‐V (green; Clontech) as recommended by the manufacturer, and then fixed and DNA stained with propidium iodide (red) before analysis.


We thank J.O.Mascorro‐Gallardo and R.Giraldo for plasmids and K.Butler for technical assistance. We are also indebted to B.Pimentel, D.Santamaría, J.Pérez‐Martín, Lorenzo and L.Ko Ferrigno for critical reading of the manuscript and helpful discussions, and to L.Sánchez‐Palazón, L.Howard, la Cueva and A.Zaldivar for encouraging support. G.C.M. was supported by the Spanish Ministry of Education and Science (BIO92‐CO2‐O2) and by Cancer Reseach UK. This work was supported by Cancer Research UK, the Louis Jeantet Foundation and the Comisión Interministerial de Ciencia y Tecnología (BIO94‐0707).


View Abstract