Infection of a plant by a pathogen induces a variety of defense responses that imply the action of several signaling molecules, including salicylic acid (SA), jasmonic acid (JA) and ethylene (E). Here we describe the role of ETHYLENE-RESPONSE-FACTOR1 (ERF1) as a regulator of ethylene responses after pathogen attack in Arabidopsis. The ERF1 transcript is induced on infection by Botrytis cinerea, and overexpression of ERF1 in Arabidopsis is sufficient to confer resistance to necrotrophic fungi such as B. cinerea and Plectosphaerella cucumerina. A positive co-operation between E and SA pathways was observed in the plant response to P. cucumerina. Infection by Pseudomonas syringae tomato DC3000, however, does not affect ERF1 expression, and activation of ethylene responses by ERF1 overexpression in Arabidopsis plants reduces tolerance against this pathogen, suggesting negative crosstalk between E and SA signaling pathways, and demonstrating that positive and negative interactions between both pathways can be established depending on the type of pathogen.
In addition to SA, ethylene (E) and jasmonic acid (JA) have also been shown to mediate pathogen responses, and JA/E-dependent (SA-independent) responses to pathogens have been reported (Knoester et al., 1998; Penninckx et al., 1996; Penninckx et al., 1998; Staswick et al., 1998; Thomma et al., 1998; Vijayan et al., 1998). Evidence for the involvement of JA/E-dependent (SA-independent) pathways in some defense reactions comes from analysis of the induction of PDF1.2 on infection by A. brassicicola. The antifungal peptide PDF1.2, which is active against this fungus, is locally and systemically induced upon infection. Induction does not depend on SA – it is not affected in NahG plants or in npr1/nim1/sai1 mutants, which are impaired in SA signaling (Cao et al., 1994; Delaney et al., 1995; Shah et al., 1997). However, induction is dependent on JA and E, as it is blocked in coi1 (JA-insensitive), ein2 and etr1-1 (E-insensitive) mutants (Penninckx et al., 1996; Penninckx et al., 1998). Genetic analysis using coi1 and ein2 mutants also demonstrated that both JA and E pathways work concomitantly in the activation of PDF1.2. Loss-of-function mutations of the COI1 gene prevented the induction of PDF1.2 by E, and mutations in EIN2 eliminated induction by JA (Penninckx et al., 1998).
The role of the JA/E-dependent pathway in the plant defense response has been further clarified by Thomma et al. (1998 and 1999), by demonstrating that JA/E-dependent and SA-dependent pathways in Arabidopsis are involved in resistance to different sets of pathogens. NahG plants or npr1 mutants, impaired in SA-signaling, which have increased susceptibility to the biotrophic fungus P. parasitica and the bacterium P. syringae (Cao et al., 1994; Delaney et al., 1994; Delaney et al., 1995; Glazebrook et al., 1996; Shah et al., 1997), were not affected in their defense response against necrotrophic fungi, such as A. brassicicola or B. cinerea (Thomma et al., 1998; Thomma et al., 1999). The opposite was true for coi1 mutants, demonstrating that JA is essential for defense against these pathogens, but is not necessary for P. parasitica resistance. As in the case of JA, E was previously shown to be dispensable for resistance to P. parasitica (Lawton et al., 1994), but important for responses to B. cinerea, as ein2 mutants showed increased susceptibility to this pathogen (Thomma et al., 1999). Nevertheless, despite the role of E (concomitantly with JA) in the activation of PDF1.2 in response to A. brassicicola, resistance to this fungus was unaffected in ein2 mutants, questioning the importance of E in the response to this pathogen, and suggesting the existence of a JA-dependent (E-independent) pathogen-response pathway.
Deciphering the genes and pathways involved in pathogen resistance has been mainly based on loss-of-function experiments. From a biotechnological viewpoint, however, it is more important to obtain gain-of-function mutants (or transgenic plants), because the character of interest can easily be transferred to crop plants by transformation (Wilkinson et al., 1997). In addition, engineering expression of transcription factors is a major objective in biotechnology, as modification of the expression of only one gene may drive to the modification of the expression of all genes regulated by this transcription factor. We previously reported the cloning of ETHYLENE-RESPONSE-FACTOR1, an early E-response gene whose expression is regulated by EIN3 and, in turn, regulates the expression of several pathogen responsive genes including b-CHI and PDF1.2 (Solano et al., 1998). In this work, we took advantage of the constitutive expression of ethylene-response genes in ERF1-overexpressing Arabidopsis transgenic plants to investigate the role of ERF1 in the regulation of E responses to pathogens. Here we show that ERF1 is upregulated upon infection of Arabidopsis by B. cinerea, and that gain-of-function of ERF1, by constitutive expression in transgenic Arabidopsis plants, is sufficient to confer resistance to several necrotrophic fungi.
ERF1 expression is induced by B. cinerea infection
It has previously been reported that ein2 Arabidopsis mutants are more susceptible than wild-type (WT) plants to infection by the necrotrophic fungus B. cinerea, suggesting that E may be involved in resistance to this fungus (Thomma et al., 1999). As ERF1 is a downstream component of the E signaling pathway (Solano et al., 1998), we tested whether ERF1 expression would be affected by B. cinerea infection.
Four-week-old WT, NahG and ein2-5 plants, and three independent 35S::ERF1 transgenic lines, were sprayed with a suspension containing 2 × 105 spores ml−1 of the fungus. Total RNAs were extracted from inoculated and mock-inoculated leaves at different times after infection, and the expression of ERF1 was analyzed by Northern blot. As shown in Figure 1(a), infection of WT plants resulted in an increase in the steady-state mRNA levels of ERF1, which reached a maximum at 5–7 days post-inoculation. The induction of ERF1 expression paralleled the appearance of necrotic lesions in the leaves (data not shown). In ein2 mutants, which are blocked in the E-response pathway, the induction of ERF1 by the fungus was strongly reduced whereas ERF1 expression in NahG plants was similar to that observed in WT plants, indicating that induction of ERF1 expression by the fungus depends on E, but is independent of SA (Figure 1a).
A similar expression pattern in response to B. cinerea infection was observed for b-CHI and PDF1.2, two genes that are likely targets of ERF1 (Solano et al., 1998). Both genes were upregulated by B. cinerea infection in WT plants, and this induction depended on an intact E-signaling pathway but not on SA (Figure 1a). By contrast, all three transgenic lines that constitutively express ERF1 also showed a high constitutive expression of b-CHI and PDF1.2 in mock-inoculated and infected plants (a representative line is shown in Figure 1a). Furthermore, the level of constitutive expression of both genes (b-CHI and PDF1.2) in the ERF1-expressing lines tested was comparable to that reached in WT plants after pathogen infection (Figure 1a).
To test the specificity of the ERF1 response to B. cinerea infection, we analyzed the expression of ERF1 after inoculation with the compatible bacterial pathogen P. syringae tomato DC3000. Four-week-old WT, NahG and ein2-5 plants, and three independent 35S::ERF1 transgenic lines, were infiltrated with a bacterial suspension (0.01 OD600), and expression of ERF1 was monitored by Northern blot. As shown in Figure 1(b), ERF1 expression was not induced in any of the plants, whereas PR1 gene expression was, as expected, upregulated by P. syringae tomato DC3000 infection in both WT and ein2-5, but not in NahG plants. Expression of b-CHI and PDF1.2 was weakly induced after P. syringae infection in WT and NahG plants, but not in the ein2-5 mutant (Figure 1b). In the three ERF1 transgenic lines tested, the expression of b-CHI and PDF1.2 was constitutive and essentially not affected by the bacterial infection, whereas PR1 expression was induced by P. syringae and not expressed in mock-inoculated plants (a representative line is shown in Figure 1b).
These results indicate that the observed induction of ERF1 expression after infection with B. cinerea was not a general, non-specific response of this gene to pathogen infection. The absence of response of ERF1 to P. syringae infection is consistent with the lack of induction of this gene in plants treated with 0.3 mm of the SAR activator BTH (data not shown; Gorlach et al., 1996).
Constitutive expression of ERF1 in Arabidopsis enhances tolerance to B. cinerea infection
The induction of ERF1 expression by B. cinerea and the regulation by ERF1 of the antifungal b-CHI and PDF1.2 proteins prompted us to test whether constitutive expression of ERF1 would affect plant tolerance to this fungus. Thus we compared the susceptibility of different E-insensitive mutants and WT plants with that of 35S::ERF1 transgenic plants. NahG and coi1 plants, whose susceptibility to the fungus has been analyzed previously (Thomma et al., 1999), were included as controls in the experiment.
Four-week-old plants grown in soil were challenged with a suspension containing 2 × 105 spores ml−1 of B. cinerea, and the progress of the infection was followed over 10 days. As shown in Figure 2, ERF1-expressing lines showed a clear reduction in infection symptoms compared with WT plants, with 40–70% of plants not showing any macroscopic or microscopic necrosis, as assessed by trypan blue staining of the inoculated leaves (Figure 2a,b). Overexpression of ERF1 in an ein2-5 mutant background resulted in a reversion of the susceptibility of this mutant to the fungus, confirming the role of ERF1 downstream of ein2 in the E-signaling pathway (Figure 2a,b; Solano et al., 1998). The differences in level of resistance of the different transgenic lines tested correlated with the strength of the ERF1-overexpression phenotype of the plants (not shown). The susceptibility of NahG plants to B. cinerea was similar to that of WT plants, whereas coi1 mutants were more susceptible, confirming that resistance to this necrotrophic fungus does not depend on SA, but needs an intact JA signal-transduction pathway. Furthermore, in addition to ein2-5 plants, ein3-1 mutants also showed higher susceptibility than WT to infection (Figure 2a). Susceptibility of ein3-1 to the fungus was lower than that of ein2-5, which correlated with the weaker E-insensitivity of this mutant compared to ein2 (Figure 2a; Chao et al., 1997).
To analyze further the ERF1-mediated resistance to B. cinerea, we determined the presence of the fungus in the plant by trypan blue staining. As shown in Figure 2(b), fungal hyphae and necrotic lesions caused by the fungus were detectable in WT and ein2-5 plants but, as expected from the macroscopic evaluation of infection, they were more numerous in ein2 mutants. In contrast, no fungal growth or necrotic lesions were detectable in the majority of leaves of Col-0;35S::ERF1 or ein2-5;35S::ERF1 plants.
These results demonstrate that in Arabidopsis an intact E-signaling pathway, together with the JA-signaling pathway, is required for full resistance to B. cinerea, and that overexpression of ERF1 is sufficient to increase resistance to this necrotrophic fungus.
ERF1 overexpression enhances tolerance against the necrotrophic fungus P. cucumerina
To test whether the tolerance to B. cinerea induced by overexpression of ERF1 is also effective against other necrotrophic pathogens, we analyzed the susceptibility of the ERF1-transgenic lines to Plectosphaerella cucumerina (Palm et al., 1995). This fungus is commonly isolated as a saprophyte from soils, and has a wide host range that includes various cucurbits, tomato and sunflower (Smither-Kopperl et al., 1999). Arabidopsis plants inoculated with P. cucumerina develop small, brown spots on the leaves and petioles 3–4 days after infection, which spread very quickly through the leaves and vascular system and lead to the complete decay of the plant 8–12 days after infection (F. Llorente and A.M., unpublished results).
Soil-grown, 4-week-old transgenic lines, as well as WT, coi1, NahG and ein2-5 plants, were challenged with a suspension of 5 × 106 spores ml−1 of P. cucumerina, and symptoms were analyzed on different days after inoculation. As shown in Figure 3(a), ein2-5 mutants were more susceptible than WT plants to infection by P. cucumerina, with approximately 80% of the ein2-5 inoculated plants decaying or showing profuse necrosis 8 days after inoculation, suggesting that E is also involved in the activation of plant defenses against this fungus. In contrast to the results obtained with B. cinerea, NahG plants also showed an enhanced susceptibility to P. cucumerina, with 100% of the inoculated plants decaying or showing profuse leaf necrosis 8 days after inoculation. Trypan blue staining of the inoculated leaves revealed the presence of spreading necrotic lesions and a large amount of fungal mycelium, mainly in the petioles, which was more pronounced in NahG and ein2-5 than in WT plants (Figure 3b; data not shown).
As previously reported (Thomma et al., 2000), coi1 mutants also showed enhanced susceptibility to P. cucumerina, which was similar to that of the NahG plants. These results indicate that SA, E and JA signal-transduction pathways are necessary in Arabidopsis to mount an effective defense response against P. cucumerina– although, as described by Thomma et al. (2000), exogenous application of MeJA to NahG and ein2-1 mutants could bypass the requirement of SA and E signaling in these plants.
As previously observed with B. cinerea, the three ERF1-expressing lines tested showed enhanced tolerance to P. cucumerina compared to WT plants, with a low percentage of the transgenic plants showing macroscopic or microscopic necrotic symptoms, as assessed by trypan blue staining of inoculated leaves (Figure 3a–c). The levels of resistance of the transgenic lines tested correlated with the strength of the ERF1-overexpression phenotype of the plants, as observed for B. cinerea tolerance (not shown). These results indicate that ERF1-induced tolerance is not restricted to B. cinerea, and further suggest a general role of the E signal-transduction pathway in plant defense responses. The results also show that although intact SA, E and JA signal-transduction pathways are needed in Arabidopsis for effective control of P. cucumerina infection, constitutive over-activation of the E pathway is sufficient to obtain enhanced tolerance to the fungus.
ERF1 overexpression does not increase tolerance against P. syringae tomato DC3000
It has been suggested that E/JA-dependent and SA-dependent pathways are involved in responses to different sets of pathogens (Thomma et al., 1998; Thomma et al., 1999). To test whether activation of E responses by ERF1-overexpression would improve Arabidopsis resistance against SAR-inducing pathogens, we checked the tolerance of the ERF1-overexpressing lines to the compatible P. syringae tomato DC3000, and compared it with that of WT and of NahG plants which show enhanced susceptibility to this bacterial pathogen (Delaney et al., 1994).
Four-week-old plants were challenged with a suspension of the bacteria (0.01 OD600), and bacterial growth in the inoculated plants was determined. As shown in Figure 4, P. syringae tomato DC3000 growth in the three ERF1-transgenic lines tested was higher than in WT plants, but lower than in the NahG plants which, as expected (Delaney et al., 1994), supported a higher level of bacterial growth than WT plants. The necrotic symptoms observed in the plants 4 days after infection with P. syringae tomato DC3000 correlated well with the differences observed in bacterial growth (data not shown). These results indicate that ERF1 overexpression does not increase tolerance against P. syringae tomato DC3000, and may have a detrimental effect on the resistance of Arabidopsis plants to the bacteria. This negative effect on resistance to P. syringae tomato DC3000 was, however, much lower than that resulting from depletion of SA in NahG plants. As previously reported for other P. syringae-compatible strains (Bent et al., 1992; Clarke et al., 2000), bacterial growth in the ein2-5 mutant did not differ significantly from that in WT plants (data not shown).
The induction of plant defense responses upon pathogen infection involves the action of several signaling molecules, including SA, JA and E (Glazebrook, 1999; Thomma et al., 2001). Here we describe the role of ERF1 as a regulator of ethylene responses after pathogen attack, and show that constitutive expression of ERF1 in Arabidopsis is sufficient to confer resistance to the necrotrophic fungi B. cinerea and P. cucumerina. In addition, overactivation of the E pathway by constitutive expression of ERF1 results in a negative effect on plant resistance to Pseudomonas syringae tomato DC3000, suggesting a negative crosstalk between the E and SA pathways.
The involvement of E in pathogen defense has been controversial for over a decade, and is beginning to be clarified. Depending on the type of pathogen and plant species, E appears to have opposing roles in the progress of disease. Exogenous application of the hormone enhances in some cases, and decreases in other cases, resistance to different pathogens (Brown and Lee, 1993; El-Kazzaz et al., 1983a; El-Kazzaz et al., 1983b). More recently, genetic analyses using different E-insensitive mutants in different plant species have led to similar conclusions. E-insensitivity leads to an apparent reduction in disease symptoms caused by the bacteria Xanthomonas sp. and Pseudomonas sp. in Arabidopsis and tomato plants (Bent et al., 1992; Lund et al., 1998), and to a partially enhanced tolerance to Pseudomonas sp. and Phythophtora sp. in soybean (Hoffman et al., 1999). In contrast, E insensitivity has been shown to enhance the susceptibility of soybean plants to several fungi, including Septoria glycines and Rhizoctonia solani (Hoffman et al., 1999); of tobacco plants to the soilborne fungus Pythium sp. (Knoester et al., 1998); and of Arabidopsis to the necrotrophic fungus B. cinerea (Thomma et al., 1999) and the soft-rot bacterium Erwinia carotovora (Norman-Setterblad et al., 2000). In addition to these apparently opposed roles of E in pathogen response, it has also been shown that disease symptoms and progression caused by some pathogens are not affected by E insensitivity, as observed in Arabidopsis for A. brassicicola and P. parasitica (Lawton et al., 1994; Thomma et al., 1999), which require intact JA and SA pathways, respectively, for efficient control of pathogen infection.
This apparent discrepancy among the roles of E in different plant–pathogen interactions may be reconciled by the different infection mechanisms of different pathogens, and by the fact that E is not only involved in pathogen response, but is a hormone implicated in many general aspects of plant development including senescence, cell death, ripening and chlorosis (Abeles et al., 1992). The detrimental effect of E in some plant–pathogen interactions may be an indirect consequence of the involvement of the hormone in the above-mentioned developmental processes. Although E synthesis occurs in response to many stresses, including pathogen attack (Penninckx et al., 1998), the emerging picture is that the E produced after pathogen infection is involved as a defense signal only in some types of plant–pathogen interactions, and may be detrimental in others. In line with this idea, we have shown that activation of E signaling may be detrimental for the interaction between P. syringae tomato and Arabidopsis, whereas it is essential for resistance against B. cinerea and P. cucumerina. Thus the loss of E signaling by mutations of genes of the pathway, or activation of E signaling by constitutive expression of one of its downstream components, ERF1, have opposite effects (susceptibility or resistance, respectively) in the response of Arabidopsis to B. cinerea and P. cucumerina, further showing that E is indeed essential to mount efficient defenses against these fungi. The activation by B. cinerea of ERF1 and its downstream targets, b-CHI and PDF1.2, which have antifungal activity (Penninckx et al., 1996), reinforces this idea.
Antagonistic interactions between SA-dependent and JA-dependent signaling pathways have already been described. SA has been shown to inhibit JA synthesis and activation of JA-responsive genes in the context of the wound response in potato and tomato (Doares et al., 1995; Harms et al., 1998; Niki et al., 1998; O'Donnell et al., 1996). Conversely, JA treatment inhibits SA-mediated expression of some PR genes (Niki et al., 1998). In addition, mutants affected in the accumulation or response to SA show enhanced expression of E/JA regulated genes, which in some cases has been shown to be reverted by addition of isonicotinic acid, an inducer of SA signaling (Clarke et al., 2000; Penninckx et al., 1996). Moreover, E insensitivity does not negatively affect SAR response in Arabidopsis and may actually potentiate SA-induced PR-1 gene expression (Lawton et al., 1994; Lawton et al., 1995). In the context of ozone-induced cell death, an antagonistic relationship between SA and JA has also been reported. SA mediates superoxide-dependent cell death, which is alleviated by JA (Rao et al., 2000).
All the cases noted above are examples of antagonistic interactions at the molecular level. These negative interactions may also operate in vivo in the resistance to a particular pathogen, as highlighted by the reduced tolerance to P. syringae tomato DC3000 of the ERF1-expressing transgenic lines, showing that overactivation of E signaling (at least by ERF1 overexpression) may have a negative effect on tolerance to P. syringae tomato DC3000.
Despite the above-mentioned antagonistic interactions, there are also examples in which both SA-dependent and JA/E-dependent pathways co-operate on defense-related gene expression. The involvement of SA, JA and E pathways in Arabidopsis resistance to P. cucumerina described in this work represents a clear in vivo example of this co-operation, and of the involvement of these three pathways in Arabidopsis defense mechanisms against a fungal pathogen. Consistent with our results, a recent cDNA microarray analysis of the expression of over 2300 genes in Arabidopsis plants treated with A. brassicicola, SA, JA or E has revealed that these stimuli mostly overlap rather than antagonize (Schenk et al., 2000). Clarke et al. (2000) have shown that components of the E and JA pathways work in combination with an SA-mediated (NPR1-independent) resistance response in the defense of Arabidopsis against P. syringae and P. parasitica.
So a general antagonistic mechanism of regulation of SA and E/JA pathways appears to be an oversimplification. Plants have probably evolved different regulatory mechanisms (antagonism and co-operation) to fine-tune particular responses to different types of pathogens.
The data presented here demonstrate the role of ERF1 in pathogen defense in Arabidopsis as a regulator of E responses to several necrotrophic pathogens, and suggest its application in engineering resistance against these pathogens in crop plants. Nevertheless, due to the enhanced susceptibility of 35S::ERF1 plants to P. syringae tomato DC3000, more work is required to separate functions and side-effects before any biotechnological application can be achieved.
The Arabidopsis thaliana wild-type (WT), mutant and transgenic plants used throughout this work were on the Columbia (Col-0) ecotype. The generation and characterization of 35S::ERF1 transgenic lines have been described previously (Solano et al., 1998). The fungal pathogen Botrytis cinerea, strain 1, was from the Escuela Tecnica Superior Ingenieros Agronomos (ETSIA) collection (Madrid). Dr B. Mauch-Mani (University of Fribourg, Switzerland) kindly provided the Plectosphaerella cucumerina isolate. The bacterium Pseudomonas syringae tomato DC3000 was from the Novartis Corporation collection (Research Triangle Park, NC, USA).
Seeds from plants were surface-sterilized in ethanol 90%, soaked for 5 min in 10% bleach and 0.5% Tween 20, washed five times in sterile water and grown on dish plates containing MS agar medium for 1 week. Then they were transplanted into soil and grown in a phytochamber with a 10 h light/14 h dark photoperiod, 60% relative humidity, and 24°C day and 22°C night temperatures.
Botrytis cinerea and P. cucumerina were grown on potato dextrose agar medium (Difco, Detroit, MI, USA) at 28°C for 8 days, and spores were collected in sterile water and stored at −80°C in 20% glycerol. Pseudomonas syringae tomato DC3000 was cultivated at 28°C on nutrient broth (Oxoid, Basingstoke, UK) with 100 µg l−1 rifampicin (Sigma, MI, USA).
Plant infection with pathogens
Four-week-old plants grown as indicated above were used for all the pathogen inoculation experiments. For inoculation with B. cinerea and P. cucumerina, plants were sprayed with a suspension of 2 × 105 and 5 × 106 spores ml−1, respectively. Mock inoculations were done with sterile water containing an equivalent amount of glycerol to the fungal spore suspension used for infection. Following inoculation, plants were covered to maintain high humidity and placed in a growth chamber under the same temperature and illumination conditions as before inoculation. The progress of fungal infection was followed for 10 days by viewing the development of necrosis in the infected leaves, which can be detected 4–5 days after inoculation with both fungal pathogens, and infection ratings were assigned to the inoculated plants (0–3: 0, no infection/necrosis; 1, 1–4 leaves showing some necrosis; 2, 5–10 leaves showing necrosis; 3, dead/decayed plant). Lactophenol trypan blue staining (Keogh et al., 1980) of individual leaves was carried out to detect fungal growth within leaf tissue and to confirm the disease rating assigned. At least 15 plants per genotype were inoculated in each experiment. Experiments were repeated at least three times with similar results.
For inoculation of P. syringae tomato DC3000, plants grown as above were dipped for 6 min in a bacterial suspension (0.01 OD600) in 10 mm MgCl2 containing 0.1% silwett L-77 (Lehle Seeds, TX, USA), and then transferred to the same growth conditions and 100% relative humidity. Mock inoculations were performed with 10 mm MgCl2 containing an identical amount of silwett L-77. Infected leaves were collected immediately after infection (day 0), and 2 and 4 days later. Four leaves from four plants of each genotype were harvested and homogenized in 10 mm MgCl2, and appropriate tenfold dilutions were plated in nutrient broth medium containing 100 µg ml−1 rifampicin in order to count bacterial colony-forming units.
Northern blot analysis
Total RNA was extracted from frozen tissues, harvested at different times after treatment, by phenol/chloroform extraction followed by precipitation with 1.5 m LiCl, as described by Lagrimini et al. (1987). Extracted RNA was subjected to electrophoresis on 1.5% formaldehyde–agarose gels, and blotted to Hybond–N+ membranes (Amersham, UK). ERF1 probes were labeled with 100 µCi of α-32P-dCTP. All other probes were labeled with 50 µCi of α-32P-dCTP. Blots were exposed for 24 h in a Phosphorimager screen (Molecular Dynamics, Amersham, UK), and the background was reduced digitally in the cases of PDF1.2 and rRNA. All probes used have been described previously (Solano et al., 1998).
We thank J. J. Sánchez-Serrano, F. García-Olmedo, J. Paz-Ares, C. Castresana, P. Rodríguez-Palenzuela, M. Pernas and O. Lorenzo for critical reading of the manuscript and stimulating discussions. We also thank Gemma López for technical assistance. This work was financed by grant 007B/0002/1999 to A.M., and grant 07G/0048/2000 to R.S. from the Comunidad de Madrid (Spain). M.B.-L. was financed by BioCT97-2120 (DGXII-SSMI) from the EU.