Genetic screens have identified a number of genes that regulate abscisic acid (ABA) responsiveness in Arabidopsis. Using a combination of suppressor screens and double mutant analysis, we have determined a genetic relationship for a number of these ABA response loci. Based on germination in the presence of exogenous ABA, the ABI1 and ABI2 phosphatases act at or upstream of the ERA1 farnesyl transferase and the ABI3 and ABI5 transcription factors act at or downstream of ERA1. In contrast with ABI3 and ABI5, the ABI4 transcription factor appears to act at or upstream of ERA1. Based on reporter gene constructs, the upstream regulation of ABI3 by ERA1 occurs at least partially at the level of transcription, suggesting that this lipid modification is required to attenuate ABI3 expression. Similar experiments also indicate that ABI3 is auxin inducible in lateral root primordia. Related to this, loss-of-function abi3 alleles show reduced lateral root responsiveness in the presence of auxin and an auxin transport inhibitor, and era1 mutants have increased numbers of lateral roots. These results suggest the possibility that genes identified through ABA responsive germination screens such as ERA1 and ABI3 have functions in auxin action in Arabidopsis.
Abscisic acid (ABA) is a hormone that regulates many complex plant processes including the induction and maintenance of seed dormancy, inhibition of root growth, control of stomatal closure, and protection of the plant from a variety of environmental stresses (Davies, 1995). Although the structure and biosynthesis of ABA is largely understood, how ABA regulates these multiple plant responses is just now beginning to be revealed through genetic analysis in Arabidopsis thaliana. To date, genetic screens for ABA-hypersensitive mutants have indicated that processes including farnesylation (era1), inositol signaling (fry1), and RNA metabolism (abh1, sad1, hyl1) are required to attenuate the ABA signal (for review, see Finkelstein et al., 2002). Additional screens for mutations that confer a reduced sensitivity to ABA have identified homologous type 2C phosphatases (ABI1, ABI2) and three disparate transcription factors (ABI3, ABI4 and ABI5). Construction of different combinations of overexpressing and loss-of-function alleles of ABI3–5 indicates that the overall ABA responsiveness of the mature seed results from a complex interaction between these transcription factors (Soderman et al., 2000).
Although first described as seed specific for altered ABA sensitivity, recently more detailed phenotypic analysis and expression studies of ABI3, ABI4 and ABI5 indicate that these genes may have functions outside of modulating ABA seed sensitivity. For example, ABI4 and ABI5 genes have been shown to have functions in both sugar and salt responses and in early seedling growth after germination (Gibson, 2000; Lopez-Molina et al., 2001; Quesada et al., 2000). Both genes are expressed outside of seed development, and loss-of-function mutations in each result in altered lateral root branching in response to nitrate (Signora et al., 2001). In the case of ABI3, it appears that this gene plays roles in plastid development and bud dormancy and interacts with genes involved in light regulation (Rohde et al., 2000, 2002). ABI3 expression is detected in many quiescent tissues including the receptacle of flowers, the axils of pedicels and axillary flower bracts, the abscission zone of siliques and rosette leaves, and stipules (Parcy et al., 1994; Rohde et al., 1999).
Although the reasons for the complex expression patterns of ABI3 are not clear, the maize ortholog of this gene, VIVIPAROUS1 (VP1), has been shown to have multiple regulatory domains (Hoecker et al., 1995) suggesting that this gene has multiple functions. Comparisons of ABI3/VP1 orthologs have identified four highly conserved amino acid domains: A1, a domain in the acidic N-terminal of the protein; and three basic regions: B1–3 (Giraudat et al., 1992). The B1 domain has been shown to be capable of protein–protein interactions with ABI5, while B2 and B3 can bind DNA (Nakamura et al., 2001; Suzuki et al., 1997). The B3 DNA-binding domain is unique among plants and is conserved among several gene families (Ulmasov et al., 1997). One of these is the auxin response factor (ARF) family. Mutations in a number of ARF genes have been shown to alter auxin responses in Arabidopsis (for review, see Liscum and Reed, 2002). Interestingly, in the presence of exogenous ABA, plants misexpressing VP1 are insensitive to auxin-induced lateral root formation suggesting that the VP1/ABI3 transcription factor may define an interaction node between ABA and auxin signaling (Suzuki et al., 2001).
To further understand the roles of ABI3 and ERA1 in plant hormone signaling, we have used a combination of suppressor and epistasis studies between ABA response mutants in Arabidopsis to define a genetic framework. At the level of seed ABA responsiveness, ABI1 and ABI2 act at or upstream of ERA1. By contrast, ABI3 and ABI5 act at or downstream of ERA1. ABI4 appears to act at or upstream of ERA1 and may define a separate genetic pathway from that of ABI3 and ABI5. These genetic studies suggest that ERA1 is a negative regulator of ABI3 and ABI5, and this is supported by the increased expression of an ABI3::GUS transgene in an era1 mutant background. These studies also suggest that ABI3 transcription may be regulated by auxin, and that auxin signaling in lateral root development may have an ABI3-dependent component. We discuss a model for a general role of ABI3 in seed and root development, and the possibility that ABA-sensitizing screens using seed germination as an assay may also enrich for auxin response mutants in Arabidopsis.
Genetic interactions between genes affecting seed ABA responsiveness
Under our assay conditions, wild-type Arabidopsis germination is permitted on concentrations of ABA lower than 1.0 μm. A mutant line is considered ABA-insensitive (abi) compared to wild type if it germinates on 3 μm ABA and has enhanced ABA sensitivity (era) if it cannot germinate on 0.3 μm ABA. To identify suppressor mutations of era1, mutagenized M2 seeds derived from 20 000 M1 era1 plants were screened for germination on 0.3 μm ABA. Of the 124 lines that re-tested in the M3 generation for suppression of era1, eight were insensitive to 3 μm ABA. Because these eight lines showed strong ABA insensitivity, they could represent new alleles of known abi mutations. Partial allelism tests were performed with a known abi3 allele (abi3-1), and three suppressor lines were classified as new alleles of abi3 (data not shown). One of these alleles was sequenced and shown to contain a missense mutation in the B3 domain of ABI3 (glycine to arginine at position 666). To corroborate suppression interaction between era1 and abi3, F2 plants from a cross of era1-2 and a severe allele abi3-6 were screened for protruding carpels, a common phenotype of homozygous era1-2 flowers (Bonetta et al., 2000). A number of these plants segregated the characteristic homozygous abi3-6 underdeveloped green seed phenotype. Plants producing green seed and protruding carpels were propagated and molecularly genotyped for the era1-2 and abi3-6 mutation, and from these, 50 immature green seeds were placed on 0.3 and 3.0 μm ABA, respectively. All seeds germinated on these concentrations of ABA, verifying that abi3-dependent ABA insensitivity was epistatic to era1 (Table 1).
Table 1. ABA responsiveness of single and double mutant seeds as measured by germination on ABA
0.3 μm ABA
3.0 μm ABA
Numbers in the numerator represent germinated seed over the total number of seeds plated. ND: not determined.
The other era1-2 suppressor lines were recessive for ABA insensitivity and not allelic to either abi4 or abi5 (data not shown). The lack of any new abi4 or abi5 alleles from the suppressor screen was not surprising because the screen was not saturating. Hence, double mutants between era1 and known alleles of abi4 and abi5 were constructed to test epistasis. Among 182 F2 seeds from an era1-2 by abi5-1 cross, 55 germinated on 3 μm ABA, which yielded a characteristic ratio for recessive epistasis for ABA insensitivity (χ2 = 2.64; P > 0.90). The era1 abi5 double mutants were established and tested for germination on 0.3 and 3 μm ABA (Table 1). These double mutants were insensitive to both concentrations indicating that the abi5 mutation is epistatic to the era1 mutation on these concentrations of ABA. By contrast, of the 126 F2 seeds generated from an abi4-1 by era1-2 cross, only 15 of the expected 31 germinated on 3 μm ABA, suggesting that abi4 is not epistatic to the era1 mutation. Furthermore, 35 of 87 F2 seeds did not germinate on 0.3 μm ABA suggesting that era1 is epistatic to abi4. This was verified by the identification of the era1 abi4 double mutant as being sensitive to 0.3 and 3 μm ABA (Table 1). Taken together, these results suggest that ABI3 and ABI5 are at or genetically downstream of ERA1 and that ABI4 is at or genetically upstream of ERA1.
The recessive nature of era1, ABI3, ABI4 and ABI5 mutations made the identification of double mutants between these mutations relatively easy. However, the identification of era1abi1-1 and era1abi2-1 double mutants was more complicated because of the dominant nature of the abi1 and abi2 mutations (Leung et al., 1997). F2 progeny from a cross between abi1-1 and era1-2 were plated on 3 μm ABA and 121 of the 187 seeds germinated. These numbers fit a 9 : 7 ratio (χ2 = 1.64; P > 0.95) suggesting that era1-2 is epistatic to abi1-1. To confirm this, individual F3 progeny were molecularly identified that were homozygous for both the abi1-1 and era1-2 mutations. These double mutants were sensitive to 0.3 and 3 μm ABA as measured by germination, indicating that the era1-2 mutation is epistatic to abi1-1 (Table 1). The close linkage of ABI2 and ERA1 genes (15 map units, chromosome 5) meant that a simple genetic ratio in the F2 progeny would not be expected. Consistent with this, the expected ratio of 9 : 7 insensitive to sensitive on 3 μm ABA was not observed (321 : 172, χ2 = 16.6). We therefore screened F2 seeds for sensitivity to 0.3 μm ABA to enrich for era1-2 homozygous mutants. From these seeds, 31 plants were genotyped for the molecular polymorphism that identified the abi2-1 mutation, and 10 plants were identified as heterozygous for the abi2-1 mutation. Four of these plants were allowed to self and F2 progeny were molecularly genotyped for abi2-1 homozygous plants. All progeny from these double mutant lines showed ABA sensitivity to 0.3 and 3 μm ABA, indicating that the era1 mutation is epistatic to abi2-1 mutation (Table 1). Together, these results show that ERA1 acts at or downstream of ABI1 and ABI2.
ERA1 negatively regulates ABI3 transcription in vegetative tissues
Loss-of-function mutations in ERA1 confer an enhanced ABA sensitivity, suggesting that farnesylation is a negative regulator of ABA sensitivity in Arabidopsis. As shown above, ABI3 and ABI5 may genetically interact downstream of ERA1, thus, farnesylation may regulate seed ABA sensitivity by modulating these gene products. To test this, the era1 mutation was crossed into a transgenic plant containing the ABI3::GUS reporter construct, and β-glucuronidase activity was monitored by histochemical staining (Parcy et al., 1994). Under our staining conditions, era1 mutant tissues showed increased GUS staining compared to wild type (Figure 1). Unexpectedly, the roots of era1 mutants also showed a punctuate staining pattern that was not observed in the wild type macroscopically (Figure 1). Using differential interference contrast microscopy (DIC), however, it was possible to discern weak GUS staining in wild type in a pattern that corresponds to specific pericycle cells of the root that are undergoing differentiation into lateral root primordia (Figure 2). Lateral roots develop when a differentiated pericycle cell adjacent to xylem begins an asymmetric transverse division leading to the formation of a convex root primordium. The lateral root primordium continues to grow through the overlying cell layers to eventually emerge from the parent root. After emergence, the lateral root continues to grow from an autonomous fully functional meristem (Laskowski et al., 1995; Malamy and Benfey, 1997). In wild-type roots, weak GUS staining is first visualized just after the pericycle cell has committed to becoming a lateral root primordium cell at approximately two-cell stage when the layer is midway through the endodermis. This pattern occurs upto approximately the five-cell-layer stage at which time the exterior border of expression appears as a crescent with staining at the outer border of the primordium and decreased staining towards the middle (Figure 2). By the time the lateral root emerges from the epidermal layer, the GUS staining pattern dissipates and is no longer visible by DIC. Therefore, it appears that in wild-type plants, ERA1 may downregulate transcription of ABI3 in the embryo, apical meristem and lateral roots, which is consistent with the negative regulation by ERA1 on ABI3 identified in the suppression studies.
Lateral root defects in ERA1 and ABI3 loss-of-function mutations
The aberrant ABI3::GUS transgene expression patterns observed in era1 mutant roots suggest that ABI3 may have functions in lateral root development. However, the number of lateral roots observed in abi3-6 loss-of-function mutants was similar to that seen in wild-type grown on minimal media (Figure 3a). This is in contrast to era1-2, which does show an increased number of lateral roots versus wild type (Figure 3a). The lack of a lateral root phenotype in abi3 mutants versus era1 may simply mean that the role of ABI3 on lateral root development is at best subtle under normal growth conditions. We, therefore, decided to test the response of abi3-6 root growth in the presence of indole acidic acid (IAA) and the auxin transport inhibitor N-1-naphthylphthalamic acid (NPA) to see if any phenotypes could be observed under conditions that perturbed auxin-regulated lateral root development. Exogenous application of auxin increases lateral root initiation, whereas inhibition of auxin transport inhibits lateral root initiation. Under these growth conditions, both genotypes responded to IAA by initiating new lateral roots. However, abi3-6 plants required higher concentrations of IAA to initiate lateral root numbers comparable to wild type (Figure 3b). Similarly, application of the auxin transport inhibitor NPA inhibited lateral root initiation in both wild type and in abi3-6, but again mutant roots were less responsive to the NPA-induced repression of lateral root initiation (Figure 3c). The abi3-6 mutant roots still retained the characteristic agravitropic response to NPA, indicating that the inhibitor was being perceived by the mutant roots (data not shown). Thus, it appears that ABI3 function is required at some level for correct auxin responsiveness in lateral roots in Arabidopsis.
ABI3::GUS expression is responsive to both auxin and ABA application in the root
To further explore this relationship of ABI3 to auxin and ABA, the expression of the ABI3::GUS transgene was monitored in response to exogenous concentrations of IAA, the synthetic auxin 2,4-dichlorophenoxyacetic acid (2,4D) and ABA. GUS precipitate was visible within 1 h of 2,4D application, and was at a maximum after 1 day of exposure (Figure 4). Unlike wild-type roots grown on minimal media, the domain of GUS accumulation was expanded in hormone-treated roots with the vascular tissue and the endodermis along with the lateral root primordia showing increased blue staining (Figure 4). In longer exposures upto 48 h, much of the ABI3::GUS expression pattern corresponded to a series of lateral roots initiated adjacent to each other in response to the hormone treatment (Figure 4). These patterns of expression were also observed for roots exposed to 1 μm IAA, but the intensity of the GUS staining was reduced substantially and a much longer exposure to IAA was required to see similar results (data not shown).
The ABI3 transgene reporter is also responsive to concentrations of ABA as low as 0.4 μm in the lateral root (Figure 4). However, in contrast to the auxin induction kinetics, 1-day incubations were required to see a clear GUS precipitate in the lateral root tissue macroscopically. Furthermore, although ABA application increased GUS expression over time, unlike 2,4D application ABA application did not lead to any increase in the domain of GUS staining outside the region of lateral root development (Figure 4). In this study, ABA did not induce ABI3::GUS expression in any other vegetative tissue, which is consistent with previous reports using this promoter fusion transgenic plant (Parcy et al., 1994).
Auxin and ABA act synergistically to inhibit germination
A physiological process such as germination is positively and negatively influenced by multiple environmental and hormonal signals. The observation that ABI3::GUS expression can be induced by auxin and that loss-of-function mutations in ABI3 show auxin-related phenotypes in lateral root development suggests that ABI3 may be involved in multiple hormonal signaling pathways. This brings into question the plethora of phenotypes observed in severe loss-of-function abi3 mutants. For example, although loss-of-function mutants in abi3 and ABA auxotrophs are both non-dormant, only abi3 mutants show reduced desiccation tolerance, loss of seed storage reserves, and the inability to complete late embryogenesis (Nambara et al., 1992; Ooms et al., 1993). Possibly, the additional phenotypes seen in abi3 mutants could be auxin related and even the decreased ABA sensitivity of abi3 alleles may be because of a change in auxin sensitivity. To test the possibility that auxin can influence seed ABA responsiveness in Arabidopsis, wild-type seeds were exposed to increasing concentrations of ABA in the presence of low concentrations of exogenous IAA. As expected, increasing concentrations of ABA progressively inhibited germination, but the presence of auxin shifted the ABA sensitivity so that less ABA was required to inhibit germination (Figure 5). Auxin only affects seed germination in the presence of ABA and does not directly inhibit seed germination.
Genetic interactions between ABA response mutations
If mutations exist that confer opposite signaling states in a particular process, double mutants can often be used to order genes into a genetic pathway (McCourt, 1999). By constructing double mutants containing era1-enhanced ABA sensitivity with a variety of ABA-insensitive abi mutations, we have been able to order a number of ABA-responsive loci into a genetic framework (Figure 6). This framework suggests that ERA1 genetically acts at or downstream of the ABI1/2 phosphatases but functions at or upstream of the ABI3 and ABI5 transcription factors. The genetic relationship between the ABI1/2 and ERA1 mutations are consistent with the ordering of these genes using stomatal conductance and patch clamp analysis (Pei et al., 1998). The assignment of ABI3 and ABI5 in the same place in this genetic pathway reflects recent work showing that these two transcription factors interact in a yeast two-hybrid assay (Nakamura et al., 2001). Furthermore, the fact that ERA1 is upstream of ABI3 as observed in the epistatic studies is supported by the observation that loss of ERA1 function causes increased transcription of the ABI3::GUS reporter construct. Interestingly, the other transcription factor ABI4 that has been implicated in seed ABA signaling through genetic screens, appears to act at or upstream of ERA1.
The genetic relationships seen here are relatively simple compared to other studies involving interactions with these genes (Finkelstein, 1994; Soderman et al., 2000). This may reflect that we only followed one response output, seed ABA responsiveness, whereas other studies have also included other physiological responses and expression of multiple genes. Secondly, while this study involved interactions between mutations, often other studies involve using lines that are misexpressing or overexpressing genes of interest or constructing double mutants with similar phenotypes. Assignment of order using over- or misexpression can be deceptive because of the artificiality of the system (Ptashne, 1988). For example, squelching, which is the titration of one transcription factor by the overexpression of a partner protein, often acts in a dominant negative manner on the normal function of the transcription factor and can often cause protein interactions that do not normally occur in vivo. Construction of double mutants using mutations that confer similar phenotypes requires that each mutation is a null, a requirement that is often difficult to attain in higher plants (Avery and Wasserman, 1992).
ABI3 and auxin interactions in lateral root initiation and germination
A relationship between lateral root initiation and ABI3 has also been uncovered in studies involving misexpression of the ABI3 ortholog VP1 in Arabidopsis (Suzuki et al., 2001). That ectopic expression of VP1 suppresses auxin-induced lateral root formation argues that this gene product may define an interaction node between multiple hormone signals such as ABA and auxin. While both ABA and auxin have been implicated in controlling overall root length, it is auxin that appears to have a direct effect on lateral root initiation (Boerjan et al., 1995; Reed et al., 1998). The reduced response of abi3 loss-of-function mutants to IAA or NPA as measured by lateral root initiation indicates that the presence of ABI3 is required for correct auxin signaling in the lateral root. Moreover, the induction of ABI3 by 2,4D application positions ABI3 as both a gene that influences auxin root responses and a gene that responds to auxin. The function of ABI3 in auxin signaling is at this time unclear but the parallels of ABI3 to other auxin response genes are intriguing. The B3-binding domain of ABI3 is conserved in many auxin response (ARF) genes and ARF proteins are thought to interact with IAA gene products to elicit a correct auxin response (Tiwari et al., 2001; Ulmasov et al., 1997). Potentially, the ABI3 transcription factor could interact with other ARF or IAA proteins to mediate auxin responses.
Although ABI3 was first described as a gene involved in ABA signaling, abi3 loss-of-function mutants show additional phenotypes compared to ABA auxotrophs (Nambara et al., 1995; Ooms et al., 1993). For example, severe abi3 alleles are desiccation intolerant, fail to complete seed maturation and cannot accumulate some storage reserves, all phenotypes not detected in mutants defective in ABA biosynthesis. If abi3 mutants have reduced auxin responsiveness, this may offer an explanation for the non-ABA-related phenotypes observed in the embryo. The fact that exogenous application of IAA to wild-type Arabidopsis seeds increases ABA sensitivity as measured by germination suggests that changes in auxin synthesis or action can translate into altered seed ABA responsiveness. Furthermore, as ERA1 is a negative regulator of ABI3 transcription, the hypersensitivity observed in era1 loss-of-function alleles may simply be the result of increased production of ABI3, which could in turn increase auxin action thereby increasing seed ABA sensitivity. Indeed, the response curve of wild-type seeds to exogenous ABA in the presence of auxin is relatively similar to the increased sensitivity of era1 mutant seeds to ABA (Cutler et al., 1996). Hence, the increased lateral root initiation observed in era1 may simply reflect the increased auxin sensitivity of this mutant. On this note, other phenotypes observed in era1 mutants such as increased apical meristem size, semi sterility and defective lateral shoot initiation may also be related to altered auxin action (Bonetta et al., 2000; Running et al., 1998).
The lack of a farnesylation site on ABI3 clearly predicts that a farnesylated intermediate must exist between ERA1 and ABI3. Interestingly, genomic analysis indicates a number of potential farnesylation target genes involved in auxin responses and a number of transcription factors (Nambara and McCourt, 1999). The use of genetic enhancer and suppressor screens of the era1 hypersensitive phenotypes should uncover these intermediates. Finally, these results further suggest that the use of germination as an assay for ABA responsiveness in genetic screens enriches for mutations that alter other hormone response pathways. To date, mutations in GA, ethylene and brassinosteroid synthesis or action have also been shown to alter seed responsiveness to exogenous ABA (Ghassemian et al., 2000; Steber et al., 1998; Steber and McCourt, 2001). Thus caution should be exercised in classifying mutations into specific hormone response pathways based simply on their germination response.
Plant material and growth conditions
Germination efficiency is dependent on seed age and the conditions in which the seeds matured on the maternal plants. We find that wild-type and mutant strains give more consistent results if the seeds are younger than 6 months post-harvesting, and only seeds within this age were used in all experiments reported here. Seeds for all genotypes were harvested from plants that had all been grown under the same conditions of constant light at 22°C. Before plating, seeds were surface sterilized in 95% ethanol for 5 min and vacuum dried and then plated onto 0.5× MS minimal media plates buffered with 500 mm MES to pH 5.7. Plants were imbibed in the cold for 3 days and then germinated at 22°C in constant light. For germination studies, plants were grown for 7 days and germination was scored as positive by radical emergence and cotyledon expansion. For the GUS histochemical studies, plants were grown for 7–14 days and then stained, or transferred to the appropriate concentration of hormone for upto 3 days and then stained (see below). For the lateral root development and initiation studies, plants were grown for 14 days and development was scored according to Zhang et al. (1999). For the IAA and NPA studies, plants were grown for 3 days and then NPA or IAA was applied to the root–shoot junction according to Reed et al. (1998), following which the plants were allowed to grow for 10 days.
For the epistatic analysis, the following strains were used: abi1-1, abi2-1, abi3-6, abi4-1, abi5-1, and era1-2. The era1-2 is a null allele while abi3-6 is a severe allele that contains an internal deletion. The abi1-1 and abi2-1 alleles are in a Landsberg erecta genetic background while abi4-1 and abi5-1 alleles are in a WS background. All other lines are in a Columbia genetic background. The PABI3:GUS line is in the C24 accession (Parcy et al., 1994). For verification of many of the genotypes involving abi1-1 and abi2-1 crosses, PCR amplification with primers that identify the abi1-1 or abi2-1 mutation was used (Leung et al., 1997). Genotyping the era1-2 mutation was performed by Southern (DNA) blot analysis using the ERA1 cDNA as a probe (Cutler et al., 1996). Often, protruding carpels could also be used to identify the era1 homozygotes. The identification of abi4-1 and abi5-1 required backcrossing of potential doubles to the abi4 and abi5 parents. F1 progeny that showed reduced ABA sensitivity were scored as homozygous for the abi mutations.
Histochemical β-glucoronidase (GUS) assay and microscopy
Detection of GUS activity was carried out using 5-bromo-4-chloro-3-indolyl-β-d-glucoronide (X-Gluc) as a substrate. Tissue was placed in a X-Gluc buffer solution (100 mm NaPO4 (pH 7), 0.5 mm K3Fe3(CN)6, 0.5 mm K4Fe(CN)6, 10 mm EDTA, 0.1% Triton X-100, 20% methanol) under vacuum for 30 min and allowed to incubate overnight at 37°C. For most analyses, samples were then cleared in ethanol and viewed under a dissecting microscope. For viewing under DIC, material was mounted in a chloral hydrate:glycerol:water clearing solution (8 : 2 : 1). GUS activity measurements were performed on 40–50 6-day-old seedlings, as described, using 4-methylumbelliferyl-β-d-glucuronide (MUG) as a substrate (Jefferson et al., 1987).
We would like to thank the members of the McCourt laboratory and J. Mattson for stimulating discussions. We would also like to thank Dr R. Finkelstein for the abi4 and abi5 alleles used in this study. P.M. holds an NSERC-Industrial Chair, and the work was funded by a grant from NSERC to P.M.