All plant channels identified so far show high conservation throughout the polypeptide sequence except in the ankyrin domain which is present only in those closely related to AKT1. In this study, the architecture of the AKT1 protein has been investigated. AKT1 polypeptides expressed in the baculovirus/Sf9 cells system were found to assemble into tetramers as observed with animal Shaker‐like potassium channel subunits. The AKT1 C‐terminal intracytoplasmic region (downstream from the transmembrane domain) alone formed tetrameric structures when expressed in Sf9 cells, revealing a tetramerization process different from that of Shaker channels. Tests of subfragments from this sequence in the two‐hybrid system detected two kinds of interaction. The first, involving two identical segments (amino acids 371–516), would form a contact between subunits, probably via their putative cyclic nucleotide‐binding domains. The second interaction was found between the last 81 amino acids of the protein and a region lying between the channel hydrophobic core and the putative cyclic nucleotide‐binding domain. As the interacting regions are highly conserved in all known plant potassium channels, the structural organization of AKT1 is likely to extend to these channels. The significance of this model with respect to animal cyclic nucleotide‐gated channels is also discussed.
Plant potassium channels form a growing family of homologous proteins. Since the initial cloning of AKT1 (Sentenac et al., 1992) and KAT1 (Anderson et al., 1992) Arabidopsis thaliana cDNAs by functional complementation of yeast mutants, four new cDNAs coding for plant channels have been reported, two from Arabidopsis: AKT2/AKT3 (Cao et al., 1995; Ketchum, 1996) and KAT2 (unpublished, DDBJ/EMBL/GenBank accession number: U25695) and two from potato (Solanum tuberosum), KST1 (Müller‐Röber et al., 1995) and SKT1 (DDBJ/EMBL/GenBank accession number X86021). All of them comprise, from their N‐terminal to their C‐terminal ends: a short hydrophilic region, a hydrophobic region structurally analogous and partially homologous to the transmembrane domain of voltage‐gated animal channels from the Shaker superfamily (Jan and Jan, 1992), a putative cyclic nucleotide‐binding domain, and a conserved C‐terminal end of unknown function. Between these last two regions, some of them (AKT1, AKT2 and SKT1) contain an ankyrin‐repeat domain with six repeats homologous to those of human erythrocyte ankyrin (Lux et al., 1990). Ankyrin repeats are found in a large variety of proteins, and have been shown to be involved in protein–protein interactions (Michaely and Bennett, 1992). Recently, ankyrin repeats have been reported in a Drosophila plasma membrane calcium channel (Wes et al., 1995). The role of such motifs in channels is still unknown.
The closest known relatives of plant channels are voltage‐gated potassium channels which belong to the Shaker superfamily (Pongs, 1992) and cyclic nucleotide‐gated channels (Zufall et al., 1994). Detailed data concerning protein structure and assembly are available for Shaker‐like channels. A tetrameric organization was proposed for these potassium channels, by analogy with sodium and calcium channels, which are composed of four tandemly associated homologous domains, each similar to a Shaker cDNA product (Caterall, 1995). Experimental evidence for association between subunits was obtained by coexpression in Xenopus oocytes of cDNAs from different subfamilies of voltage‐gated potassium channels (Isacoff et al., 1990; Ruppersberg et al., 1990; Li et al., 1992). A stoichiometry of four subunits per channel protein was deduced thereafter from the properties of heteromultimeric channels formed by the association of wild type and mutant toxin‐resistant subunits (MacKinnon, 1991) and from the expression of multimeric constructs (Liman et al., 1992). Electron microscopy images of purified Shaker proteins revealed a symmetrical tetrameric complex organized around a central vestibule (Li et al., 1994).
In voltage‐gated Shaker‐type channels, a domain involved in subunit assembly, located in the N‐terminal region, was identified by deletion experiments combined with biochemical approaches (Li et al., 1992; Shen et al., 1993) or with electrophysiological studies (Li et al., 1992; Hopkins et al., 1994). This domain is capable of self‐assembly (Li et al., 1992; Shen et al., 1993; Pfaffinger and DeRubeis, 1995) and is responsible for subfamily‐specific recognition between subunits (Li et al., 1992; Shen et al., 1993; Shen and Pfaffinger, 1995; Xu et al., 1995). It also seems to be absolutely required for subunit assembly of a Kv1.1 channel (Shen et al., 1993). However, in Kv1.5 (Attali et al., 1993), Kv1.4 (Lee et al., 1994) and Kv1.3 channels (Tu et al., 1995, 1996), this domain can be removed without leading to disappearance of currents measured in Xenopus oocytes, thus confirming the earlier results of Van Dongen et al. (1990) who observed that deletions removing N‐ and C‐terminal hydrophilic parts did not prevent the formation of functional Kv2.1 channels. This correlates with results obtained in dominant‐negative experiments in oocytes, where coexpression of the tetramerization domain with wild type subunits has been observed to have a strong effect on channel activity in the case of a Kv1.1 channel (Li et al., 1992; Babila et al., 1994) but no significant effect for Kv1.3 or Kv1.5 channels (Babila et al., 1994; Tu et al., 1995). Differences in channel isoform properties could explain these discrepancies (Tu et al., 1996). In agreement with the assumption that interactions occur inside the hydrophobic core, the first transmembrane segment (S1) was shown to play an important role in Shaker channel oligomerization (Shen et al., 1993; Babila et al., 1994), and experiments performed with a mouse Kv1.3 channel revealed the participation of other segments (probably S2, S3 and S4) (Tu et al., 1996). The four H5 regions also associate in a hydrophobic environment to form the pore of the channel (Kirsch et al., 1993; Peled and Shai, 1993). It has been suggested (Babila et al., 1994) that the conserved N‐terminal sequences would drive specific recognition between subunits and their assembly in the cytoplasm, while interactions between hydrophobic regions would stabilize the complex in the membrane.
In plant K+ channels, there is no equivalent of the N‐terminal region of voltage‐gated animal channels. The similarity between AKT1 and voltage‐gated Shaker channels is restricted to the hydrophobic domain, which displays a common structural design (six transmembrane segments, a H5 sequence forming the pore between S5 and S6, and the S4 segment forming the voltage sensor), whereas homology to cyclic nucleotide‐gated channels covers a larger region from the first transmembrane segment to the end of the cyclic nucleotide‐binding domain. Although the high degree of conservation found between H5 regions of all potassium channels suggests a common arrangement of the pore, the spatial organization of plant channel subunits remained unresolved.
This paper concerns the organization of the channel AKT1. This channel is expressed in the surface tissue of roots and is probably involved in K+ uptake from the soil solution (Lagarde et al., 1996). AKT1 is functional and behaves as an inwardly rectifying voltage‐gated channel when expressed in Sf9 insect cells using a recombinant baculovirus (Gaymard et al., 1996). However, it has not been possible to detect any associated current for this channel, or its homologue AKT2 which also possesses an ankyrin domain (Cao et al., 1995), in the Xenopus oocyte expression system. In order to study AKT1 structure, we developed a strategy using two different approaches: biochemical analysis of the purified protein, and twohybrid system experiments. Unlike animal channels, AKT1 has a short hydrophilic N‐terminus (60 amino acids), but displays a very long peptide sequence (562 amino acids) downstream from the last transmembrane segment. Our results support evidence for a tetrameric structure of the protein, and show that this long C‐terminal region is by itself able to form a tetramer in vitro. Experiments using the two‐hybrid system have delineated the structural elements responsible for the interactions. A central role for the large region, located in AKT1 between the hydrophobic domain and the ankyrin repeats, was revealed. This region is conserved in all known plant potassium channels as well as animal cyclic nucleotide‐gated channels.
Production of the AKT1 protein in Sf9 cells has been described previously (Gaymard et al., 1996). AKT1 was solubilized from the microsomal fraction of Sf9 cells with lysophosphatidylcholine, and partially purified using an anion‐exchange column. In order to obtain an estimate of the molecular weight of AKT1, the purified polypeptide was loaded onto a size‐exclusion column (Sephacryl S‐300, Pharmacia). The chromatography was carried out under non‐denaturing conditions in the presence of lysophosphatidylcholine. An aliquot of each eluted fraction was tested by dot blot with an anti‐AKT1 serum, and spot intensities were quantified by densitometry. The elution profile revealed three peaks (Figure 1A). The highest one corresponded to a complex whose molecular weight was estimated to range between 385 and 425 kDa, consistent with that expected for a tetramer of identical 95 kDa subunits. To eliminate the possibility that the elution of AKT1 as a high molecular weight complex resulted from its association with other proteins, the peak fraction was analysed by SDS gel electrophoresis. After silver nitrate staining of the gel, only the 95 kDa AKT1 polypeptide was detectable (Figure 1B). The other two peaks could be ascribed to the presence of a dimer (mol. wt 173–213 kDa) and a monomer (81–121 kDa).
Assembly of the C‐terminal cytoplasmic region of AKT1 (AKT1 C‐Term)
AKT1 C‐Term is defined as the sequence (supposed to be intracytoplasmic) lying downstream from the last transmembrane segment (ranging from H294 to the end of the polypeptide). A large amount of correctly folded polypeptide was required for further analyses of its biochemical properties. For this reason, the baculovirus expression system was chosen, rather than production in Escherichia coli, because the polypeptide is denatured when expressed in bacteria (our unpublished data). The sequence coding for AKT1 C‐Term was inserted into the baculovirus genome in place of the polyhedrin gene, by cotransfection of the wild type baculovirus DNA with the plasmid containing the cDNA sequence. The recombinant baculovirus was used to infect Sf9 cells. From these cells, a specific polypeptide, which after 2 days of infection constitutes the major band detectable on a Coomassie blue‐stained SDS–polyacrylamide gel, was observed in the total soluble protein extract (Figure 2A, lanes 3 and 4). This polypeptide was not found in extracts of non‐infected cells or in wild type baculovirus infected cells (Figure 2A, lanes 1 and 2). Its apparent molecular weight was consistent with that (∼60 kDa) calculated for the peptide sequence of AKT1 C‐Term. Furthermore, the Western blot analysis performed with the anti‐AKT1 serum revealed only the 60 kDa band (Figure 2A, right).
An extract enriched in the 60 kDa polypeptide was obtained by chromatography of the Sf9 soluble protein on an ion exchange column. This extract was then fractionated by passing through a size‐exclusion column. Dot blot analysis of the different fractions revealed a major peak which was maximal in fractions 23 and 24 corresponding to molecular weights from 230 to 245 kDa (Figure 2B). A shoulder on this peak was also present at ∼120 kDa. No signal was detected for fractions corresponding to the monomer molecular weight (60 kDa). Electrophoretic analysis of the relevant peak fractions 20–27 revealed a few polypeptides. The 60 kDa AKT1 C‐Term polypeptide was detected by silver nitrate staining and Western blot (Figure 2C and D). As a single 60 kDa band was observed in peak fractions 23 and 24 (Figure 2C), the most likely explanation for the elution profile shown in Figure 2B is that four identical AKT1 C‐Term polypeptides assembled to form a 240 kDa homotetrameric complex. The 120 kDa shoulder visible on the chromatogram (Figure 2B) could correspond to a homodimeric form.
Delineation of the AKT1 domains involved in the interactions
The two‐hybrid system was used to test the interactions between the various cytoplasmic domains present in the hydrophilic C‐terminus of the protein. The sequence coding for AKT1 C‐Term was divided into three regions, called 2, 3 and 4 (Figure 3). Region 2 (amino acids H294–D517) comprises the putative cyclic nucleotide‐binding domain (estimated from sequence homologies to include approximately amino acids 393–477) and adjacent sequences. Region 3 (M516–I724) includes the ankyrin repeat domain (amino acids 522–693), and region 4 (I724 to the last amino acid, S857) is the extreme C‐terminus. Combinations of these structural elements, corresponding to regions 2–3–4 (complete C‐Term sequence) and 2–3, were also investigated. In region 2, a segment (∼100 amino acids) is present upstream from the putative cyclic nucleotide‐binding domain. Cleavage of the 2–3 region cDNA sequence using the AccI site resulted in a deletion covering most of this segment. The region (from Y371 to I724) corresponding to this truncated cDNA sequence was called δ2–3. The deleted sequence (amino acids H294–V370) was called 2A, and the rest of region 2, 2B (Figure 3). Similarly, using the SspI site in the sequence coding for the 2–3–4 region, another deletion was generated which led to the elimination of the sequence upstream from I473 (including the putative cyclic nucleotide‐binding domain). The corresponding region was named Δ2–3–4. Finally, the NheI site present in the DNA sequence of region 4 was used to create region Δ4, comprising the last 81 amino acids (Figure 3.)
In‐frame fusions were made between cDNA sequences coding for GAL4 domains (DNA‐binding or activator domain) and the cDNA fragments corresponding to the different regions described above. The vectors used, pGBT9 and pACTII, enable the synthesis of fusion proteins with either the DNA‐binding domain or the activator domain of GAL4, respectively. Plasmids resulting from insertion of AKT1 sequences downstream from that coding for the GAL4 domain were designated by the name of the vector followed by the region encoded by the cDNA fragment [e.g. pGBT9(2–3–4): pGBT9 containing the 2–3–4 coding sequence]. The recipient yeast strain was cotransformed with different combinations of plasmids as shown in Figures 4 and 5. The β‐galactosidase activities were systematically detected by colorimetric tests on solid medium (example presented in Figure 4), and quantitative measurements were carried out after protein extraction. Control experiments were performed using empty plasmids in place of each one of the two plasmids. Confirmations of observed positive and negative reactions were obtained by testing the reciprocal combinations of plasmids (obtained by cloning the insert of plasmid pGBT9 in pACTII and reciprocally that of pACTII in pGBT9). In the case of positive results, such a reciprocity is considered as a strong argument for the existence of a true physical association (Allen et al., 1995).
Results obtained with the complete 2–3–4 region are presented in Figure 4. Control experiments using pGBT9(2–3–4) and pACTII, or pGBT9 and pACTII(2–3–4) did not reveal any signal above the background level (0.8 unit of β‐galactosidase activity) obtained with the combination pGBT9/pACTII. In contrast, yeast transformed with both plasmids pGBT9(2–3–4) and pACTII(2–3–4) turned blue rapidly (2 h) on X‐Gal plates and displayed a high level of β‐galactosidase activity. This result confirms that interactions occur between the AKT1 C‐Term regions.
In order to identify which domains in the 2–3–4 region are able to bind to the complete 2–3–4 polypeptide sequence, regions 2, 3 and 4 were tested independently against 2–3–4. Regions 2 and 4 were found to be involved in intramolecular interactions (Figure 5). As expected from these results, regions 2–3 and Δ2–3–4 also interacted with region 2–3–4. No interaction could be detected with region 3.
In addition, each single region was tested against each of the others. Two identical region 2s exhibited a positive reaction in this assay. The β‐galactosidase activity was ∼5‐fold lower than that observed with the complete 2–3–4 region, but clearly above the background level (Figure 5). Another interaction was detected between regions 2 and 4. The signals were significantly higher than those of the controls, and the two plasmid combinations gave similar levels of β‐galactosidase activity (Figure 5). In agreement with these observations, the following couples of AKT1 fragments gave positive results: 2–3 and 2–3, 2–3 and 2, 2–3 and 4. In contrast, the negative results obtained with 4 and 4, or Δ2–3–4 and Δ2–3–4 suggest an absence of interaction between homologous 4 regions (Figure 5).
Deletions within the regions 2–3 and 4 leading to δ2–3 and Δ4 enabled the sequences that are more specifically involved in the interactions to be localized. As shown in Figure 5, the deletion in region 4 did not prevent association with region 2 (positive results with tests combining Δ4 and 2), showing that the extreme end of the protein is by itself able to bind to region 2. The region δ2–3 was still able to react with another δ2–3 but not with region 4; the 2A region is therefore required for association with the C‐terminal end, but it is dispensable for the interaction between regions δ2–3. This interaction between two δ2–3 regions is lost when further 5′ sequence is deleted so that the cyclic nucleotide‐binding domain is also removed, as a negative result is obtained with two Δ2–3–4 regions (Figure 5).
Comparison of AKT1 channel structure with that of animal potassium channels
Two independent methods have been used to establish that the AKT1 gene product is capable of self‐association: the estimation of the native protein and C‐Term molecular weights, and the determination of the binding properties of protein regions. Furthermore, from the molecular weights of the polypeptide complexes, it can be deduced that AKT1 shares the tetrameric organization common to the voltage‐gated animal potassium channels of the Shaker family (Pongs, 1992; Caterall, 1995). This is the first demonstration of such a structure in a plant channel.
Although the overall structure of AKT1 is similar to that of voltage‐gated animal channels, the mode of subunit assembly appears to be different, with a major role for the C‐terminal domains in the case of AKT1. In channels of the Shaker family, the C‐terminal cytoplasmic tail has been shown not to be involved in tetramerization (Li et al., 1992; Hopkins et al., 1994). AKT1 C‐Term tetramers appear to be very stable, as the monomeric form is undetectable in gel filtration column eluates (Figure 2B). Thus, interactions involving the C‐terminal hydrophilic part are likely to be important for AKT1 tetramerization. These results show that animal and plant potassium channels have developed independent ways to achieve their correct subunit assembly resulting in a tetrameric organization which is compatible with gating inside the channel pore.
Using the two‐hybrid system, we have defined two kinds of interactions: one between identical regions 2 (amino acids 294–516), and another one between region 2 and region Δ4 (amino acids L777 to the last one, S857). All the results obtained are consistent with the existence of these two interactions (Figure 5). They are reproducible and always reciprocal (i.e. the occurrence of a positive or negative result does not depend on the Gal4 domain with which the AKT1 fragments are fused, as demonstrated by the symmetry of measurements with respect to the diagonal of the table in Figure 5). In no combination did the ankyrin‐repeat domain (included in region 3) show any interaction, either with itself or with other regions of AKT1 C‐Term. Interaction with itself was not expected since ankyrin domains are known to interact with sequences which are not related to them (Hoffman, 1991). The negative results obtained with other parts of the molecule could be due to incorrect folding of the ankyrin domain when isolated from its normal polypeptide environment. However, the structure formed by the six adjacent ankyrin repeats would be expected to be globular and stable, from the data obtained on human erythrocyte ankyrin (Michaely and Bennett, 1993). Furthermore, negative results were obtained when region 3 was included in large parts of the molecule (see Δ2–3–4 with Δ2–3–4, and Δ2–3–4 with δ2–3 in Figure 5). Thus, the higher level of β‐galactosidase activity that resulted in some cases when ankyrin domain was present (e.g. in tests with region 2–3 compared with the same tests performed with region 2) is likely to arise from an increased stability of the fusion protein and/or a better folding of the interacting domain, rather than from interactions involving ankyrin domains.
The two‐hybrid experiments performed with region 2 revealed different binding properties between its N‐ and C‐terminal segments. The first one (2A), was found to be involved in the interaction with the C‐terminal end (Δ4) of the protein, whereas the other (2B, including the putative cyclic nucleotide‐binding domain), appeared to be mainly responsible for the observed self‐association of regions 2–3 (and consequently of region 2), since deletions of domain 2A or region 3 from region 2–3 (generating regions δ2–3 and 2, respectively) did not severely affect the interaction. Interestingly, in the AKT1 gene (Basset et al., 1995), two introns flank a segment corresponding to amino acids R323–L384. This provides further support for the hypothesis that the inserted sequence, corresponding for the junction between the transmembrane domain and the putative cyclic nucleotide‐binding domain, codes for a real domain.
Region 2B, which is likely to be involved in subunit tetramerization, is homologous to the cAMP‐binding domain of cAMP‐dependent protein kinases (Kaupp et al., 1989) and to the E.coli cAMP receptor protein (CRP) (Weber and Steitz, 1984). All these proteins are dimers. According to crystallography data, intersubunit contacts in the CRP protein are mainly found at the level of the C helix located downstream from the cAMP‐binding domain (Weber et al., 1987). However, a proteolytic fragment which includes the N‐terminal (cAMP‐binding) domain, but lacks 85% of the C helix, is still able to form stable dimers (Heyduk et al., 1992). Studies concerning a cAMP‐dependent protein kinase have shown that an isolated cAMP‐binding domain expressed in E.coli is able to form dimers, and that the dimer displays a much higher nucleotide exchange rate than the monomer (Shabb et al., 1995). Similarly, interactions between cyclic nucleotide‐binding domains could be involved in associations between AKT1 subunits.
Concerning the interactions involving domain 2A and region 4, the two‐hybrid system does not predict whether they occur within a single polypeptide chain or between two different subunits. However, if they can occur between regions 2 and 4 of the same polypeptide, one would expect in a test involving regions 2–3–4 and 4, a competition between the two region 4s for binding to region 2. This does not appear to be the case, since the complete 2–3–4 sequence gives an even higher response than 2–3 when tested against region 4 in two‐hybrid assays (Figure 5). We tend therefore to favour the hypothesis of inter‐chain associations, attributing this preference to physical constraints preventing intra‐chain folding. This interpretation is based on the assumption that the hybrid proteins containing region 4 alone are not favoured compared with those containing the complete 2–3–4 sequence, due to e.g. a higher stability.
According to our results, the interaction of the C‐terminal ends of each subunit of the functional channel with the proximal regions of the C‐Term regions would lead to the formation of four loops. This would result in an external localization of the four ankyrin domains that would favour interactions between these domains and as yet unidentified proteins. It is worth noting that in erythrocyte ankyrin, the 24 ankyrin repeats are thought to be folded in four globular domains of six ankyrin repeats each (Michaely and Bennett, 1993). The four six‐repeat ankyrin domains of the AKT1 channel could form a similar structure.
Extension of the AKT1 structural model to other plant channels and to cyclic nucleotide‐gated channels
All plant channels identified so far share high sequence homologies in the domains implicated, in this study, in intramolecular interactions: the cyclic nucleotide‐binding domain and its junction with the hydrophobic domain (Cao et al., 1995; Müller‐Röber et al., 1995), and region Δ4 (Figure 6). In AKT1 C‐Term, these are the only regions which show sequence conservation between all plant channels of both the AKT1 and KAT1 types (presence and absence of an ankyrin domain, respectively); no homology can be found within sequences located between regions 2B and Δ4. We suggest therefore that the conserved sequences in regions 2 and Δ4 will play similar roles in all plant channels, and that the interactions responsible for channel tetramerization and protein folding involve common mechanisms. It is interesting that the ankyrin domain, which is specific to AKT1 family channels, was not found to be involved in interactions within AKT1.
Like potassium channels, cyclic nucleotide‐gated cation channels are thought to be multimeric (Eismann et al., 1993), probably tetrameric (Gordon and Zagotta, 1995; Liu et al., 1996). They can function either as homomultimers of α subunits or as heteromultimers combining homologous α and β subunits (Chen et al., 1993; Korschen et al., 1995). Apart from homologies in the hydrophobic domain, all known α and β subunits share 22 to 25% identity [Fasta program (Pearson and Lipman, 1988), data not shown] with AKT1 throughout the region extending from T336 to V445 in AKT1 (end of region 2A + putative cyclic nucleotide‐binding domain). Thus, subunit assembly in cyclic nucleotide‐gated channels might be mediated, at least in part, by sequences homologous to this AKT1 segment. While looking for equivalents of region Δ4 in cyclic nucleotide‐gated channels, we found a weak homology (38% identity in a 21 amino acid overlap) with the C‐terminal end of some β subunits sequences (Figure 6).
In conclusion, we propose a global structural organization, very different from that of animal voltage‐gated potassium channels, which would be common to all known plant potassium channels. To define the exact contribution of C‐terminal ends in the tetramer more precisely, a detailed structural analysis of the crystallized protein will probably be required. Further investigations will also be needed to test whether the plant channel model can be extended to cyclic nucleotide‐gated channels. However, the similarity of primary structure organization suggests that these channels could adopt a mode of assembly partly analogous to that of AKT1, with the participation of the C‐terminal intracytoplasmic tail, specifically in the region including the cyclic nucleotide‐binding domain.
Materials and methods
Expression of cDNA products in insect cells
Sf9 and Sf21 cell lines were propagated as described by Gaymard et al. (1996). The two recombinant transfer vectors, harbouring AKT1 cDNA or the cDNA sequence coding for AKT1 C‐Term, were derived from pGmAc34T (Davrinche et al., 1993). The insertion of AKT1 cDNA into pGmAc34T has been described previously (Gaymard et al., 1996). The sequence coding for AKT1 C‐Term was introduced into pGmAc34T–SmaI. This vector was created from pGmAc34T by insertion of a SmaI site at position +45 (from the ATT triplet replacing the ATG initiation codon of the polyhedrin gene). The partial AKT1 cDNA sequence was obtained from plasmid pUC9–AKT1, containing the NotI AKT1 fragment from pHS41 (Sentenac et al., 1992) inserted at a NotI site previously introduced into the pUC9 polylinker. This plasmid was digested at the NcoI site present at the end of the S6 segment, blunt‐ended with the Klenow fragment of E.coli DNA polymerase I, and religated. This created an ATG codon and an NsiI site. An NsiI–SmaI fragment encoding AKT1 C‐Term was isolated by partial NsiI digestion and use of the SmaI site of the pUC9 polylinker. This fragment was cloned into pJRD184 (Heusterspreute et al., 1985) using the NsiI and SmaI sites of this vector. SmaI sites were introduced at both ends of the cDNA sequence by subcloning firstly into pSP72 (Promega), using BglII and SmaI sites, and then into pBluescript SKII (Stratagene), using BglII and XhoI sites. The SmaI fragment containing the AKT1 partial cDNA sequence was then introduced into pGmAc34T–SmaI.
Sf21 cells were transfected with wild‐type viral DNA and the recombinant transfer vectors (Davrinche et al., 1993). Recombinant baculoviruses were purified, and amplified in Sf9 cells to 108 p.f.u./ml for protein production.
Preparation of protein extracts
Sf9 cell extracts were prepared according to Gaymard et al. (1996). AKT1 was solubilized from the microsomal fraction (100 000 g pellet) with lysophosphatidylcholine [ratio detergent/protein of 5 (w/w), in grinding medium minus glycerol and leupeptin]. AKT1 C‐Term was extracted following the same protocol as for AKT1, except that after 1 h centrifugation at 100 000 g it was obtained in the soluble protein fraction. Both polypeptides were partially purified by anion exchange chromatography (Mono‐Q, Pharmacia). The elution buffer contained 10% glycerol, 1 mM EDTA and either 20 mM histidine, pH 8 plus 0.05% lysophosphatidylcholine (for AKT1) or 20 mM piperazine, pH 9 (for AKT1 C‐Term). Proteins were eluted using a NaCl step gradient from 0 to 1 mol/l, with steps of 0.1 mol/l.
A 250 ml Sephacryl S‐300 column (Pharmacia) was equilibrated (200 μl/min) with either 20 mM bis–tris propane–HCl pH 7 buffer, containing 100 mM NaCl, 1 mM EDTA, 10% glycerol and 0.05% lysophosphatidylcholine for AKT1, or 20 mM piperazine pH 9 buffer, containing 100 mM NaCl, 1 mM EDTA and 10% glycerol for AKT1 C‐Term. The elution was carried out in the same buffer, after loading 100 μg of protein. Columns were calibrated with molecular weight standards from Sigma (MW‐GF‐1000): carbonic anhydrase (29 kDa), bovine serum albumin (66 kDa), alcohol dehydrogenase (150 kDa), β‐amylase (200 kDa), apoferritin (443 kDa) and thyroglobulin (669 kDa).
Polyclonal antibodies were raised against the C‐terminal region of AKT1, as described by Gaymard et al. (1996). Elution fractions from the size‐exclusion columns were analysed by dot blots, a 1 μl aliquot of each fraction being deposited onto a nitrocellulose membrane (Sartorius). For Western blot analyses, proteins were separated by SDS–PAGE according to Laemmli (1970) and electroblotted to a nitrocellulose membrane (Sartorius) at 100 V for 1 h in a medium containing 25 mM Tris, 192 mM glycine, 0.1% SDS and 20% methanol.
Membranes were processed as described by Gaymard et al. (1996). AKT1 or AKT1 C‐Term were detected by chemiluminescence (Aurora kit, ICN). For the dot blot quantifications, intensities of the spots were measured by densitometry using ‘The Imager™’ (Appligene) and the NIH‐Imager program.
Plasmid vectors pGBT9 (Bartel and Fields, 1995) and pACTII [derived from pACT (Durfee et al., 1993) and kindly provided by S.Elledge, Baylor college of Medicine, Houston, TX] were used for the generation of fusion proteins with the DNA‐binding domain and the activator domain of GAL4, respectively.
In‐frame fusions were made between the DNA‐binding or activator domain of GAL4 and fragments from AKT1 cDNA in pHS41 (Sentenac et al., 1992). Cloning procedures involved DNA digestions at appropriate restriction sites (see Figure 3), end modifications using the Klenow fragment of DNA polymerase I, and intermediate subclonings into pBluescript SKII (Stratagene). All constructions were checked by DNA sequencing of the vector–insert junction region.
All constructions were introduced into strain Y526 (Mata, ura3‐52, his3‐200, ade2‐101, lys2‐801, trp1‐901, leu2‐3, 112, canr, gal4‐542, gal80‐538, URA3::Gal1–lacZ) (Bartel and Fields, 1995). Yeast were routinely grown on YPD medium (Sherman, 1991). For transformation, an exponential phase culture (0.6 absorbance unit) was pelleted by centrifugation, and resuspended in 1/20 culture volume of YPD medium supplemented with DTT (25 mM) and HEPES (20 mM, pH 8). After a 30 min incubation at 30°C under constant agitation, cells were washed twice with EB buffer (10 mM Tris, 270 mM sucrose, 1 mM MgCl2, pH 8) and resuspended in 1/50 culture volume of EB buffer. Fifty microlitre aliquots of the yeast suspension were mixed with 1 μg of each plasmid, and deposited between two plate electrodes (type D03N, Jouan, Saint Herblain, France). Plasmids were introduced into yeast cells by electroporation with a Jouan GHT 1287B apparatus, by a 12 ms pulse at 725 V. After transformation, yeast cells were plated onto a selective medium lacking leucine and tryptophan, composed of SD minimal medium (Sherman, 1991), complemented with isoleucine (30 mg/l), valine (150 mg/l), adenine (20 mg/l), arginine (20 mg/l), histidine (20 mg/l), lysine (30 mg/l), methionine (20 mg/l), phenylalanine (50 mg/l), threonine (200 mg/l), tyrosine (30 mg/l) and uracil (20 mg/l).
Assay for reporter gene expression
Transformed yeast were replicated onto selective medium containing raffinose (2%) in place of glucose, incubated for 3 days at 30°C and tested for β‐galactosidase activity directly on the agarose plate, as described by Werner et al. (1993). Quantitative tests were performed after protein extraction, according to Bartel and Fields (1995), using o‐nitrophenyl‐β‐d‐galactopyranoside (ONPG) as substrate.
We are grateful to Helen Logan for critical reading of the manuscript. Our work is partly supported by the Bio Avenir programme (funded by Rhône‐Poulenc and the French Ministry in charge of Research and Industry) and by the European Communities' BIOTECH Programme (Project of Technological Priority 1993–1996).
- Copyright © 1997 European Molecular Biology Organization