Human GKβ (pancreatic β-cell glucokinase) is the main glucose-phosphorylating enzyme in pancreatic β-cells. It shares several structural, catalytic and regulatory properties with Hxk2 (hexokinase 2) from Saccharomyces cerevisiae. In fact, it has been previously described that expression of GKβ in yeast could replace Hxk2 in the glucose signalling pathway of S. cerevisiae. In the present study we report that GKβ exerts its regulatory role by association with the yeast transcriptional repressor Mig1 (multicopy inhibitor of GAL gene expression 1); the presence of Mig1 allows GKβ to bind to the SUC2 (sucrose fermentation 2) promoter, helping in this way in the maintenance of the repression of the SUC2 gene under high-glucose conditions. Since a similar mechanism has been described for the yeast Hxk2, the findings of the present study suggest that the function of the regulatory domain present in these two proteins has been conserved throughout evolution. In addition, we report that GKβ is enriched in the yeast nucleus of high-glucose growing cells, whereas it shows a mitochondrial localization upon removal of the sugar. However, GKβ does not exit the nucleus in the absence of Mig1, suggesting that Mig1 regulates the nuclear exit of GKβ under low-glucose conditions. We also report that binding of GKβ to Mig1 allows the latter protein to be located at the mitochondrial network under low-glucose conditions.
- glucose repression
- hexokinase 2 (Hxk2)
- multicopy inhibitor of GAL gene expression (Mig1)
- pancreatic β-cell glucokinase
- Saccharomyces cerevisiae
GK (glucokinase), a member of the Hxk (hexokinase) family (Hxk IV) , plays an essential role in glucose phosphorylation in the liver and endocrine β-cells of the pancreas . GK also acts as a glucose sensor, integrating blood glucose levels and insulin release in pancreatic β-cells  and regulating glucose metabolism in the liver . Hepatic GK shows dynamic nuclear–cytoplasmic localization depending on the amount of glucose in the blood. The nuclear localization of hepatic GK is regulated by its binding to a GKRP (GK regulatory protein) under low-glucose conditions [4,5]. In addition, it has been previously described that hepatic GK could also be associated with liver mitochondria [6,7]. The GK subcellular distribution in the pancreatic β-cell is also regulated by glucose: low levels of glucose cause an association of GK with mitochondria , whereas high glucose levels prevents GK translocation to this organelle. In β-cells, GK has also been localized on the secretory granules and, although it has been suggested that glucose causes dissociation of GK from this location , later studies suggest that GK is an integral component of the granule and does not translocate during glucose stimulation .
GKβ (pancreatic β-cell GK) and the yeast Hxk2 proteins have several structural and functional similarities. For example: (i) GKβ is a monomer without allosteric regulation by glucose 6-phosphate; (ii) yeast Hxk2, like GKβ, has a glucose-regulated subcellular distribution [11,12]; in higher glucose conditions Hxk2 shows a nuclear enrichment that is absent under low-glucose conditions; and (iii) both Hxk2 and GKβ proteins play a vital role in glucose signalling in the yeast and the pancreatic β-cell respectively. Current evidence suggests that the main role of Saccharomyces cerevisiae Hxk2 in the glucose signalling pathway is achieved by its interacting with both Mig1 (multicopy inhibitor of GAL gene expression 1) and Snf1 (sucrose non-fermenting 1). It has been proposed that Hxk2 inhibits the phosphorylation of the Mig1 repressor, when the cells are growing in high-glucose conditions, maintaining in this way the transcriptional repression of target genes [13,14]. Since GKβ plays a prominent role in pancreatic β-cell signalling (inactivation of one GK allele leads to maturity-onset diabetes of the young, whereas loss of both alleles is associated with permanent neonatal diabetes [15,16]), GKβ is considered as a promising drug target for diabetes therapy. Thus the study of the regulatory properties of GKβ is an important matter to be examined.
In the present study, we took advantage of the ability to manipulate the S. cerevisiae genetic system and the similarity between the yeast Hxk2 and pancreatic GKβ to study the role of GKβ in the yeast glucose signalling pathway and also to study the regulation of the different subcellular localization that this enzyme presents in yeast cells.
MATERIALS AND METHODS
Strains and growth medium
GFP (green fluorescent protein) and RFP (red fluorescent protein)-fluorescence experiments utilized yeast strain H250 (MATα SUC2 ade2-1 can1-100 his3-11,15 leu2-3,112 trp1-1 ura3-1 mig1-δ2::LEU2) , W303-1A (MATa ura3-52 trp1-289 leu2-3,112 his3-Δ1 ade2-1 can1-100) , DBY1315 (MATα ura3-52 leu2-3,2-112 lys2-801 gal2) , DBY2052 (MATα hxk1::LEU2 hxk2-202 ura3-52 leu2-3,2-112 lys2-801 gal2) , Δsnf1 , Δsnf1 Δhxk1 Δhxk2 , Y03694 (Matα his3Δ1 leu2Δ0 met15Δ0 ura3Δ0 msn5::kanMX4) (euroscarf) and FMY388 (MATα his3Δ1 ura3Δ0 leu2Δ0 met15Δ0 MIG1::gfp), the last containing a GFP-tagged MIG1 ORF (open reading frame) at its chromosomal location. Yeast two-hybrid experiments employed strain Y187 (MATα ura3-52 his3-200 ade2-101 trp1-901 leu2-3,112 gal4Δ gal80Δ URA3::GAL1UAS-GAL1TATA-lacZ) .
Escherichia coli DH5α [F Ø80dlacZ ΔM15 recA1 endA1 gyrA96 thi-1 hsdR17(rk-rk-) supE44 relA1 deoRΔ (lacZ YA-argF)U169] was the host bacterial strain for the recombinant plasmid constructions.
Yeast cells were grown in the following media: YEPD, high-glucose (2% glucose, 2% peptone and 1% yeast extract); YEPE, low-glucose (0.05% glucose, 3% ethanol, 2% peptone and 1% yeast extract); and synthetic medium containing the appropriate carbon source and lacking appropriate supplements to maintain selection for plasmids [2% glucose (SD), or 3% ethanol and 0.05% glucose, and 0.67% yeast nitrogen base without amino acids]. Amino acids and other growth requirements were added at a final concentration of 20–150 μg/ml. The solid medium contained 2% agar in addition to the components described above.
Plasmids YEp352-GKβ and pWS93-GKβ carried a 1398 bp DNA fragment with the complete coding region of the human GKβ gene. Plasmid YEp352-GKβ was generated by cloning a 1398 bp DNA fragment, synthesized by PCR using the plasmid pWS-GST/GKβ  as the template and the primer pair OL1 (sense: 5′-CAGATAGGATCCATGCTGGACGACAGAGCCAGG-3′) and OL2 (antisense: 5′-CTTCAGAATTCCTATCACTGGCCCAGC- ATACAGGC-3′) into the BamHI-EcoRI site of YEp352-HXK2. In all sequences, the underlining corresponds to the restriction endonuclease sites. The YEp352-GKβ plasmid expresses the complete coding region of the GKβ gene from the HXK2 promoter. To make plasmid pWS93-GKβ, an EcoRI-SalI fragment obtained from pWS-GST/GKβ was subcloned into the EcoRI-SalI site of the pWS93 vector.
Plasmid YEp352-GKβ/gfp and YEp352-GKL/gfp were constructed as follows: a 969 bp BamHI-BglII fragment containing the gfp gene was subcloned into the YEp352-HXK2 plasmid first cleaved with BamHI. Then, a 1398 bp BamHI PCR fragment (OL3: 5′-CGTAGGATCCATGGCTATGGATACTACAAG-3′ and OL4: 5′-GCTAGGATCCCGCCT GGGCCAGCATGCAAGCC-3′) containing the complete coding region of the GKβ gene or a 1398 bp BamHI PCR fragment (OL5: 5′-CGTAGGATCCATGGCTATGGATACTACAAG-3′ and OL6: 5′-GCTAGGATCCCGCCTGGGCCAGCATGCAAGCC-3′) containing the complete coding region of the GKL (liver GK) gene was subcloned into a previously BamHI-digested YEp352-HXK2/gfp plasmid. The resulting plasmids expressed GKβ or GKL from the HXK2 promoter as fusion proteins with GFP.
For two-hybrid analysis, plasmids pGBKT7-MIG1, pGBKT7-MIG1S311A and pGBKT7-HXK2 were constructed as indicated previously [13,14,23]. Plasmids pGBKT7-GKβ and pGADT7-GKβ carried a 1398 bp DNA fragment with the complete coding region of the GKβ gene. To make plasmid pGBKT7-GKβ an EcoRI-SalI fragment obtained from pWS93-GKβ carrying the complete GKβ gene was subcloned into the EcoRI-SalI sites of pGBKT7. To make plasmid pGADT7-GKβ, an EcoRI-SalI fragment containing the complete coding region of the GKβ gene obtained from pWS93-GKβ plasmid was subcloned into a previously EcoRI-XhoI cleaved pGADT7 vector. The DNA sequence of all PCR-generated constructs was verified by sequencing and all of the clones used were verified by sequencing analysis of fusion points.
Yeast strains expressing the GKβ–GFP, GKL–GFP, Hxk2–GFP, Mig1–GFP or Su9 (F0F1 ATPase subunit 9)–RFP fusion proteins were grown to early-exponential phase (D600 of less than 0.8) in synthetic high-glucose medium (SD-ura). Half of the culture was shifted to synthetic low-glucose medium (SE-ura) for 1 h. The medium contained the appropriate carbon source and lacked the appropriate supplements to maintain the selection of plasmids. Cells (25 μl) were loaded on to poly-L-lysine-coated slides, and the remaining suspension was immediately withdrawn by aspiration. A 1 μl aliquot of DAPI (4′,6-diamidino-2-phenylindole; 2.5 μg/ml in 80% glycerol) was added, and a cover slide was placed over the microscope slide. GFP, RFP and DAPI localization in live cultures was monitored by direct fluorescence using a Leica DM5000B microscope. To avoid the non-linear range of fluorescent signals, cells highly overexpressing GFP-tagged fusion protein were excluded from further analyses. The localization of proteins was monitored by visual inspection of the images. At least 100 cells were scored in each of at least three independent experiments. The distribution of fluorescence was scored in the following way: N indicated a nuclear fluorescence signal; C indicated a cytoplasmic fluorescence signal without nuclear or mitochondrial fluorescence signals; and M indicated a mitochondrial fluorescence signal.
To stain cells with Mitotracker® Red 580, yeast were immobilized on poly-L-lysine-coated slides. Then they were incubated with 0.5 μM MitoTracker® diluted in DMSO and glycerol 80% (v/v) for 20 min at room temperature (22 °C). Finally, MitoTracker® solution was aspired and a glass cover slide was immediately placed over the microscope slide. Images were processed in Adobe Photoshop CS.
Yeast two-hybrid analysis
The yeast two-hybrid analysis  employed yeast vectors pGADT7 and pGBDKT7 and host strain Y187 (described above), in accordance with the Matchmaker two-hybrid system 3 from Clontech. Transformed yeasts were grown in high-glucose (SD/−Leu, Trp) medium. Assays for β-galactosidase activity followed protocols as described previously . Expression levels of the GAD (Gal4-activating domain) and GBD (Gal4-binding domain) fusion proteins were controlled by Western blot analysis. Experiments were performed a minimum of three times.
ChIP (chromatin immunoprecipitation) assay
ChIP assays were performed essentially as described previously . Cells were harvested and disrupted by vortex mixing in the presence of glass beads, and the lysate was sonicated to generate DNA fragments that ranged in size from 200 to 400 bp. To immunoprecipitate HA (haemagglutinin)-tagged proteins, we incubated the extract overnight at 4 °C with anti-HA antibodies (Santa Cruz Biotechnology). To immunoprecipitate Hxk2 protein, we incubated the extract overnight at 4 °C with anti-Hxk2 antibodies . The sequence primers for PCR to amplify the SUC2 (sucrose fermentation 2) promoter region containing the MIG1 element were 5′-TTATTACTCTGAACAGGA-3′ (sense) and 5′-AAGTCGTCAAATCTTTCT-3′ (antisense).
Subcellular localization of human GKβ in yeast cells
It has been previously described that human GKβ complements the glucose signalling defect of S. cerevisiae Δhxk2 mutants, indicating that GKβ may substitute for Hxk2 in its role in regulating glucose signalling in yeast . To gain insight into the molecular mechanism by which GKβ regulates this process in yeast, the glucose-dependent subcellular location of GKβ was determined in S. cerevisiae cells.
To detect GKβ in live yeast cells, a GKβ–GFP protein fusion was expressed under the control of the HXK2 gene promoter, and its subcellular location was determined by fluorescence microscopy. The GKβ–GFP fusion protein maintained repression functions in the glucose signalling pathway (results not shown). In cells grown overnight in high-glucose conditions, a fraction of GKβ–GFP was enriched in the nucleus (Figures 1A and 1C). Similar results were obtained with an Hxk2–GFP protein fusion, as previously described . However, in cells grown overnight in low-glucose medium, GKβ–GFP was found mainly associated to multiple organelles distributed through the cytoplasm, reminiscent of a typical mitochondrial network. In contrast, the Hxk2–GFP protein presented an even distribution between the nucleus and the cytoplasm, as previously described  (Figure 1A). In order to determine whether the GKβ–GFP protein presented a mitochondrial localization under low-glucose conditions, we labelled the mitochondrial network with a Mitotracker probe and observed a similar pattern of subcellular distribution of GKβ–GFP and the Mitotracker (Figure 1B), suggesting a mitochondrial distribution of GKβ–GFP under these conditions. To confirm the mitochondrial localization of the GKβ–GFP protein under low-glucose conditions, an RFP fused with the mitochondrial targeting signal of Neurospora Su9  was co-expressed with the GKβ–GFP fusion protein in wild-type S. cerevisiae cells. As can be seen in Figure 1(B), the expressed proteins were associated with the mitochondria and the fluorescence signals completely overlapped.
Since liver and β-cell GK isoforms differ only in the first 15 N-terminal residues, it was important to determine whether the observed location of GKβ was also shared by the liver isoform. As can be seen in Figures 1(A) and 1(C), in cells grown overnight in high-glucose conditions, a fraction of GKL–GFP was enriched in the nucleus, and in cells grown overnight on low-glucose medium GKL–GFP was found mainly associated to mitochondria (Figures 1A–1C). Therefore our results suggest that, in yeast, the subcellular localization of β-cell and liver GK isoforms was similar and did not depend on their N-terminus.
The nuclear export of human GKβ in the yeast cell is Mig1-dependent
We have recently described that, under low-glucose conditions, Hxk2 exits the nucleus using a pathway in which Mig1 is involved [13,14]. In order to determine whether Mig1 is also affecting the exit from the nucleus of GKβ–GFP under the same conditions, we determined the subcellular distribution of GKβ–GFP in Δmig1 mutant cells (Figure 2A). The results indicated that GKβ–GFP fusion protein was enriched in the nucleus in both high- and low-glucose-grown cells, suggesting that in the absence of Mig1, GKβ–GFP was not able to exit the nucleus and reach the mitochondria. The introduction in these cells of a plasmid expressing MIG1 under its own promoter (YEp351-MIG1) allowed GKβ–GFP to exit the nucleus under low-glucose conditions (Figure 2A), as in wild-type cells (Figure 1A), suggesting that in low-glucose conditions, Mig1 is necessary to export the GKβ–GFP fusion from the nucleus to the mitochondria.
We also studied the subcellular distribution of GKβ–GFP in a Δhxk1 Δhxk2 double-mutant strain (Figure 2B). The results demonstrated that, in the absence of Hxk1 and Hxk2, GKβ–GFP was targeted to the mitochondria at both high- and low-glucose concentrations. Since, in the absence of Hxk1 and Hxk2, the protein kinase Snf1 is constitutively activated, we repeated the experiment in a triple Δhxk1 Δhxk2 Δsnf1 mutant (Figure 2C). In this case, no mitochondrial distribution of GKβ–GFP was observed under any growth conditions (similar results were obtained with a single Δsnf1 mutant; Figure 2C). These results suggested that Snf1 played a major role in regulating the exit of GKβ–GFP from the nucleus. When Snf1 was activated, either by growing the cells under low-glucose conditions or by deleting negative regulators such as Hxk1 and Hxk2, GKβ–GFP exited the nucleus and presented a mitochondrial distribution.
It is known that Snf1 is involved in the phosphorylation of Mig1 under low-glucose conditions, leading to its exit from the nucleus to the cytoplasm . Since GKβ–GFP followed the same pattern of distribution as Mig1 and, as we have described above, GKβ–GFP did not exit the nucleus in the absence of Mig1, even under conditions where Snf1 is activated (low-glucose), we suggest that Mig1 could be involved in the exit of GKβ–GFP from the nucleus under low-glucose conditions. In agreement with this suggestion, we observed that GKβ–GFP fusion protein was enriched in the nucleus, in both high- and low-glucose-grown cells, in the absence of Msn5 (multicopy suppressor of SNF1 mutation 5) (Figure 3A), a member of the importin family of nuclear transport proteins which is required to export Mig1 from the nucleus in low-glucose conditions . This result suggests that, in the absence of Msn5, GKβ–GFP was not able to exit the nucleus and reach the mitochondria. Thus, taken together, these results demonstrated that Mig1 and GKβ–GFP could form part of a nuclear complex whose export to the cytoplasm was dependent on both the phosphorylation of Mig1 by Snf1 and the exportin Msn5. The close relationship between Mig1 and GKβ–GFP was confirmed when we observed that a Mig1–GFP protein fusion was dragged to the mitochondrial network under low-glucose conditions only if GKβ was expressed in the same cells (Figure 3B). These results suggested that Mig1 and GKβ may form a complex in which Mig1 confers the determinants for nuclear export and GKβ confers the determinants for mitochondrial localization under low-glucose conditions.
Human GKβ interacts with the yeast Mig1 transcriptional repressor
The experiments described above strongly suggest that Mig1 and GKβ may form part of an in vivo complex. To confirm the possible physical interaction between Mig1 and GKβ, we have used a yeast two-hybrid assay. Plasmid pGADT7-GKβ (expressing a fusion of the GAD to GKβ) was co-transformed with a plasmid expressing a fusion of the GBD with Hxk2, Mig1, Mig1S311A or GKβ, into an appropriate reporter strain. The interaction between the selected proteins was monitored by measurement of the β-galactosidase activity. As shown in Figure 4, the GAD–GKβ fusion protein produced a strong interaction with GBD–Mig1. However, this interaction was absent in a GBD–Mig1S311A mutant, indicating that the Ser311 residue of Mig1, previously identified as essential for the Hxk2 interaction , was also required for interaction with GKβ. No interaction was observed between GAD–GKβ and GBD–Hxk2 or GBD–GKβ, indicating that these proteins did not form dimers. Similar levels of all protein fusions were detected in all of the cases (results not shown). To confirm the interaction we performed co-immunoprecipitation and GST-pulldown assays, but we were not able to observe any interaction between GKβ and Mig1, probably because it was too weak or transient to be detected by these methods (results not shown).
Human GKβ binds in vivo to the SUC2–Mig1 repressor complex
Previous reports have demonstrated that Hxk2 participates in the SUC2–Mig1 repressor complex by interaction with Mig1 protein and not by direct binding to DNA [14,31,32]. As GKβ and Mig1 may form an in vivo complex (see above), we investigated the biological significance of this interaction by determining whether GKβ localized in the SUC2–Mig1 repressor complex in a glucose-dependent manner. To test this hypothesis, we used ChIP assays. Our results show that in cells grown in high-glucose medium, both HA–GKβ and Hxk2 proteins were recruited to a DNA fragment of the SUC2 promoter containing the Mig1-binding site (Figure 5A, lanes 1 and 3). Conversely, in low-glucose medium, HA–GKβ and Hxk2 binding to the SUC2–Mig1 complex was abolished (Figure 5B, lanes 2 and 4). Binding of HA–GKβ to the Mig1 site of the SUC2 promoter was dependent on the presence of Mig1, since in Δmig1 cells, no binding of HA–GKβ to the SUC2 promoter was observed under any growth conditions (Figure 5B). No DNA amplification was observed when we used cells with untagged GKβ or cells with HA-tagged Rgt1 protein (results not shown).
Taking all of these results together, we suggest that GKβ forms a similar complex with Mig1 as Hxk2 does, and that this complex is involved in the glucose-regulated expression of the SUC2 gene. This complex exits the nucleus under conditions of Snf1 activation, the presence of Mig1 and Msn5 being necessary for the nuclear exit.
In the present study, we have taken advantage of the facility to manipulate the S. cerevisiae genetic system to determine the mechanisms by which GKβ regulates glucose signalling in yeast cells and to examine the glucose-dependent localization of the GKβ in yeasts.
Reports suggest that Hxk2, the yeast orthologue of GKβ, has a dual function in yeast cells, first initiating the intracellular metabolism of glucose by its enzymatic activity and secondly, signalling glucose repression by its interaction with nuclear Mig1 repressor, regulating its phosphorylation status . Since Hxk2 protein mutations without phosphorylating activity but retaining activity in glucose repression signalling have not been achieved [33,34], the idea that the sugar kinase activity and the sugar signalling properties are mediated through separate domains of Hxk2 is still controversial. In the present study, we demonstrate that recombinant human GKβ is able to signal glucose repression in S. cerevisiae; both GKβ and Hxk2 have a nuclear location during growth in high-glucose conditions, and both are able to regulate the glucose-responsive SUC2 gene in S. cerevisiae. We also report that GKβ is physically associated with the Mig1 protein repressor and ChIP assays confirmed that both GKβ and Hxk2 interacted with Mig1 in a cluster with DNA fragments containing the MIG1 site of the SUC2 promoter. Therefore our results suggest that GKβ and Hxk2 use similar mechanisms to regulate glucose signalling in yeast.
The other important finding of the present study regards the subcellular location of GKβ. We report that GKβ moves between the nucleus and mitochondria in the yeast cell in response to the glucose concentration in the medium: under high-glucose conditions, GKβ is enriched in the nucleus and resides there in association with Mig1; however, under low-glucose conditions, GKβ exits the nucleus and eventually reaches the mitochondria, either because it has uncharacterized mitochondrial determinants (GKβ, lacks the hydrophobic N-terminal sequence present in Hxks I and II that allows their binding to the mitochondria ), or because it interacts with another protein that has these mitochondrial localization determinants, as in the case of the mammalian pro-apoptotic factor BAD (Bcl-2/Bcl-XL-antagonist, causing cell death) which is involved in the mitochondrial targeting of GK in both liver and pancreatic β-cells . For GKβ to reach this subcellular location, first Mig1 must be present in the cell and Snf1 must be active. Then, binding of Mig1 to GKβ allows the Mig1–GKβ complex to exit the nucleus by a mechanism dependent on the exportin Msn5. Finally, the complex is located in the mitochondrial network, a location that has never been reported for either Mig1 or Hxk2.
Since the subcellular distribution of GK in mammalian pancreatic β-cells is also regulated by glucose: low levels of glucose causes an association of GK with mitochondria, whereas high-glucose levels prevents GK translocation to this organelle , the yeast system offers a possibility to study the nature of the determinants present in GKβ or the identification of interactive proteins that target GKβ to the mitochondria under low-glucose conditions.
The role of Mig1 in regulating GKβ exit from the nucleus could be compared with the role of the GKRP in mammalian cells [4,5]. However, these two proteins present different domain structures and only 17% similarity (assessed by Clustal analysis), suggesting that the relationship between the two proteins is only at the level of their interaction with GK.
Taken together, the results of the present study indicate that despite the strong phylogenetic difference between the human GKβ and the yeast Hxk2 proteins, the mechanism of glucose signalling is maintained, suggesting that the function of the regulatory domain of these proteins has been conserved throughout evolution.
We are very grateful to J. Guinovart for fruitful discussions. This work was supported by grants BFU2004-02855-C02-02 and BFU2007-66063-C02-02/BMC from the MEC-DGI (Spain). A. R. was supported by MEC-DGI (Spain). D. A. was supported by the Fundación Carolina (Spain). This work was also supported by grants from the CIBER de Enfermedades Raras, an initiative of the ISCIII, to P. S.
Abbreviations: ChIP, chromatin immunoprecipitation; DAPI, 4′,6-diamidino-2-phenylindole; GAD, Gal4-activating domain; GBD, Gal4-binding domain; GFP, green fluorescent protein; GK, glucokinase; GKβ, pancreatic β-cell GK; GKL, liver GK; GKRP, GK regulatory protein; GST, glutathione transferase; HA, haemagglutinin; Hxk, hexokinase; Mig1, multicopy inhibitor of GAL gene expression 1; Msn5, multicopy suppressor of SNF1 mutation 5; ORF, open reading frame; RFP, red fluorescent protein; SD, synthetic complete medium containing glucose as the carbon source; Snf1, sucrose non-fermenting 1; Su9, F0F1 ATPase subunit 9; SUC2, sucrose fermentation 2
- © The Authors Journal compilation © 2008 Biochemical Society