Characterization of Glucokinase-binding Protein Epitopes by a Phage-displayed Peptide Library IDENTIFICATION OF 6-PHOSPHOFRUCTO-2-KINASE/FRUCTOSE-2,6-BISPHOSPHATASE AS A NOVEL INTERACTION PARTNER*

The low affinity glucose-phosphorylating enzyme glucokinase shows the phenomenon of intracellular translocation in beta cells of the pancreas and the liver. To identify potential binding partners of glucokinase by a systematic strategy, human beta cell glucokinase was screened by a 12-mer random peptide library displayed by the M13 phage. This panning procedure revealed two consensus motifs with a high binding affinity for glucokinase. The first consensus motif, LSA XX VAG, corresponded to the glucokinase regulatory protein of the liver. The second consensus motif, SLKVWT, showed a complete homology to the bifunctional enzyme 6-phosphofructo-2-kinase/fructose-2,6-bisphosphatase (PFK-2/FBPase-2), which acts as a key regulator of glucose metabolism. Through yeast two-hybrid analysis it became evident that the binding of glucokinase to PFK-2/FBPase-2 is conferred by the bisphosphatase domain, whereas the kinase domain is responsible for dimerization. 5 (cid:1) -Rapid amplification of cDNA ends analysis and Northern blot analysis revealed that rat pancreatic islets express the brain isoform of PFK-2/FBPase-2.

The low affinity glucose-phosphorylating enzyme glucokinase (hexokinase type IV) plays a pivotal role for the coupling of physiological millimolar glucose concentration changes to glycolysis in liver, pancreatic beta cells, and neuroendocrine cells (1)(2)(3)(4)(5)(6)(7). In pancreatic beta cells glucokinase, acting as a glucose sensor, catalyzes the rate-limiting step for glucosestimulated insulin secretion (1,2,4,8). Studies in glucokinase knock-out mice, as well as the metabolic profile of diabetic patients with mutations of the glucokinase gene, provide evidence that the glucose sensor function of this enzyme in pancreatic beta cells can be fulfilled only within a narrow range of enzyme activity (5, 9 -12). Thus principles of posttranslational regulation of beta cell glucokinase have gained special interest in recent years (13)(14)(15)(16)(17)(18). Liver glucokinase is regulated by insulin and glucagon on the transcriptional and translational level and by the glucokinase regulatory protein (GRP) 1 via translocation between cytosol and nucleus (6, 19 -23). In particular this translocation confers a short term adaptation of glucose metabolism to changes of the glucose concentration in the physiological range (20 -25). In pancreatic beta cells the situation is different from that in liver. The glucokinase activity level in insulin-producing cells is regulated by glucose. However, this modulation is not mediated by the liver type GRP, which was not detected in pancreatic beta cells (13,15,17,18). Studies with digitonin-permeabilized insulin-producing RINm5F cells provide evidence that glucokinase interacts with an as yet unidentified protein factor capable of modulating the activity (15). The existence of this protein is supported by morphological data that report changes in the spatial distribution of glucokinase immunostaining in response to the nutritional status (16,26). To identify proteins that interact with glucokinase, we used a random peptide phage display library to select epitopes with high binding affinity. Through this strategy two glucokinase-binding peptide consensus sequences were identified that showed a high homology to (i) the GRP of the liver (27) and (ii) to the bifunctional regulatory enzyme 6-phosphofructo-2-kinase/fructose-2,6-bisphosphatase (PFK-2/FBPase-2) (28,29). In particular, the binding of glucokinase to PFK-2/FBPase-2 may help to explain recent observations of metabolic channeling of glycolysis in liver (30). The interaction of glucokinase with PFK-2/FBPase-2 might as well have a regulatory role in metabolic stimulus-secretion coupling in beta cells because the bifunctional enzyme is also expressed in pancreatic islets.

EXPERIMENTAL PROCEDURES
Materials-Collagenase P, restriction enzymes, the SP6/T7 transcription kit and the DIG nucleic acid detection kit were obtained from Roche Molecular Biochemicals. Hybond N nylon membranes, the ECL detection system, and the autoradiography films were from Amersham Pharmacia Biotech. Guanidine thiocyanate was from Fluka (Neu-Ulm, Germany). Restriction enzymes and modifying enzymes for the cloning procedures were from New England Biolabs (Beverly, MA) or Fermentas (Fermentas, St. Leo-Rot, Germany). Custom oligonucleotide synthesis was done by Life Technologies, Inc. or MWG Biotech (Ebersberg, Germany). Media and supplements for culture of yeast were from CLONTECH (Palo Alto, CA). Columns for DNA purification were from Qiagen (Hilden, Germany). All other reagents of analytical grade were from Merck.
Immobilization of Recombinant Human Beta Cell Glucokinase-Human beta cell glucokinase protein was expressed in M15 Escherichia coli bacteria and purified by Ni-NTA metal chelate chromatography as described (31). The recombinant protein was immobilized by His tag on Ni-NTA-HisSorb TM strips (Qiagen). 200 l of the protein solution (50 g/ml glucokinase protein in 0.1 M NaHCO 3 ) were incubated overnight at 4°C with gentle agitation in the microplate wells. Thereafter the supernatant was removed, and the wells were blocked with 300 l of blocking buffer (5 mg/ml bovine serum albumin in 0.1 M NaHCO 3 ) for 1 h at 4°C with gentle agitation.
Phage Display Screening-The selection of peptides able to bind to glucokinase was performed with the Ph.D.-12 TM phage display peptide library (New England Biolabs). The combinatorial library of random 12-mer peptides was incubated in glucokinase-coated and blocked wells for 1 h at room temperature. A control experiment using only blocked wells was also performed to exclude unspecific binding. The following biopanning procedure was carried out according to the instructions in the manufacturer's manual. Specifically bound phages were eluted by disruption with glycine buffer or by incubation with 100 l of the protein solution (200 g/ml glucokinase protein in 0.1 M NaHCO 3 ). The recovered phages were amplified in ER2537 E. coli bacteria, and the phage titer was expressed as plaque-forming units. The whole procedure was repeated three to four times, and individual phage clones of each panning were selected. Phage DNA of single colonies were prepared using the QIAprep Spin M13 kit (Qiagen) and characterized by dideoxy chain sequencing (32).
The programs PeptideSearch (EMBL, Protein & Peptide Group, Heidelberg, Germany; www.mann.embl-heidelberg.de) and Bic2 (EBI, Wellcome Trust Genome Campus, Cambridge, UK; www.ebi.ac.uk) were used for homology searches comparing amino acid sequences of glucokinase interacting peptides with sequences listed in the data bases Swall, Swissprot, and Trembl.
Phage ELISA-The conditions in the ELISA experiments were essentially the same as those used in the panning procedure, and all washing steps were performed as described in the manufacturer's manual. Blocking buffer was prepared with 2.5% skimmed milk in Trisbuffered saline. 2-fold serial dilution of consensus sequence displaying phages starting from 10 15 phages/well were analyzed for binding to glucokinase. 200 l of horseradish peroxidase-conjugated anti-M13 antibody (Amersham Pharmacia Biotech; diluted 1:5000 in blocking solution) was added to each well, and bound phages were visualized using 2Ј,2Ј-azino-bis(3-ethylbenz-thiazoline-6-sulfonic acid) (ABTS) in citrate buffer as substrate for the peroxidase. The A 405 values were measured after 20 min of incubation using a htIII microplate reader (Anthos, Köln, Germany).
5Ј-RACE of Rat Islet PFK-2/FBPase-2-2 g of total RNA from freshly isolated islets from Wistar rats were subjected to cDNA synthesis by LD PCR (18 cycles) according to the protocol of the SMART cDNA library construction kit (CLONTECH). 1 or 5 l from the spin columnpurified islet cDNA (corresponding to 50 and 250 ng of cDNA) were used for the specific PFK-2/FBPase-2 RACE PCR reaction (3 min of initial denaturation at 94°C; 25 cycles of 30 s at 94°C, 30 s at 55°C or 58°C, and 60 s at 72°C; 5 min of end extension at 72°C in a total volume of 100 l) with the SMART 5Ј PCR forward primer (5Ј-AAGCAGTGGTAT-CAACGCAGAGT-3Ј) and the rat PFK-2/FBPase-2 consensus reverse primer (5Ј-(G/C)AC(A/G)TGGAT(A/G)TTCAT-(G/C)AGGTA(A/G)TA-3Ј), which contained wobble bases according to the various isoforms. 10 l of the PFK-2/FBPase-2 RACE PCR products were separated in a 1% Sea-Plaque GTG low melting gel. Specific bands were sliced from the gel (20 -30 l) and used for Taq polymerase-amplified cloning into the pTOPO vector (Invitrogen, Groningen, The Netherlands) according to the manufacturer's protocol. The sequences of the PFK-2/FBPase-2 RACE PCR fragments were obtained by the dideoxy chain termination method (32) using a LiCor 4200 automated sequencer (MWG Biotech). Homology plots were generated by Fasta3 and NCBI-Blast2 searches (EBI) on the basis of GenBank TM cDNA sequences of different PFK-2/ FBPase-2 isoforms from the rat. A full-length rat islet PFK-2/FBPase-2 cDNA was amplified from LD islet cDNA with specific primers coding for the brain isoform of rat PFK-2/FBPase-2 (5Ј-TAGCAGGATCCAT-GTCTGAGAATAGTACATTTTCCA-3Ј; 5Ј-TAGTAGTCGACTCAG-GAGAGCAAAGTGAGG-3Ј) and subcloned with BamHI and SalI restriction sites into the pQE30 vector (Qiagen) followed by sequence analysis.
Yeast Two-hybrid Analysis-Yeast two-hybrid analysis was performed with the Matchmaker GAL4 system 2 (CLONTECH) as described in the manufacturer's manual. Cloning procedures were performed as described previously (33,34). The coding cDNA sequence of the human beta cell glucokinase was amplified by PCR from the pQE30-GK plasmid (31) with primers containing NcoI and XhoI restriction sites and subcloned in frame to the activation domain (AD) into pACT2 (CLONTECH). PFK-2/FBPase-2 coding cDNAs for rat liver (35) (GenBank TM accession number Y00702) and rat brain/islet (36) (Gen-Bank TM accession number S67900) were amplified by PCR from the corresponding pQE30 plasmid with composite rimers containing SmaI and BamHI restriction sites and subcloned in frame to the binding domain (BD) into pAS2-1 (CLONTECH). Additionally the kinase domain (amino acid residues 1-257) and the bisphosphatase domain (amino acid residues 258 -470) of the liver PFK-2/FBPase-2 were subcloned separately into the pAS2-1 (BD) vector by the same method. All constructs were verified by sequence analyses of the inserts. The yeast Saccharomyces cerevisiae strain CG1945 (CLONTECH) was simultaneously transformed with pACT2 or pACT2-GK and each of the pAS2-1-GRP wild type and mutant plasmids by using the lithium acetate procedure (37). Activation of the HIS3 reporter construct was determined by growth on selection SD agar plates lacking leucine, tryptophan, and histidine after 5 days. For the quantitative measurement of ␤-galactosidase reporter gene activity, the host strain Y190 (CLON-TECH) was transformed in the same way, and yeast was grown on SD agar plates without leucine and tryptophan.
Quantitative Chemiluminescent ␤-Galactosidase Assay of Yeast Twohybrid Interaction-The resulting transformed yeast clones were grown in SD medium lacking leucine and tryptophan overnight. The yeast protein was isolated from the cell pellet by vigorous vortexing with glass beads (Roth, Karlsruhe, Germany) in phosphate buffer as described (38). Insoluble material was pelleted at 14,000 rpm and 4°C for a period of 30 s. Protein content of the solution was determined using a Bio-Rad protein assay. For the quantitative chemiluminescent ␤-galactosidase reporter gene assay, the protein solution was added to Galacto Star reaction buffer (Tropix, Bedford, MA) as described by the manufacturer. The light emission was recorded as a 1-s integral in a microplate using a Victor 2 luminometer (Wallac, Freiburg, Germany), and the specific activity was calculated in relation to ␤-galactosidase standard values and protein content.
Northern Blot Analyses-Pancreatic islets from fed male Wistar rats (300 -400 g of body weight) were isolated by collagenase digestion. When other tissues were used they were also taken from Wistar rats. Rat pancreatic islets were purified by Ficoll gradient centrifugation and were used immediately for isolation of RNA. Total RNA from rat tissues was isolated by a combined water saturated phenol-chloroform-isoamyl alcohol extraction with an addition of ultrapure glycogen to achieve full precipitation of islet RNA (39). 20 g of total RNA/lane was subjected to electrophoresis on denaturing formamide/formaldehyde 1% agarose gels and transferred to nylon membranes. Hybridization was performed as described before (15) using 11-DIG-UTP-labeled antisense cRNA probes coding for rat liver PFK-2/FBPase-2 and rat islet/brain PFK-2/ FBPase-2 (36). The DIG-labeled hybrids were detected by an enzymelinked immunoassay using an anti-DIG-alkaline-phosphatase antibody conjugate. The hybrids were visualized by chemiluminescence detection on a light-sensitive film for quantification by densitometry with the National Institutes of Health Image 1.58 program. Ribosomal bands were used as control markers for gel loading (not shown).
Statistical Analyses-The data are expressed as the mean values Ϯ S.E. Statistical analyses were performed by ANOVA followed by Bonferroni's test for multiple comparison using the Prism analysis program (Graphpad, San Diego, CA). The sequence homologies between the islet and the liver PFK-2/FBPase-2 proteins were calculated by the DNASIS V2.1 program (Amersham Pharmacia Biotech).

RESULTS
Isolation of Glucokinase-binding Phages-A 12-mer random peptide M13 phage display library was used to screen immobilized human beta cell glucokinase for potential interaction partners. For this purpose recombinant glucokinase protein was fixed through the N-terminal His tag sequence to Ni-NTAcoated microplates. Aliquots of the phage library were allowed to interact with recombinant protein at room temperature for 1 h followed by wash procedures of increasing stringency (panning). Bound phages were recovered and reamplified for the next round of panning. Sequence analysis of enriched phages from the second to the fourth round of panning showed peptide inserts that matched to the consensus sequences SLKVWT (Table I, (Table I). Through screening of the peptide data bases Swall, Swissprot, and Trembl by the PeptideSearch routine of the EMBL, the LSAXXVAG consensus motif showed a complete homology to the human (swissnew Q14397 GCKR human glucokinase regulatory protein) and rat (swissnew Q07071 GCKR rat glucokinase regulatory protein) GRP of the liver, which is a well characterized interaction partner of the glucokinase. The phage peptide sequence HGMKVWTLPATS, which was found in 10 clones by the panning procedure, could be aligned to the consensus motif SLKVWT (Table I, part A). Data base analysis revealed a complete match to the enzyme 6-phosphofructo-2kinase/fructose-2,6-bisphosphatase, which is also known as phosphofructokinase type 2 (swissprot P0793 F261 rat 6PF-2-K/Fru-2,6-P2ase; swissprot P16118 F26L human 6PF- The sequences of peptides displayed by glucokinase-binding phage were isolated in three independent experiments each consisting of four rounds of high stringency panning from a 12-mer random peptide M13 phage display library. Two consensus peptide sequences A and B (shown with black background) were identified which were subjected to homology searches by the Peptide (EMBL) and Bic2 (EBI) program in the databases Swall, Swissprot, and Trembl. Amino acids 298 -303 of the bifunctional enzyme PFK-2/FBPase-2 correspond to consensus sequence A, and amino acids 181-188 of the rat liver GRP correspond to consensus sequence B.
In the next step the interaction of the PFK-2/FBPase-2 and GRP consensus phage peptides were quantified by an ELISA binding assay (Fig. 1). The titration curves with the preferentially enriched phage clones HGMKVWTLPATS (corresponding to PFK-2/FBPase-2) and EYLSAIVAGPWP (corresponding to GRP) clearly show a significant binding affinity to immobilized glucokinase protein. The GRP motif exhibited a strong interaction with glucokinase that was significantly different from the background level at a concentration of 3.1 ϫ 10 13 phages/well (Fig. 1B), whereas a significant binding of the PFK-2/FBPase-2 consensus motif was detectable at a concentration of 1.2 ϫ 10 14 phages/well (Fig. 1A). Control phages displaying 12-mer peptides with a random distribution of amino acids did not show any specific interaction with glucokinase protein (Fig. 1). Furthermore, the specific phages with the PFK-2/FBPase-2 and GRP consensus peptide sequences did not exhibit a specific binding to microplates that were blocked by skimmed milk. Recombinant glucokinase protein, which was eluted from the wells with 300 mM imidazoline, retained full catalytic activity, excluding nonspecific hydrophobic interaction with denatured glucokinase protein during the panning procedures (data not shown).
Yeast Two-hybrid Interaction of Glucokinase with Liver Type PFK-2/FBPase-2-To study the interaction of glucokinase with the novel interaction partner PFK-2/FBPase-2 and, for internal control, with the GRP of liver, the cDNAs were subcloned into the pAS2-1 (fusion to binding domain) or the pACT2 (fusion to activation domain) vector of the GAL4 system. Control experiments excluded nonspecific interactions of glucokinase, PFK-2/FBPase-2, and the GRP of liver with the truncated activation and binding domains of the vectors (data not shown). Furthermore, interactions of the proteins were not affected by the location within the binding domain or the activation domain.
Quantitative analysis of the protein interactions by the ␤-galactosidase reporter enzyme clearly showed a strong binding of the GRP to glucokinase with a 12-fold increase of the reporter activity compared with background by the truncated activation domain and binding domain interaction (Fig. 2). The PFK-2/ FBPase-2 protein exhibited a 4-fold increase of ␤-galactosidase activity above the background level (Fig. 2). The glucokinase protein did not bind to itself as shown by reporter activities that were not significantly different from the background level (Fig. 2).
The PFK-2/FBPase-2 enzyme consists of two functional domains that confer the kinase and bisphosphatase activity. From the phage display peptide consensus motif the binding to glucokinase is conferred by the bisphosphatase domain. To verify this epitope for the separate domains of the bifunctional enzyme, yeast two-hybrid interaction studies were conducted with the kinase or the bisphosphatase domain fused to the binding domain of the pACT2 vector (Fig. 3). These studies clearly indicated that the kinase domain of the PFK-2/ FBPase-2 enzyme did not interact with the glucokinase protein as shown by the low ␤-galactosidase reporter gene activity (Fig.  3A). In contrast the bisphosphatase domain and the holoenzyme of PFK-2/FBPase-2 exhibited clear interaction with glucokinase (Fig. 3A).
Because the PFK-2/FBPase-2 acts as a dimer, the yeast two-hybrid system was used to evaluate the molecular basis for this interaction. The PFK-2/FBPase-2 holoenzyme showed a strong interaction with a 50-fold increase of ␤-galactosidase reporter activity above the background level (Fig. 3B). In contrast, the bisphosphatase domains, which bind to glucokinase, did not interact with each other as indicated by reporter activities, which were not significantly different from the background activities (Fig. 3B). On the other hand, the kinase domains of PFK-2/FBPase-2 showed a strong interaction with a 20-fold increase above the basal level (Fig. 3B). These data provide direct evidence that dimerization of the PFK-2/ FBPase-2 protein takes place within the kinase domain of this bifunctional enzyme consistent with the crystal structure of the rat testis isoform (40). A comparison of the ␤-galactosidase reporter activities provides evidence that the binding affinity between the kinase domains of PFK-2/FBPase-2 is significantly stronger than that of glucokinase with the holoenzyme (Figs. 3  and 4). Thus it is likely that glucokinase interacts with an already dimerized PFK-2/FBPase-2 protein.

PFK-2/FBPase-2 Expression in Rat Pancreatic Islets-
The kinetic characteristics and the regulatory properties of the PFK-2/FBPase-2 enzyme show significant differences between the various tissue-specific isoforms (28,29). Because the binding motif of PFK-2/FBPase-2 with glucokinase is highly conserved between the liver, brain, kidney, and testis isoform, a 5Ј-RACE of the variable N-terminal region was performed with islet cDNA as template. Using a PFK-2/FBPase-2 wobble reverse primer, a 900-base pair PFK-2/FBPase-2 cDNA fragment could be amplified at an annealing temperature of 55°C or 58°C (Fig. 4). The length of the amplification product corresponded to the expected length of various PFK-2/FBPase-2 cDNA isoforms (28,29). Sequence analysis of this islet PFK-2/ FBPase-2 cDNA revealed in five of six clones a complete homology to the rat brain isoform (Fig. 5) (36). In one clone (K6) a splice variant was observed in which additional 24 bases were detected upstream from nt 832 of the rat brain cDNA (Gen-Bank TM accession number X67900) (36). Surprisingly, this sequence variation could be also found in the rat EMBL RN-PFKBX9 PFK-2/FBPase-2 clone (GenBank TM accession number X65958) (41), which codes for intron 8 to exon 11 of the rat PFKFB2 gene (Fig. 5). Only 24 nucleotides from the 3Јterminal end of intron 8, which comprises altogether 343 nucleotides within the PFKB2 gene, are inserted between exon 8 and exon 9 of the K6 cDNA clone of the islet PFK-2/FBPase-2. Interestingly, the sequence of clone K6 showed a complete homology to the 3Ј end of intron 8 of the mouse kidney PFK-2/ FBPase-2 gene (42). Thus this sequence variation of the islet PFK-2/FBPase-2 cDNA can be interpreted as a novel splice variant, in addition to those that have been described previously for the PFKB2 gene, which codes for the brain, heart, and kidney isoforms of the PFK-2/FBPase-2 enzyme (29,43). The islet PFK-2/FBPase splice variant with the additional 8 amino acid residues showed an overall homology to the liver isoform of 70% (Fig. 6). The sequence comparison of the kinase domain of the islet and liver isoform revealed a homology of 61%, whereas the sequence homology of the phosphatase domain was higher with 79% (Fig. 6). The glucokinase-binding motif SLKVWT of the liver PFK-2/FBPase protein corresponds to the DLKVWT motif of the islet PFK-2/FBPase protein in which the serine is replaced by a aspartic acid residue (Fig. 6).
Northern blot analyses of various rat tissues using the islet PFK-2/FBPase-2 0.9-kb cRNA probe from clone K6 (islet probe), which contained the splice variant of the brain isoform, revealed strong 4.2-and 2.1-kb hybridization signals in insulinproducing RINm5F cells, INS-1 cells, and rat pancreatic islets (Fig. 7, upper panel). A distinct expression of the islet/brain PFK-2/FBPase-2 mRNA could be also observed in brain, testis, heart, and kidney, however, with a weaker intensity compared with insulin-producing cells (Fig. 7, upper panel). The islet/ brain-specific probe did not provide any hybridization signal with liver or muscle tissue (Fig. 7, upper panel). Hybridization with a 0.6-kb cRNA probe coding for the N-terminal region of the rat liver PFK-2/FBPase-2 isoform (44) showed strong 2.1-kb bands only in liver and muscle tissue but not in insulinproducing cells (Fig. 7, lower panel).
Yeast Two-hybrid Interaction of Glucokinase with Islet Type PFK-2/FBPase-2-A full-length cDNA of the islet PFK-2/ FBPase-2 isoform was amplified from a rat pancreatic islet cDNA pool with specific primers from the rat brain PFK-2/ FBPase-2 gene (36). Sequence analysis of the 1.4-kb amplification product from pancreatic islets showed a complete homol- ogy to the rat brain cDNA (data not shown) and was subcloned into GAL4 yeast two-hybrid vectors for interaction studies. The rat islet PFK-2/FBPase-2 holoenzyme exhibited a significant interaction with beta cell glucokinase with a significant growth of yeast on the selection plates (Fig. 8). This interaction was conferred by the bisphosphatase domain in analogy to the data obtained with the liver isoform of PFK-2/FBPase-2 ( Figs. 3 and  8). The co-transfection of the kinase domain and the glucokinase revealed no yeast colonies on the selection plates (Fig. 8). Thus glucokinase interacts with the bisphosphatase domain of PFK-2/FBPase-2 in liver as well as in pancreatic beta cells.

DISCUSSION
Glucokinase (hexokinase type IV) plays a key role in glucose sensing both in liver and in pancreatic beta cells (2)(3)(4)(5)19). The present study was prompted by recent observations in liver and insulin-producing cells that provide evidence that translocation and interaction of glucokinase with proteins are regulatory principles that may explain nutrient-dependent adaptation of enzyme activity (15,21,(23)(24)(25)(45)(46)(47). Random peptide libraries displayed on the surface of phages proved to be a useful tool to identify, by a systematic approach, epitopes specific for protein interaction. Following a stringent selection strategy we identified two peptide consensus motifs that showed high binding affinity for recombinant human beta cell glucokinase immobilized on a solid surface by a Ni-NTA His tag interaction. Data base analysis of the sequence motif LSAXXVAG revealed a complete homology to the hepatic GRP, which is a well characterized anchor protein for glucokinase conferring translocation between cytoplasm and nucleus (27,48,49). Irrespective of the interesting molecular aspects of the glucokinase interaction with the regulatory protein, the identification of this consensus motif can be regarded as a validation of the peptide phage display strategy for the characterization of glucokinase-binding partners.
The second peptide consensus motif SLKVWT showed a complete homology to the liver type of the bifunctional enzyme PFK-2/FBPase-2, which is also known as phosphofructokinase type 2 (28,29,50). Isoforms of this enzyme are expressed in liver, muscle, heart, kidney, testis, and also pancreatic islets. Through regulation of intracellular levels of fructose-2,6bisphosphate, the PFK-2/FBPase-2 enzyme is a very potent modulator of glucose metabolism in various cell types (28,29). Glucokinase interacts with a region of the bisphosphatase domain that is highly conserved among the different isoforms (50). The binding of the PFK-2/FBPase-2 SLKVWT motif was weaker than that of the GRP of liver to glucokinase. This was demonstrated by titration ELISAs of phage binding. Furthermore, the yeast two-hybrid assays confirmed that the interaction site with the glucokinase is in the bisphosphatase domain of liver PFK-2/FBPase-2, whereas the kinase domain mediates the dimerization. These functional data correspond well to the crystallographic studies upon purified PFK-2/FBPase-2 enzyme from rat testis, which proposed the kinase domain for dimerization (40). The crystal structure also confirms that the consensus glucokinase-binding motif is surface-exposed on the bifunctional enzyme (Fig. 9A). Furthermore, the consensus sequences are located on opposite sides of the bifunctional enzyme dimer (Fig. 9B), which suggests a binding stoichiometry of one glucokinase molecule/bifunctional enzyme monomer. At the same time, the location of the glucokinase-binding consensus motif is proximal to the kinase domains in the threedimensional structure, which raises the possibility that the PFK-2 and glucokinase activities may be modulated by the interaction of the two proteins.
The role of PFK-2/FBPase-2 is complex because of its opposing synthetic and degradative catalytic activities and its contrasting regulation by hormones and allosteric effectors (28,29,50). The functional role of this enzyme is determined by the properties of the tissue-specific isoform in its respective tissue. The liver type enzyme is one of the best characterized isoforms that is modulated by phosphorylation and dephosphorylation (28). Current data indicate that PFK-2/FBPase-2 proteins are expressed from at least four different genes denominated PFKFB1-PFKFB4 (29). The liver and skeletal muscle isoforms of PFK-2/FBPase-2 are transcribed from the PFKFB1 gene locus. The PFKFB2 gene codes for the brain, kidney, and heart isoforms of PFK-2/FBPase-2, which do not appear to be regulated by protein kinases or phosphatases at the N-terminal end of the protein (29). Interestingly, this PFKFB2 locus transcribes splice variants, particularly in the brain, which mostly affect the C-terminal ends of the kinase domain and bisphosphatase domains. To characterize the PFK-2/FBPase-2 isoform in rat pancreatic islets, a 5Ј-RACE with a consensus primer from the C-terminal end of the kinase domain was performed with a cDNA pool from isolated rat pancreatic islets. Sequence analysis revealed that the brain isoform is expressed in pancreatic beta cells. The majority of the transcripts showed a complete homology to the E10 splice variant described by Watanabe et al. (51) for PFK-2/FBPase-2 transcripts in brain. Furthermore, we identified a novel splice variant that consists of an additional 24 bases at the 5Ј end of exon 9, apparently because of an alternative splicing of intron 8. It will have to be clarified by investigations upon ribosomal-bound mRNA from insulin-secreting cells that the splice variant is really translated into a PFK-2/FBPase-2 enzyme protein and therefore comprises a differentially spliced PFK-2/FBPase-2 pre-mRNA rather than a partially spliced PFK-2/FBPase-2 pre-mRNA. The islet PFK-2/FBPase-2 enzyme is coded by the PFKB2 gene, which is also expressed in brain, kidney, and heart (29). Hybridization with the 0.9-kb islet PFK-2/FBPase-2 cRNA probe, which consisted of the kinase domain of the brain isoform plus the additional 24 nucleotides from alternative splicing that were found in pancreatic islets, resulted in strong hybridiza-FIG. 4. 5-RACE amplification of PFK-2/FBPase-2 cDNA from rat pancreatic islets. A representative cDNA pool was generated from 2 g of total RNA from rat pancreatic islets by SMART LD-PCR amplification. 1 or 5 l (50 or 250 ng cDNA) of the SMART islet LD-cDNA pool were amplified in a second step at 55°C or 58°C annealing temperature with the 5Ј SMART forward primer and a PFK-2/FBPase-2 wobble reverse primer comprising sequence variations of PFK-2/ FBPase-2 isoforms. The expected PFK-2/FBPase-2 specific amplification product of 900 base pairs was subcloned for verification of the isoform by sequence analysis. tion signals in insulin-producing cells compared with brain, heart, and kidney tissue. This may reflect a somewhat higher mRNA expression level in insulin-producing cells than in brain tissue. However, it cannot be excluded that under high stringency conditions a better hybridization efficacy occurred with insulin-producing cells because of the complete homology to the 24 different nucleotides that were specifically expressed in islet tissue (Fig. 5). Thus the intensity of the bands may not necessarily reflect the relative PFK-2/FBPase-2 enzyme activities in insulin-producing cells and brain, respectively. Sakurai and colleagues (52) identified the heart isoform of PFK-2/FBPase-2 as that expressed in rat pancreatic islets by a selective PCR strategy. However, the design of the specific primers used in this study (52) could not distinguish between the brain, kidney, and heart isoforms, which share a high degree of homology because these isoforms are transcripts from the PFKB2 gene (29). Furthermore, expression of a heart type isoform of PFK-2/FBPase-2 in pancreatic islets would not correspond to the observed band pattern in the Northern blot analyses.
In the present study we have described the interaction of the glucokinase and PFK-2/FBPase-2 proteins on the molecular level. This interaction may be physiologically relevant. Two recent studies using hepatocytes have provided evidence that the interaction of glucokinase and PFK-2/FBPase-2 plays an important role for the regulation of glycolysis and glycogen synthesis (30,53). Permeabilized hepatocytes from rats showed a partial channeling between glucokinase and phosphoglucoisomerase/phosphofructokinase in the presence of saturating FIG. 5. Splice variant of the PFK-2/FBPase-2 cDNA in rat pancreatic islets. 5Ј-RACE PCR was performed with rat islet cDNA using a PFK-2/FBPase-2 consensus reverse primer coding for the N-terminal region of the kinase domain. A 900-base pair fragment was subcloned, and six clones (K1-K6) were subjected to sequence analysis. Nucleotide sequences and the encoded amino acid sequences of the clones K1-K6 from rat pancreatic islets are shown in bold type for comparison with the nucleotide sequences of the rat brain PFK-2/FBPase-2 isoform (top line) and the rat PFKFB2 genomic sequence (clone RNPFKBX9, GenBank TM accession number X65958, intron 8 to exon 11) of the PFKFB2 gene (bottom line). Clones K1-K5 showed a complete homology to the rat brain PFK-2/FBPase-2 isoform (RNF6P2KBI, GenBank TM accession number S67900). Clone 6 revealed a splice variant with 24 additional bases upstream of nucleotide 832 of the rat brain cDNA. This sequence variation could also be found in the rat PFKFB2 gene.
FIG. 6. Sequence alignment of the rat islet and rat liver PFK-2/ FBPase-2 proteins. The sequence of the rat islet PFK-2/FBPase-2 protein was obtained from cloned cDNA products. The sequence of the rat liver PFK-2/FBPase-2 protein was taken from the GenBank TM data base (accession number Y00702). The GK-binding motif and the additional 8 amino acids of the islet splice variant of the PFK-2/FBPase-2 protein are marked by a black background. ࡗ, amino acid residues that for the catalytic site of the kinase or bisphosphatase domain are depicted in bold type. ‹, kinase domain (PFK-2) and bisphosphatase domain (FBPase-2). concentrations of fructose 2,6-bisphosphate (30). This is consistent with different pools of glucose-6-phosphate, which can be channeled either into glycolysis or glycogen synthesis (30). This contention is supported by the observation that overexpression of liver type PFK-2/FBPase-2 by recombinant adenovirus modulated glycolysis and glycogen synthesis in a different manner in nondiabetic control and streptozotocin-diabetic rats (53). The mechanism by which fructose 2,6-bisphosphate increases glycogen synthesis in liver from streptozotocin-diabetic rats remains so far unclear but could be associated with up-regulated glucokinase activity levels (53). Notably fructose 2,6-bisphosphate, which is a strong allosteric activator of phosphofructokinase type 1, has no effect on glucokinase enzyme activity in liver extracts (54) nor on purified recombinant human beta cell and liver glucokinase. 2 Thus it is unlikely that the interaction of glucokinase and PFK-2/FBPase-2 will exert metabolic effects solely through promoting the generation of fructose 2,6-bisphosphate. Furthermore, the interaction of recombinant glucokinase with liver type PFK-2/FBPase-2 did not inhibit glucokinase activity under in vitro conditions like the interaction with the GRP of the liver. 2 From this scenario it can be concluded that the interaction of glucokinase with PFK-2/ FBPase-2 may be required for metabolic channeling without modifying glucokinase enzyme activity. It will have to be clarified in future functional studies how the PFK-2/FBPase-2 is associated with the proposed metabolic channel complex because it is unlikely that the bifunctional enzyme acts through a substrate-product transfer because fructose 2,6-bisphosphate has a regulatory effect on putative metabolic channel components such as phosphofructokinase type 1.
Although the liver isoform of PFK-2/FBPase-2 is under tight control by phosphorylation/dephosphorylation, the brain isoform expressed in insulin-producing cells does not contain the protein kinase A-sensitive serine consensus motif in the Nterminal region of the enzyme. This is consistent with metabolic studies on rat pancreatic islets, which did not show the effects of glucagon upon fructose 2,6-bisphosphate levels. Glucose in the physiological concentration range induces signifi-cant increases of fructose 2,6-bisphosphate in pancreatic islets (55)(56)(57), however, without a lag in the time course typical for PFK-2/FBPase-2 enzyme conversion by phosphorylation/dephosphorylation in liver (56). The interaction of glucokinase with the brain type PFK-2/FBPase-2 enzyme may have significant impact upon the regulation of glucose metabolism in pancreatic beta cells. Firstly, a channeling of the first steps of glycolysis through glucokinase could explain the phenomenon that glucose metabolism in beta cells is dependent upon glucokinase activity even though the high affinity hexokinase isoenzyme comprises 50 -70% of total glucose phosphorylation activity in pancreatic islets (1,58,59). Secondly, the effector molecule fructose 2,6-bisphosphate may participate in metabolic oscillations in beta cells through an allosteric activation of phosphofructokinase that may form a regulatory cycle through a complex of glycolytic enzymes with the PFK-2/FBPase-2 (60,61). This is of particular interest because enzyme tunneling of hexose 6-phosphates has been already demonstrated in pancreatic islets from rats (62).
Through a systematic selection technique we have identified the bifunctional regulatory enzyme PFK-2/FBPase-2 as a new glucokinase interaction partner. In contrast to the GRP of the liver, the interaction of glucokinase with the PFK-2/FBPase-2 enzyme does not result in an inhibition of glucokinase enzyme activity but opens the possibility of an integrative regulatory 2 M. Tiedge, unpublished observation. The cDNAs of full-length rat islet PFK-2/ FBPase-2 and kinase domain (PFK-2) or bisphosphatase domain (FBPase-2) were cloned as fusion proteins together with the BD of the GAL4 yeast two-hybrid system. The glucokinase cDNA was fused to the AD. Yeast two-hybrid interactions were evaluated semiquantitatively through growth selection on SD agar plates lacking leucine, tryptophan, and histidine for 5 days after plating. concept of glycolysis through a multi-enzyme complex that channels not only metabolites but also allosteric modulators. It will be the challenge of future studies to determine the importance of the interactions of PFK-2/FBPase-2 isoenzymes in liver and pancreatic beta cells with glucokinase and other enzymes of the glycolytic pathway as it relates to glucose metabolism.