Molecular forms of 6-phosphofructo-2-kinase/fructose-2,6-bisphosphatase expressed in rat skeletal muscle.

The rat cDNA for the muscle-type (M) isozyme of 6-phosphofructo-2-kinase (PFK-2)/fructose-2,6-bisphosphatase (FBPase-2) contains two putative translation initiation sites. To determine whether the M isozyme expressed in rat skeletal muscle corresponds to the short (PFK2M-sf) or the long (PFK2M-lf) isoform, we have expressed them in Escherichia coli. A third construction was also expressed in which the second ATG codon was deleted (PFK2M-lf delta ATG) to ensure that initiation started at the first ATG. The properties of these recombinant proteins were compared with those of the PFK-2/FBPase-2 present in rat skeletal muscle and liver. The recombinant proteins displayed PFK-2 and FBPase-2 activities and the M(r) values of the subunits measured by SDS-polyacrylamide gel electrophoresis were compatible with the calculated ones. The purified recombinant lf form contained not only the expected lf band (54,500 M(r)) but also the sf band (52,000 M(r)), indicating that the expression system could synthesize the long and the short isoforms from the same mRNA. The kinetic properties of the recombinant sf form were not different from those of the rat muscle enzyme. By contrast, lf delta ATG PFK-2 displayed a higher Km for its substrates and a lower Vmax. Immunoblotting with an antibody directed against the long isoform revealed a 54,500 M(r) band both in the lf and the lf delta ATG recombinant, but no band in rat skeletal muscle extracts. In these extracts, one band of 52,000 and a minor one of 54,500 M(r) were detected by an anti PFK-2/FBPase-2 antibody. The 54,500 M(r) band was recognized by an antibody directed against the L isozyme, suggesting that a small amount of the latter is expressed in skeletal muscle. Thus, the M isozyme differs from the L isozyme by replacement of the first 32 amino acids of the L isozyme by an unrelated nonapeptide.

The nucleotide sequencefs) repoded in this paper has been submitted M26215.

to the GenBankTM/EMBL Data Bank with accession numberfs)
To whom correspondence should be addressed HORM unit, ICP-UCL 7529, avenue Hippocrate, 75, B-1200 Brussels, Belgium. Tel.: 322-764-75-32; Fax: 322-762-74-55. phofructo-2-kinase (PFK-2)'/fructose-2,6-bisphosphatase (FBPase-2) are present in mammalian tissues (1-4). They are referred to as L-(liver), M-(muscle), H-(heart), and T -(testis) type according to their tissue prevalence. They differ in molecular mass, in kinetic and immunological properties, and in response to phosphorylation by protein kinases (4,5 ) . They contain two identical subunits, each of which bears the two catalytic sites. The amino acid sequences of these isozymes have been established from the corresponding cDNAs (4,(6)(7)(8). The rat L and M isozymes have in common an identical sequence of 438 amino acids. They differ at the Nterminal end which, in the L isozyme, contains 32 residues including Ser-32. Phosphorylation of the latter by the cyclic AMP-dependent protein kinase inactivates PFK-2 and activates FBPase-2. In the M isozyme, the N-terminal end is unrelated to the unique part of the L isozyme. It does not contain Ser-32, and thus the M isozyme is not a substrate for CAMP-dependent protein kinase. These two isozymes are encoded by the same gene. They arise from different promoters with splicing of either the first (M isozyme) or the second (L isozyme) exon with the 13 other common exons (9).
The first exon encoding the unique 5' sequence of the rat M-type mRNA contains two in-frame ATG codons (Fig. l ) , which are putative translation initiation sites, and for which Kozak (10) has proposed the following consensus sequence, 5'-(GCC)GCCA/GCCmG-3'. In this respect, the second ATG ( Fig. 1) is in a more favorable context for initiation than the first one. However, the participation of the first translation initiation site has not been ruled out, since the Nterminal amino acid sequence of rat muscle PFK-2IFBPase-2 could not be established (11). Therefore, two isoforms of the M isozyme could exist, a long one starting at the first inframe ATG with a calculated molecular mass of 53,971 Da, and a second one, 21 amino acids shorter, starting at the second in-frame ATG with a calculated molecular mass of 51,880 Da. It is very unlikely that translation starts also at the third in-frame ATG corresponding to Met-41, because the sequence of one of the "tryptic" peptides obtained from PFK-2IFBPase-2 purified from rat skeletal muscle (11) started a t Thr-29 of the sequence shown in Fig. 1.
SDS-PAGE analysis of purified preparations of PFK-2/ FBPase-2 from pigeon skeletal muscle displayed a major band of 53,000 M, and a minor constituent of 54,000 M, (3). The latter had the properties of the L isozyme (3). Immunoblots of rat skeletal muscle PFK-2/FBPase-2 preparations partially isozyme of PFK-P/FBPase-P. Numbers on the left refer to nucleotides, with the two putative initiation codons underlined, those on the right to amino acids. The sequence of cDNA RL2K-5c is identical to that of cDNA 22c, which codes for the L isozyme, except that the first 126 bp of the sequence shown here for the M isozyme replace the first 269 bp of cDNA 22c (7). Thus, the amino acid sequence downstream from Ala-30 (or Ser-32 in the L isozvme) is common to the L and the M isozymes. The anti-Mlf antibody was raised against the peptide whose sequence is underlined.
purified by DEAE-cellulose showed two bands, a major one of 54,000 M,, which was not phosphorylated by CAMP-dependent protein kinase, and a minor one of 55,000 M,, which was a substrate of CAMP-dependent protein kinase and could correspond to the L isozyme (12). Purified preparations of PFK-2IFBPase-2 from rat muscle contained a single 54,000 M , band (11). This band corresponds to the calculated molecular mass (53,971 Da) of the long form of the M isozyme, but is not very different from the L isozyme, which has a molecular mass of 54,570 Da. This study was undertaken to determine whether the M isozyme expressed in rat skeletal muscle corresponds to the short or to the long isoform and whether some L isozyme could be present in skeletal muscle. The cDNAs corresponding to the two isoforms ( I f and s f ) of the M isozyme were cloned in an expression vector, expressed in Escherichia coli, and the biochemical and immunological properties of these two isoforms were compared with those of the enzyme purified from rat skeletal muscle. In a third construction, the second ATG was removed by site-directed mutagenesis to ensure that initiation started at the first ATG codon, and the properties of this obligatory long isoform (ZfAATG) expressed in E. coli were studied. Finally, antibodies were raised against two synthetic peptides, one corresponding to the N-terminal portion unique to the L isozyme (13), and one corresponding to a sequence located between the two putative ATG codons of the M isozyme (Fig. 1). These two antibodies were used to detect the L isozyme and the long isoform of the M isozyme by immunoblotting in various purified preparations.
Remaut (Laboratory of Molecular Biology, State University of Ghent, Belgium). The pBluescript(KS)II+ and XL1-Blue E. coli host were from Stratagene. Helper phage M13K07 was from Promega. Tmmobilon polyvinylidine difluoride hybridization membranes were from Millipore. Q-Sepharose high performance (HiLoad 16/10) and Blue Sepharose CL-GB were from Pharmacia LKB Biotechnology Inc. All other materials were reagent grade.

DNA Manipulations
Standard DNA manipulations were carried out as described (14). Single-stranded phagemid DNA was extracted from packaged phagemid, obtained from overnight incubation of a culture superinfected with M13K07 at a helper phage/host ratio of 10, when the Ae10 of the culture was 0.5 (15). The cloning host for a-complementation analysis and isolating single-stranded phagemid DNA, XL1-Blue, contains the lacZAM15 on a fertility (F') episome (selectable on minimal medium) encoding pili formation. Sequencing was made using reagents from U. S. Biochemicals, except that Sequenase was replaced by T7 DNA polymerase. Oligonucleotide-directed mutagen-esis (16) was performed using the Amersham kit and according to the manufacturer's instructions.
Construction of the pPL-PFK2M Expression Plasmids As a consequence of the cloning procedure, the cDNA corresponding to the M isozyme of PFK-2/FBPase-2 was obtained as two EcoRI fragments, 5c2 (5' end of the cDNA) and 5cl (7), each of which had been cloned in pBR322 plasmid. These two fragments were then reassembled to create the phagemid pBluescript(KS)II+/PFK2M as follows (Fig. 2). A 1346-bp EcoRI-Ssp1 fragment was excised from 5c1, thus removing 65 bp of the 3"untranslated region, and inserted in pBluescript(KS)II+ cleaved with EcoRI and SmaI. The 5c2 cDNA fragment was then inserted after restriction of the phagemid with EcoRI and checked for correct orientation by restriction analysis. A single-stranded (+)-form of this phagemid was prepared (15). It was used as a template for mutagenesis (16) to create a restriction site suitable for cloning the coding sequence immediately downstream of the translation initiation site of the expression vector. Three constructions were made. The first one corresponded to the short form (PFK2M-sf) and the second to the long form (PFKZM-lf) of the M isozyme, in which the translation began a t the second or the first inframe ATG codon, respectively. The mutagenic oligonucleotides chosen to replace the second ATG by a BstBI site and the first ATG by a StuI site had the following sense sequence. (1st ATG codon) Antisense oligonucleotides were used for the site-directed mutagenesis reaction. The mutants were screened by digestion with the appropriate restriction enzyme and sequenced by the Sanger's dideoxy chain termination method (17) on alkali-denatured double-stranded phagemid DNA (14). The procedure for the construction of the expression vectors, schematized in Fig. 3, was as follows. The BstBI mutant was digested with RstBI, its 5' ends were filled-in with DNA polymerase I (Klenow) in the presence of dCTP only, and the 5'protruding C was removed by the mung bean nuclease (14). The phagemid was then digested with BamHI to release the corresponding portion of the M-type cDNA beginning a t the GAA codon next to the second in-frame ATG. The StuI mutant was double digested with StuIIBamHI, thus providing a cDNA fragment beginning at the CCT codon immediately following the first in-frame ATG. The third recombinant phagemid, PFK2M-lfAATG, was obtained by deleting the second ATG from the StuI long form mutant. The three constructs were introduced in an expression vector, pPL, using the inducible leftward promoter of coliphage X (18). pPLcmu299 (19) was digested with NcoI, generating 5' ends which were filled-in with DNA polymerase I (Klenow) in the presence of the four dNTPs. Subsequent digestion with BamHI led to the creation of the compatible ends with the cDNAs. Ligation of the cDNAs into the NcoIIBamHI vector yielded three constructs, pPL/PFKBM-sf, pPL/PFK2M-lf, and pPL/ PFK2M-lfAATG, which contained the cDNAs linked directly to the translation initiation codon of pPL. thermosensitive XcI857 mutant product of a gene present on a compatible multicopy plasmid pcI857 derived from pACYC187 (20). The three expression plasmids were used to transform E. coli MC1061X[pcI857] cells (21), according to Ref. 18. Repression is complete at the nonpermissive temperature (28 "C) and the fully induced state is obtained at 42 "C. Cultures of MC1061X[pcI857] carrying the expression vectors were grown at 28 "C in L broth medium until was 0.15, and protein synthesis was induced by shifting the temperature to 42 "C and shaking the flasks for 150 min a t this temperature. The three expression plasmids directed measurable synthesis of soluble bifunctional enzyme. Under these conditions, no inclusion bodies were detected by phase-contrast microscopy.
Purification of PFK-BIFBPase-2 The purification of PFK-Z/FBPase-P was adapted from described procedures (3,22,23) to be completed within 2 days. The starting material consisted of 3 liters of culture corresponding to 0.1-0.2 unit of PFK-PIFBPase-2. The bacteria were collected by centrifugation (1,000 X g, 15 min) and resuspended in a buffer (25 ml/liter of culture) containing 100 mM KC], 50 mM Hepes at pH 7.5, 1 mM EDTA, 1 mM dithiothreitol, 0.1 mg of phenylmethylsulfonyl fluoride/ml, and 2 pg o f aprotinin/ml. The bacteria were lysed in a motor-driven French pressure cell at an internal cell pressure of 10,000 p.s.i. and the resulting lysate was purified by the following three steps.
Precipitation by Polyethylene Glycol (PEG)-Solid PEG 6000 was added to reach a concentration of 5% (w/v) and the solution was stirred for 30 min and centrifuged at 10,000 X g for 10 min. The supernatant was collected, its pH was adjusted to 7.0, and the concentration of PEG was increased to 20% (w/v). After stirring for 30 min, the 5-20% protein fraction was collected by centrifugation (30,000 X g, 30 rnin). The pellet was resuspended in a buffer (5 ml/  Fig. 4B. PFK-2/FBPase-2 was purified according to Ref. 23 from rat hind leg muscles and liver. The steps of this purification were the same as above.

Peptide Synthesis and Antibodies
Two decapeptides, GELTQTRLQK, corresponding to the N-terminal region unique for the L isozyme, and PTGPALGVCK, corresponding to the unique N-terminal region of the long isoform (lf) of the M isozyme (see Fig. 1) were prepared as described (13) by solid phase on a Beckman 990 B automatic peptide synthesizer. The synthetic peptides were used to raise antibodies in rabbits as described (13). These anti-L and anti-Mlf antibodies were used after partial purification by precipitation with 45% ammonium sulfate. They  (Fig. 2). A HstBI site and a StuI site were created in the vicinity of the second or the first ATG codon by oligonucleotide-directed mutagenesis, yielding pBluescript(KS)-II+/PFKZM (BstBI) and pBluescript(KS)II+/PFKBM (StuI), respectively. The rest of the procedure is explained in the text. For the construction of pPLIPFKZM-lfAATG, in which the second ATG codon has been deleted, the oligonucleotide 5"GAGGCTTTTTCT TCCGCTAATTTTCCAATG-3', was annealed with pBluescript(KS)lI+/ PFKZM (StuI) single-stranded DNA and the mutagenesis reaction was performed as described under "Experimental Procedures." The strategy used to construct the corresponding expression vector was the same as the one described here for pPL/PFKZM-lf.

K h o w + dcTp only
Mung Bern N u e l m

'
EwRV EcoRI I recognized specifically their cognate synthetic antigen without crossreaction in an enzyme-linked immunosorbent assay test. Another antibody, raised against a sequence (EEKASKTRA) of the short isoform, failed to react satisfactorily in immunoblots, although it did recognize its cognate antigen in an enzyme-linked immunosorbent assay test (13).
Protein Electrophoresis and Immunoblotting Proteins were separated by SDS-PAGE (10%) (24). The gel was electroblotted onto a Immobilon polyvinylidine difluoride membrane. The antigens were detected by specific polyclonal antibodies. These included the anti-L and anti-Mlfantibodies and MCL-2, an antibody which recognizes epitopes common to the L and M isozymes (25).
Bound antibodies were detected by a peroxidase-antiperoxidase complex (PAP from Prosan) revealed by the ECL Western blotting detection system (Amersham).

Measurement of Enzyme Activities
PL. The assay of PFK-2 activity was adapted from methods described PFK-2 activity was measured by the rate of formation of Fru-2,6-pPLcmu299/PPKZMJ/ amH1 previously (26, 27), as follows. In a total volume of 0.2 ml, 0-50-pl samples of the enzyme preparation were incubated at 30 "C for up to 15 min in the presence of 2 mM fructose 6-phosphate, 5 mM MgATP, 5 mM potassium phosphate, 1 mM dithiothreitol, 100 mM KC1, 20 mM KF, 1 mg of albumin/ml, and 50 mM Tris-HC1 at pH 8.5 (optimal pH). The reaction was stopped by the addition of 0.2 ml of 50 mM NaOH and by heating a t 80 "C for 5 min. The amount of Fru-2,6-P2 formed was then measured as described (28).
FBPase-2 activity was measured by the formation of ['"PIP, from [2-"'P]Fru-2,6-P2 which was synthesized as described (29). The samples (0-50 pl) of the enzyme preparation were incubated a t 30 "C for 15 min in a total volume of 0. For all the enzymes studied, the linearity of the reaction as a function of the time of incubation and of the concentration of enzyme was checked. One unit of enzyme activity corresponds to the formation of l pmol of product/min under the assay conditions. For the measurement of V,,,.,, K,,,, and Ki the concentrations of ligands were varied accordingly and the constants were calculated by nonlinear least squares fitting to an hyperbola using a computer program ("Ultrafit") from Biosoft.

Other Methods
Protein was measured (30) with y-globulin as a standard. The SDS-PAGE gels were stained with silver nitrate (31). The molecular masses were calculated by a computer program ("The DNA Inspector II+" from Textco).

Preparation of Bacterial Lysates-Different techniques of
lysis, such as three freeze-thaw cycles in a dry ice/ethanol bath, sonication (three times 10 s, with 20-s intervals on ice), or the French pressure cell were tested. Total protein concentration and PFK-2 activity varied in the same proportion for a given lysis procedure. We routinely used the French pressure cell because it gave the highest specific PFK-2 activity (1-2 milliunit/mg of protein), it lysed more than 90% of the cells and it yielded about 10 mg of total protein from 50 ml of a culture induced for 2.5 h, while protein recovery was 5 to 10-fold less after sonication, and 100-fold less after freeze-thawing.
Purification Immunological Characterization-We have used the technique of immunoblotting to assign the bands seen on SDS gels to a specific isoform. The antibodies used were MCL-2, which cross-reacts with the L and M PFK-2IFBPase-2 isozymes, anti-L, which is specific for the L isozyme, and anti-Mlf, which is specific for the long form of the M isozyme. Detection was performed on bacterial lysates for the recombinant forms and on PEG extracts or Q-Sepharose fractions for PFK-2/FBPase-2 from rat tissues (Fig. 7). As expected, MCL-2 reacted with the three recombinant PFK-S/FBPase-2 (Fig. 7a). It showed a 52,000 MI band with sf (lane 5 ) , a 54,500 MI band with lfAATG (lane 7), and not only the 54,500 M, band but also a small amount of 52,000 M, with lf (lane 6). As expected, the anti-Mlf antibody (Fig. 7c) showed no  Mr) isoform of the M isozyme. PEG extracts from liver showed the same 54,500 MI band when incubated with the MCL-2 (Fig. 7a, lane 1) or with anti-L (Fig. 76, lane 1 ) antibody, consistent with the presence of the L isozyme. No signal was obtained with the anti-Mlf antibody (Fig. 7c, lane 1 ). The 52,000 M, band seen in these extracts by Fru-2,6-P2 labeling (Fig. 6, lane 1 ) could not be detected by immunoblotting. PEG extracts from skeletal muscle showed a major band of 52,000 M , and a minor band of 54,500 MI with the MCL-2 antibody (Fig. 7a, lane 2). As to the major band (52,000 M,) it was the main component of the first peak of Q-Sepharose, as expected for the M isozyme (Fig. 7a, lane 3 ) , but was also present in the second peak (lane 4 ) , consistent with the overlapping character of these peaks. No band was revealed with the anti-Mlf antibody in the Q-Sepharose fractions (Fig. 7c, lanes 3  and 4 ) , confirming that the long isoform of the M isozyme is absent in PEG extracts of rat muscle. Therefore, the major 52,000 MI band present in muscle extracts is the short isoform of the M isozyme. As to the minor 54,500 M , band, it could be assigned to the L isozyme because it was recognized by the anti-L antibody (Fig. 7b, lane 2), but not by the anti-Mlf antibody (Fig. 7c, lane 2) and because it was more abundant in the second than in the first peak of Q-Sepharose (Fig. 7,  lanes 3 and 4 ) . Biochemical Properties-The kinetic properties of the purified PFK-2/FBPase-2 are given in Table I. Except for the lfAATG, the kinetic parameters of PFK-2 and FBPase-2 did not differ between the isoforms and were similar to those reported previously for rat skeletal muscle PFK-Z/FBPase-2 (11). Therefore, the heterologous expression of the short form of the M isozyme results in a protein that displays all the biochemical features of the original enzyme. The interpretation of the results obtained with lf is ambiguous since the purified preparation of lf contained a large amount of sf (Fig.  5, lane 5). Thus, it is expected that lf resembles sf rather than 1fAATG. The obligatory long isoform (1fAATG) contained slightly less PFK-2 activity than sf. This resulted from a decrease in VmaX and from an increase in the K,,, values for Fru-6-P and MgATP. It is not known whether this lower PFK-2 activity is due to the N-terminal region or to the deletion of the second ATG and thus of a methionine residue.

DISCUSSION
Our previous cloning and characterization of the M-type cDNA showed that translation of the M isozyme could start at two AUG codons separated by 20 amino acids (7). In order  t o ascertain whether the short or the long isoform of the M isozyme is expressed in skeletal muscle, the most straightforward way would have been to establish the N-terminal sequence of PFK-2IFBPase-2 purified from rat muscle. This could not be obtained because the N terminus was blocked (11). We have therefore resorted to an indirect way. Based on covalent labeling and immunological detection, the present comparison of PFK-2IFBPase-2 from muscle with that from liver and with recombinant M isoforms expressed in E. coli now shows that the M isozyme expressed in rat skeletal muscle, and to a lesser extent in liver, is the short isoform.
Earlier work has shown that skeletal muscle might contain the L isozyme in addition to the M isozyme (3,11,12). Our results show that this is indeed the case. The small amount of L isozyme present in rat skeletal muscle was detectable not only by labeling, but also by immunoblotting. Our results on the relative abundance of the L and M isozymes in muscle are also consistent with Northern blot analysis using cDNA probes specific for the corresponding isozymes (7).
The differences in properties between the L and M isozymes all result from the replacement of the N-terminal 32 amino acids of the isozyme by a nonapeptide in the short form of the M isozyme. Moreover, the kinetic properties of the M isozyme resemble those of the phosphorylated L isozyme (3). One way in which this nonapeptide could mimic phosphorylation of the L isozyme is by providing (an) extra negative charge(s). While the specific N-terminal sequence of the L isozyme provides seven positive and two negative charges, the nonapeptide provides only three positive and two negative charges.
The scanning mechanism for initiation of translation in eukaryotes requires that both the position (proximity to the 5' end of the mRNA) and the context are important for selection of the initiation site (33). In the consensus sequence for initiation in higher eukaryotes, (GCC)GCCA/GCCBG, ;e/Fructose-2,6-bisphosphatase a purine (usually A) in position -3 is the most highly conserved nucleotide, otherwise a G at position +4 is essential (10). Of the 699 vertebrate mRNA sequences that have been analyzed only 5-10% diverge from the "first AUG rule" (10) because their first AUG is in an unfavorable context, so that translation is initiated at a second AUG. The present study shows that translation of the M-type PFK-BIFBPase-2 mRNA is restricted to the second AUG codon in muscle. It is not surprising that both the short and the long M isoforms were synthesized in our expression system. In the latter, initiation is expected to occur not only at the first AUG codon, which belongs to the vector and has been designed for efficient translation but also at an internal AUG codon if it is in a good Shine-Dalgarno context and not too far from the 5' end. The nucleotide sequence upstream from the second AUG indeed fits with the purine-rich Shine-Dalgarno sequence.
Finally, in-frame AUG codons upstream from the translation initiation site have been shown in a few instances to regulate translation in yeasts (34). Evidence that a similar regulation occurs in vertebrates is, however, lacking.