The p53-induced protein TIGAR [TP53 (tumour protein 53)-induced glycolysis and apoptosis regulator] is considered to be a F26BPase (fructose-2,6-bisphosphatase) with an important role in cancer cell metabolism. The reported catalytic efficiency of TIGAR as an F26BPase is several orders of magnitude lower than that of the F26BPase component of liver or muscle PFK2 (phosphofructokinase 2), suggesting that F26BP (fructose 2,6-bisphosphate) might not be the physiological substrate of TIGAR. We therefore set out to re-evaluate the biochemical function of TIGAR. Phosphatase activity of recombinant human TIGAR protein was tested on a series of physiological phosphate esters. The best substrate was 23BPG (2,3-bisphosphoglycerate), followed by 2PG (2-phosphoglycerate), 2-phosphoglycolate and PEP (phosphoenolpyruvate). In contrast the catalytic efficiency for F26BP was approximately 400-fold lower than that for 23BPG. Using genetic and shRNA-based cell culture models, we show that loss of TIGAR consistently leads to an up to 5-fold increase in the levels of 23BPG. Increases in F26BP levels were also observed, albeit in a more limited and cell-type dependent manner. The results of the present study challenge the concept that TIGAR acts primarily on F26BP. This has significant implications for our understanding of the metabolic changes downstream of p53 as well as for cancer cell metabolism in general. It also suggests that 23BPG might play an unrecognized function in metabolic control.
The p53 protein is a transcription factor that acts as the central hub of a tumour suppressor pathway . It is stabilized when cells are exposed to a large number of stresses, such as oxidative stress, nutrient starvation, oncogene activation and DNA damage. This leads to the transcriptional activation of many target genes. Upon mild stress, these target genes induce cell-cycle arrest, reduce oxidative stress and activate DNA damage-repair pathways . This seems to help sublethally damaged cells to repair the incurred insults. Upon stronger stress, cell death and permanent cell-cycle arrest (or senescence) remove damaged cells from the proliferative pool . In previous years, the effects of p53 on metabolism and cellular redox status have received ever increasing attention . One of the milestones was the identification of TIGAR [TP53 (tumour protein 53)-induced glycolysis and apoptosis regulator; also known as C12orf5 (chromosome 12 open reading frame 5)] as a direct target gene of p53 . On the basis of a distant sequence similarity to proteins with F26BPase (fructose-2,6-bisphosphatase) activity, TIGAR was also hypothesized to have this activity. F26BP (fructose 2,6-bisphosphate) is an allosteric activator of one of the key enzymes of glycolysis, PFK1 (phosphofructokinase 1), and an inhibitor of fructose-1,6-bisphosphatase . As such, this activity of TIGAR could reduce F26BP levels and glycolytic flux under conditions of cellular stress. Consistent with this hypothesis, Bensaad et al.  showed that overexpression of TIGAR leads to a modest decrease in F26BP and reduced glycolytic flux. In contrast, shRNA-mediated knock down of TIGAR led to the opposite effect. Functionally, the inhibition of glycolysis leads to reduced apoptosis and autophagy [4,6,7]. This was attributed to increased NADPH production in the pentose phosphate pathway and the ensuing resistance to oxidative stress.
Regulation of PFK1 activity by F26BP is central to the regulation of glycolytic flux . Synthesis and degradation of F26BP is performed by bifunctional proteins that have 6-phosphofructo-2-kinase activity (PFK2 for the synthesis using ATP and fructose 6-phosphate) and F26BPase 2-phosphatase activity (F26BPase for the dephosphorylation to fructose 6-phosphate). These bifunctional proteins are encoded by four distinct genes [PFKFB1 (6-phosphofructo-2-kinase/F26BPase 1)–PFKFB4] that differ in their tissue-specific expression patterns, their relative PFK2 and F26BPase activities, and their regulation by numerous stimuli and metabolites .
The postulated function of TIGAR as an F26BPase has the potential to influence glycolytic flux. However, contribution of TIGAR to the regulation of F26BP levels and glycolytic flux is dependent on the activity of TIGAR towards F26BP relative to that of other cellular F26BPases present in cells. Curiously, little information is available on the kinetic properties of TIGAR. When describing the crystal structure of TIGAR, Li and Jogl  provided kinetic data for TIGAR revealing that the catalytic efficiency (i.e. kcat/Km) of TIGAR as a F26BPase is several orders of magnitude lower than that of the bifunctional liver enzyme PFK2/F26BPase . This led us to hypothesize that F26BP might not be the physiological substrate of TIGAR, and prompted us to re-evaluate the biochemical activity of TIGAR.
The ORF of human TIGAR was amplified by PCR from cDNA derived from the cell line RKO using the primer pair 5′-ATTAC-ACATATGGCTCGCTTCGCTCTGACT-3′ (forward) and 5′-ATTACACTCGAGCCTTAGCGAGTTTCAGTCAGTC-3′ (reverse). The product was digested and inserted into the restriction endonuclease sites NdeI and XhoI of the vector pET28a (EMD Millipore) yielding the plasmid pOH84. All nucleotide sequences were confirmed by sequencing and details of these constructs are available upon request.
Production of recombinant proteins
Human TIGAR protein was produced as a fusion protein with an N-terminal His6 tag in Escherichia coli cells. We transformed the plasmid pOH84 (see above) into BL21(DE3)-CodonPlus® cells (Agilent). Single colonies were inoculated in liquid cultures and grown to a D600 of 0.4. Expression of protein was induced by the addition of 1 μg/ml IPTG and continued for 4 h at 30°C. Cells were lysed using a French press in 20 mM Tris/HCl (pH 8.5), 200 mM NaCl, 5 mM 2-mercaptoethanol and 5% glycerol. After clarification by centrifugation (20000 g for 15 min at 4°C), we purified N-terminally His6-tagged TIGAR using a metal-affinity column (Pierce) on an ÄKTA liquid chromatography system. Elution was performed with increasing concentrations of imidazole in wash buffer [20 mM Tris/HCl (pH 8.5), 200 mM NaCl and 5% glycerol]. Fractions were desalted using G25 Sepharose columns (GE Healthcare) and protein concentration was determined using the BCA assay (Pierce).
Evaluation of TIGAR phosphatase activity
Unless mentioned otherwise, all chemicals and enzymes were obtained from Sigma or Roche. Pi formation was determined in an assay mixture (400 μl) comprising 20 mM Hepes (pH 7.1), 25 mM KCl, 1 mM MgCl2, 1 mM DTT and adequate amounts of TIGAR. The reaction was stopped by the addition of 500 μl of 10% trichloroacetic acid. Pi was assayed in a final volume of 1 ml using the methods described by Fiske and Subbarow  or Itaya and Ui .
The formation of 3PG (3-phosphoglycerate) from 23BPG (2,3-bisphosphoglycerate) was measured spectrophotometrically in 1 ml of an assay mixture comprising 25 mM Hepes (pH 7.1), 25 mM KCl, 1 mM MgCl2, 0.15 mM NADH, 1 mM DTT, 1 mM ATP-Mg, the equivalent of 1 μl each of rabbit muscle GAPDH (glyceraldehyde-3-phosphate dehydrogenase; 4000 units/ml; Sigma) and yeast PG kinase (6300 units/ml; P7634; Sigma), with or without 23BPG. The formation of pyruvate from PEP (phosphoenolpyruvate) was determined spectrophotometrically in the same assay mixture, but with 1 μl of rabbit muscle lactate dehydrogenase (2750 units/ml; Roche) as a coupling enzyme.
Evaluation of dephosphorylation of F16BP (fructose 1,6-bisphosphate) and F26BP
Dephosphorylation of F16BP was determined through the production of Pi (see above) or spectrophotometrically through the formation of fructose 6-phosphate and fructose 1-phosphate. The assay mixture comprised 25 mM Hepes (pH 7.1), 25 mM KCl, 1 mM MgCl2 and 1 mM DTT. Yeast phosphoglucose isomerase (Roche), yeast glucose-6-phosphate dehydrogenase (Roche) and 0.5 mM NADP were included to measure the formation of fructose 6-phosphate. The reaction was stopped by the addition of perchloric acid (final concentration 0.3 M). After neutralization of the samples with K2CO3, fructose 1-phosphate was assayed spectrophometrically with fructose 1-phosphate kinase , in an assay mixture containing 25 mM Hepes (pH 7.1), 25 mM KCl, 2 mM MgCl2, 0.15 mM NADH, 1 mM ATP-Mg and 0.25 mM PEP, as well as rabbit muscle pyruvate kinase and rabbit muscle lactate dehydrogenase.
The F26BP was obtained from Sigma several years ago. Its integrity was verified by checking that fructose 6-phosphate was absent from the preparation, but formed stoichiometrically upon acid treatment. F26BP hydrolysis to fructose 6-phosphate and fructose 2-phosphate was measured by monitoring the formation of fructose 6-phosphate as above. Fructose 2-phosphate was then measured spectrophotometrically as the free fructose in the neutralized perchloric extracts, using yeast glucose-6-phosphate dehydrogenase, yeast phosphoglucose isomerase and yeast hexokinase in the presence of 0.5 mM ATP-Mg.
Analysis of 23BPG phosphatase activity in muscle
Mouse muscle tissue was snap-frozen in liquid nitrogen and homogenized in four volumes of 20 mM Hepes (pH 7.1), 20 mM KCl, 5 μg/ml leupeptin and 5 μg/ml antipain. The lysate was clarified first by centrifugation (15000 g for 10 min at 4°C), and then by the addition of PEG 6000 (Sigma) to a final concentration of 2% followed by centrifugation (15000 g for 10 min at 4°C). The addition of PEG was omitted in the experiments where TIGAR was immunodepleted. Aliquots of the lysates (1 ml) were then loaded on to a 0.5 ml (bed volume) SP–Sepharose column (GE Healthcare) equilibrated with 20 mM Hepes (pH 7.1). Elution was performed with 1 ml of the indicated concentrations of NaCl in 20 mM Hepes (pH 7.1) and analysed for 23BPG phosphatase activity as described above using GAPDH and PG kinase as coupling enzymes in a spectrophotometric test.
To immunodeplete TIGAR, we added 2 μg of an antibody against TIGAR or 2 μg of an antibody against an unrelated protein [SREBP2 (sterol-regulatory-element-binding protein 2)] to 100 μl of the fraction eluted from the SP–Sepharose column with 100 mM NaCl. Triton X-100 was added to a concentration of 0.1% to reduce the non-specific binding of TIGAR to Protein A/G–Sepharose beads. Immunodepletion was performed by adding Protein A/G–Sepharose beads (Santa Cruz Biotechnology) for 1 h at 4°C. After brief centrifugation (800 g for 5 min at 4°C), 23BPG phosphatase activity was measured in the supernatant.
After removal of the tissue culture medium, cells were immediately lysed in 0.5 M perchloric acid and centrifuged for 10 min at 15000 g and 4°C. Pellets were resuspended in 0.1 M sodium hydroxide and protein concentration was determined using the BCA assay. Supernatants were neutralized with 3 M potassium carbonate and clarified by centrifugation for 10 min at 15000 g and 4°C. 23BPG was assayed at 30°C through the stimulation of PGAM (PG mutase) activity [13,14]. The assay mixture (1 ml) comprised 25 mM Hepes (pH 7.1), 25 mM KCl, 0.15 mM NADH, 1 mM DTT, 1 mM MgCl2, 0.5 mM ADP-Mg, 0.1 mM 3PG, the equivalent of 0.33 μl each of rabbit muscle lactate dehydrogenase (2750 units/ml; Roche), rabbit muscle pyruvate kinase (2000 units/ml; Roche) and rabbit muscle enolase (160 units/ml; Sigma), and either samples (typically 2–5 μl of cell extract) in which 23BPG was to be determined or 23BPG standards (0.5–5 nM final concentration in the assay). PGAM activity in this assay came from contaminating activities in the preparations from rabbit muscle enolase and pyruvate kinase. A340 was measured over ~15 min. Under these conditions, the rate of the reaction was proportional with the 23BPG concentration and the concentration of 23BPG in samples could be calculated by comparison with standards. Doubling of the blank value was observed with 0.5 nM 23BPG.
Purification of 3PG
Commercial 3PG was contaminated with ~1% (mol/mol) 23BPG. Before using it for the 23BPG assay, it was therefore critical to remove this contaminant. An aliquot of 3PG (3 ml of 20 mM) in 10 mM Hepes (pH 7.1) was added to a 10 ml Dowex AG1-X8 column [pre-treated with 10 ml of 1 M HCl, and washed with water and 10 mM Hepes (pH 7.1)]. A salt gradient was applied [0–500 mM in 2×100 ml of 10 mM Hepes (pH 7.1)] and 3PG was assayed spectrophotometrically in the fractions. The first half of the peak was free from 23BPG as indicated by low blank values in the 23BPG assay.
For the F26BP assay, tissue culture medium was removed and cells were immediately lysed in 0.2 M NaOH. The resulting extract was collected and used, after appropriate dilution, as described previously .
The cell lines HCT116, U2OS, RKO and HEK (human embryonic kidney)-293T were obtained from Professor Eric Fearon (University of Michigan, Ann Arbor, MI, U.S.A.) and were cultured in DMEM (Dulbecco's modified Eagle's medium) containing 10% FBS, 2 mM L-glutamine and penicillin/streptomycin (Life Technologies). Lentiviral shRNA expression constructs targeting human TIGAR were obtained as a set of seven constructs from Open Biosystems in the vector pGIPZ (allowing expression of shRNA, GFP and puromycin resistance from the same transcript). Lentivirus production was performed as described previously and the indicated cell lines were transduced in the presence of 8 μg/ml polybrene (Sigma) . Selection was performed with puromycin for 4 days.
Western blot analysis
Cells were washed once with PBS and lysed in RIPA buffer [150 mM NaCl, 20 mM Tris/HCl (pH 7.5), 1% Nonidet P40, 0.5% sodium deoxycholate and 0.1% SDS] containing a proteinase inhibitor cocktail (Complete™; Roche). Protein concentration was determined using the BCA assay. Equal amounts were resolved on 10% polyacrylamide gels and transferred on to PVDF membranes using semi-dry transfer (Millipore). Membranes were blocked for 1 h with 5% non-fat dried skimmed milk powder in TBST (Tris-buffered saline containing 0.1% Tween) at room temperature (22°C). Incubations with primary antibodies were performed overnight at 4°C in Tris-buffered saline containing 2% BSA. Antibody concentrations were 1:1000 for anti-TIGAR (Santa Cruz Biotechnology), 1:5000 anti-FLAG (Sigma) and 1:5000 anti-β-actin (Sigma). Subsequently, membranes were washed and incubated in HRP (horseradish peroxidase)-coupled secondary antibodies in TBST containing 5% non-fat dried skimmed milk powder. Signals were revealed using ECL reagents (Pierce) and detected using autoradiography (Fujifilm).
RNA isolation, reverse transcription and qPCR (quantitative PCR)
Human tissue RNA samples were obtained commercially as a mixture from three individual donors (Ambion/Life Technologies). Mouse tissues were obtained from three C57BL6N mice and snap-frozen in liquid nitrogen. RNA extraction using TRIzol® reagent (Life Technologies), reverse transcription using RevertAID reverse transcriptase (Thermo Scientific) and qPCR using SYBR Green qPCR mix (Kapa Biosystems) were performed as described previously . Relative expression levels were calculated using the 2−ΔΔCT method. The primers used were: hTIGARs, 5′-CAGGTGAA-AATGCGTGGAAT-3′ hTIGARas, 5′-TCCAGACAGTTGCTT-GGAGA-3′; mTigar_s, 5′-CAGTTTACCCACGCCTTCTC-3′; mTigar_as, 5′-TTCACCGCCATGTCTTTACA-3′; mTbp_s, 5′-ACCTTATGCTCAGGGCTTGG-3′; mTbp_as, 5′-GCCATAAG-GCATCATTGGAC-3′; hTBPs, 5′-CTGGTTTGCCAAGAAGA-AAG-3′; and hTBPas, 5′-GGGTCAGTCCAGTGCCATAA-3′.
Establishment of Tigar-knockout ES (embryonic stem) cells
A targeting construct for the mouse Tigar (also known as 9630033F20Rik) locus was established by recombineering in DH10b cells. To this end, a BAC (bacterial artificial chromosome) construct containing the genomic locus of mouse Tigar was obtained from Thermo Scientific. An EGFP expression cassette with an intronic neomycin-resistance gene was amplified by PCR from the construct R6K-NFLAP  using the primers TIG_EGFP_s, 5′-ATACATAAGATCTACTGCAAGCATGGTG-TCCAAG-3′ and TIG_EGFP_as, 5′-ATTACACATATGTTAGC-TCTTGTACAGCTCGTCCAT-3′. This cassette was fused with short homology arms amplified by PCR from BAC DNA using the primer pairs TIG_BAC_LH_as, 5′-ATACATAAGATCTCG-CATCTACGCCTTGTCCTGCGGTGAGA-3′ and TIG_BAC_LH_s, 5′-GGCCTGCAGCCTGCAGGAACAG-3′ and TIG_BAC_RH_as, 5′-AGCGAGTGTGAAATCATACAA-3′ and TIG_BAC_RH_s, 5′-ATTATTTCATATGCCGTGCAGTGGAGCCT-CTGAAG-3′. The construct was introduced into DH10b cells containing the BAC clone and the Red/ET recombination system plasmid pSC101-BAD-gbaA-tet (Genebridges). Recombination was induced by the addition L-arabinose [18,19]. Drug-resistant colonies (pIG248) were screened for correct insertion with the primer pairs SENSE_LH_BAC, 5′-GCGTGCTGACCTTCAGA-CAG-3′ and REVERSE_EGFP_AS, 5′-CACGCCGGTGAAC-AGTTCCT-3′ and SENSE_EGFP_S, 5′-GGGCATGGACGAGC-TGTACA-3′ and REVERSE_RH_BAC, 5′-ATACGCAGCGTAT-TACTGA-3′. We then introduced into the vector pGK-DTA  (plasmid number 13440; Addgene) two short homology arms generated by PCR from BAC DNA (using the primer pairs TIG_LH_s, 5′ATTACATGAATTCCTGGAATTGGAGTTGCA-GATAT-3′ and TIG_LH_as, 5′-ATCCCTTGTGGGTACACGG-AGGCAA-3′ and TIG_RH_s, 5′-ATCCTTTGTCCTCATGAAG-GATCTC-3′ and TIG_RH_as, 5′-ATTACATGGTACCGGCGCG-CCAAGTACAGATTCTATCAGAGTTCA-3′) via the restriction endonuclease sites EcoRI and KpnI (pOH201). This allowed us to recover the EGFP-NEO cassette with a flanking sequence from the Tigar genomic locus by Red/ET recombination (pIG250).
|Substrate .||Km (mM) .||kcat (s−1) .||kcat/Km (mM−1·s−1) .||Assay type (formation of) .||Activity at 1 mM (μmol/min per mg of protein) .|
|L-2-Phosphoglycerate||1.5, 1.0||1.39, 1.28||0.93, 1.28||Pi||1.2|
|Fructose 1,6-bisphosphate (F16BP)||1.35, 1.61||1.05, 1.20||0.78, 0.78||Pi||0.9|
|1.25||0.88||0.71||F6P and F1P†||0.75|
|Fructose 2,6-bisphosphate (F26BP)||1.25||0.43||0.34||F6P and F2P†||0.4|
|Substrate .||Km (mM) .||kcat (s−1) .||kcat/Km (mM−1·s−1) .||Assay type (formation of) .||Activity at 1 mM (μmol/min per mg of protein) .|
|L-2-Phosphoglycerate||1.5, 1.0||1.39, 1.28||0.93, 1.28||Pi||1.2|
|Fructose 1,6-bisphosphate (F16BP)||1.35, 1.61||1.05, 1.20||0.78, 0.78||Pi||0.9|
|1.25||0.88||0.71||F6P and F1P†||0.75|
|Fructose 2,6-bisphosphate (F26BP)||1.25||0.43||0.34||F6P and F2P†||0.4|
The resulting targeting construct was linearized with AscI. The mouse ES cell line JM8A3.N1 was obtained from the European Mouse Mutant Repository  and electroporated with 50 μg of the linearized plasmid pIG250 using 0.4 cm cuvettes at 800 V and 3 μF. Culture conditions were as described previously . Single cell clones were obtained by selection in 150 μg/ml G418 (Invivogen). Correct insertion on both sides was demonstrated by long-range PCR (Kapa Biosystems) on genomic DNA from 14 out of 96 clones. The second allele was targeted by culturing two different heterozygote (E10 and F5) clones at high concentrations of G418 . Loss of the wild-type allele was verified by PCR on genomic DNA of the resulting clones using the primers TIG_GENO_rev, 5′-ATGGCAGCAGCAGAGGCCA-AGT-3′, TIG_GENO_s1, 5′-CGGGACCACATGGTGCTGAA-3′ and TIG_GENO_S2, 5′-TGAGCCATGGCGCTTACATGA-3′ in five out of 96 clones. Three homozygous knockout clones (E10D6, F5C12 and F5F11), four heterozygous clones (E10, F5, F5C8 and E10A5) and three wild-type clones (E9, C6 and D3) were characterized further.
Substrate specificity of recombinant TIGAR
Human TIGAR was expressed in E. coli cells as a fusion protein with an N-terminal His6 tag and purified by metal-affinity chromatography. Except for fructose 6-phosphate, F16BP and F26BP, no other potential substrates had been investigated in studies published previously. We tested activity on a series of intracellular phosphate esters (Table 1). The best substrate we found was 23BPG, which was hydrolysed to 3PG. The catalytic efficiencies for 2PG, phosphoglycolate and PEP were lower by about 4-, 10- and 30-fold respectively (Table 1). The structural similarity of the best substrates indicates that TIGAR acts preferentially on a phosphate esterified to the second carbon of a 2-hydroxy-carboxylic acid (Figure 1).
TIGAR substrates share structural elements
Several other compounds were dephosphorylated at lower rates by TIGAR. This included ADP, GDP, ribulose-1,5-bisphosphate, F26BP and F16BP. Dephosphorylation of the last two compounds was investigated in more detail by following the formation of the two monophosphates that could result from their hydrolysis (Figure 2). Remarkably, TIGAR produced a mixture of fructose 6-phosphate and fructose 1-phosphate at a 1:2 ratio, whereas it converted F26BP essentially (>95%) into fructose 2-phosphate and only to a minor extent into fructose 6-phosphate. Thus, unlike the specific F26BPase associated with PFK2, which exclusively forms fructose 6-phosphate , TIGAR mainly forms fructose 2-phosphate, as found previously with non-specific phosphatases in plants and yeasts [24–26]. The catalytic efficiencies of F16BP and F26BP hydrolysis (taking into account both reactions) were close to 200- and 400-fold lower than that observed with 23BPG respectively.
Formation of different fructose monophosphates from F26BP and F16BP
23BPG is a physiological substrate of TIGAR
23BPG is largely known for its role in red blood cells (erythrocytes) where its concentration is much higher than in the rest of the body. It regulates the affinity of haemoglobin to O2 and is synthesized and degraded by the bifunctional enzyme BPGM (bisphosphoglycerate mutase) . In addition, degradation of 23BPG might be due to the MINPP1 (multiple inositol-polyphosphate phosphatase 1), which similarly to BPGM, is highly expressed in erythrocytes . TIGAR protein is not detectable in human erythrocytes (results not shown), suggesting that it does not play a role in regulating 23BPG in these cells.
Except for its role as a cofactor of PGAM, little is known about the functional relevance of 23BPG outside of red blood cells . Concentrations in tissues are considered to be very low (i.e. below 50 μM), but precise measurements are difficult due to the contamination of tissues with red blood cells that contain up to 10 mM 23BPG . To assess the effect of TIGAR on cellular 23BPG levels, we therefore turned to cell culture models. We introduced shRNA to efficiently reduce TIGAR protein levels in several cancer cell lines. In Western blot analysis we observed excellent knockdown of TIGAR with two independent shRNAs (Figure 3A). In all cases, significant increases in 23BPG levels were observed (Figure 3B). In contrast, levels of F26BP were less affected or unchanged (Figure 3C).
Effect of TIGAR knockdown on cellular 23BPG and F26BP levels
Although we obtained excellent knockdown using shRNAs, we were concerned that a residual TIGAR activity might obscure an effect of TIGAR on cellular F26BP levels. Furthermore, it was conceivable that other F26BPase activities were dominant in the tested cell lines under our experimental conditions. To address these concerns, we sought to knock out TIGAR in an experimental system with the highest possible endogenous TIGAR levels. Mouse ES cell lines show Tigar mRNA expression levels similar to those in skeletal muscle, the organ with the highest Tigar expression level among all mouse tissues analysed (Figure 6A, and results not shown). To abolish Tigar expression, we inactivated both alleles of Tigar in the ES cell line JM8A3.N1. The first allele was inactivated by replacing a large part of the Tigar gene with a neomycin-resistance cassette. Homozygosity was achieved subsequently by culture with high concentrations of G418 . We analysed three clones with bi-allelic Tigar inactivation derived from two different heterozygous clones. TIGAR protein was undetectable (Figure 4A) and 23BPG levels were increased approximately 5-fold in the knockout cell lines (Figure 4B). Under these conditions, a 3.5-fold increase in F26BP levels was observed (Figure 4C). Overall, our tissue culture experiments are consistent with 23BPG being a physiological substrate of TIGAR, and suggested that the effect on F26BP is context-dependent.
Knockout of Tigar in ES cells leads to increased 23BPG and F26BP levels
The 23BPG phosphatase activity of TIGAR corresponds to the phosphoglycolate-independent 23BPG phosphatase activity described previously in muscle
The substrate spectrum of TIGAR is reminiscent of a 23BPG phosphatase activity described previously in rat, rabbit and chicken muscle [29,30]. Cellular 23BPG phosphatase activity results from several distinct proteins from the histidine phosphatase family. Together with PGAM and BPGM, TIGAR belongs to this family of enzymes that can accept a phosphate group from 23BPG to form a phosphohistidine intermediate . This phosphate can then be hydrolysed (in the case of BPGM and TIGAR) or transferred on to 3PG (in the case of PGAM) [9,32]. By itself, PGAM has weak 23BPG phosphatase activity. However, upon the addition of phosphoglycolate as a substrate mimic that cannot accept a phosphate group, the phosphohistidine intermediate of PGAM is hydrolysed. The overall action of PGAM on 23BPG in the presence of phosphoglycolate therefore corresponds to a 23BPG phosphatase activity .
In addition to this, several groups have described a 23BPG phosphatase activity distinct from PGAM that was not stimulated, but rather competitively inhibited, by phosphoglycolate. This activity, which we will term ‘phosphoglycolate-independent 23BPG phosphatase’, was biochemically separable from the phosphoglycolate-dependent 23BPG phosphatase activity, and co-purified with a phosphatase activity for PEP and 2PG [13,29,30,33]. This substrate spectrum is similar to the one observed for recombinant TIGAR. Furthermore, the latter was found to be competitively inhibited in its action on 23BPG by phosphoglycolate with a Ki value of 1.1 mM (results not shown). Therefore we hypothesized that TIGAR could be responsible for the ‘phosphoglycolate-independent’ 23BPG phosphatase activity. To test this hypothesis, we first analysed whether PG-independent 23BPG phosphatase activity co-purified with mouse TIGAR protein. We loaded mouse muscle extract on to a cation exchanger (SP–Sepharose) and eluted with increasing salt concentrations. Owing to its high isoelectric point (estimated pI=8.45), mouse TIGAR was expected to be retained on the SP–Sepharose column at pH 7.1. Indeed, using Western blot analysis with a TIGAR-specific antibody we could show that TIGAR protein was retained on the cation exchanger (Figure 5A, bottom panel, lanes FT and 0) and eluted with 50 or 100 mM NaCl (Figure 5A, bottom panel, lanes 50 and 100). Analysis of 23BPG phosphatase activity in the different fractions showed that phosphoglycolate-dependent phosphatase activity was present in the first fractions eluted from the column (Figure 5A, bottom panel, lanes FT and 0), whereas an activity that was not stimulated by phosphoglycolate was eluted in the fractions that contained TIGAR (Figure 5A, bottom panel, lanes 50 and 100). To prove that TIGAR was responsible for this activity, we immunodepleted TIGAR from fractions of muscle extracts eluted from the SP–Sepharose at 100 mM NaCl using a monoclonal antibody directed against TIGAR. As shown in Figures 5(B) and 5(C), immunodepletion of TIGAR led to a 2.5-fold reduction in 23BPG phosphatase activity. These data indicated that TIGAR indeed represents the phosphoglycolate-independent 23BPG phosphatase activity.
Tigar protein represents the phosphoglycolate-independent 23BPG phosphatase activity in muscle
Transcriptional regulation of TIGAR
As described above, the phosphoglycolate-insensitive 23BPG phosphatase activity has been described previously primarily in the skeletal muscle of pigs, rats and chickens [13,29,30,33]. Consistent with this, we found mouse TIGAR mRNA and protein expressed most highly in skeletal muscle (Figure 6A, and results not shown). Surprisingly, the expression pattern in human organs is remarkably different from the expression pattern in mouse organs. Human TIGAR seems to be expressed more ubiquitously (Figure 6B). Furthermore, similarly to Bensaad et al. , we noted that the dependence of TIGAR on p53 is less stringent in mice than in humans (results not shown). This suggests that the relative importance of the regulatory elements driving TIGAR expression has changed during evolution. Hence, studies of TIGAR function in mouse models or murine cell lines might not represent faithfully the situation in human physiology or disease.
The expression pattern of mouse and human TIGAR mRNA in tissues is markedly different
TIGAR is responsible for the phosphoglycolate-independent 23BPG phosphatase activity
TIGAR has been postulated to be a F26BPase on the basis of a distant homology with the F26BPase domains of PFKFB1–PFKFB4. However, a similar degree of similarity is also observed with other members of the histidine phosphatase family, particularly PGAM and BPGM (results not shown). We have found 23BPG to be the best substrate for TIGAR, with weaker activity towards 2PG, phosphoglycolate and PEP. Several lines of evidence suggest that TIGAR is responsible for a phosphatase activity that has been described previously by the group of Carreras and colleagues as phosphoglycolate-independent 23BPG phosphatase [13,29,30,33]. It has the same substrate specificity (acting best on 2-phospho-hydroxy-carboxylic acid), is expressed highly in the muscle (at least in some species, such as rat, mouse and rabbit), has the same behaviour on ion exchangers and its action on 23BPG is not increased, but competitively decreased, by phosphoglycolate. Last, but not least, we observed that immunodepletion of TIGAR from muscle extracts led to a reduction in phosphoglycolate-independent 23BPG activity.
Is 23BPG the physiologic substrate for TIGAR?
Reducing TIGAR protein levels leads to an increase in cellular 23BPG levels suggesting that 23BPG is indeed a physiological substrate for TIGAR. Although 23BPG is clearly the best substrate among those that we have tested, significant activity is also observed towards PEP, phosphoglycolate and 2PG. This apparent lack of specificity might be surprising at first sight. It should, however, be noted that PEP and 2PG are glycolytic intermediates subject to rapid turnover. As such, the weak activity of TIGAR as a PEP phosphatase and a 2PG phosphatase in most cases would be without consequence (see below). The lack of selective pressure to develop an enzyme with higher specificity might explain why TIGAR has retained these activities during evolution.
The low F26BPase activity that we observe for TIGAR is in good agreement with the value described by Li and Jogl . Several lines or reasoning have suggested that the catalytic pocket of TIGAR has not evolved as a F26BPase. First, the catalytic efficiency of TIGAR for this substrate is approximately 400-fold lower than that for 23BPG. Secondly, TIGAR has phosphatase activity towards positions 2 and 6 in F26BP and positions 1 and 6 in F16BP, indicating that these substrates do not adopt a defined orientation in the catalytic pocket, as previously suggested by Li and Jogl . Thirdly, the structure of F26BP differs markedly from that of the substrates on which TIGAR acts best, with up to 400-fold higher catalytic efficiency (Figure 1). Finally, from the data obtained in muscle extracts, we calculated that the activity of TIGAR in skeletal muscle amounted to 0.2 μmol/min per g of muscle (at 200 μM 23BPG at 30°C), indicating that this enzyme has the capacity for decreasing the tissue concentration of 23BPG by ~50% in 1 min. Calculations on the basis of relative catalytic efficiencies using these two substrates indicate that the same amount of TIGAR would only be able to reduce the F26BP concentration by ~0.2% in 1 min. This suggests that TIGAR is much better suited to regulate 23BPG than F26BP levels.
TIGAR as regulator of glycolytic flux, redox balance, apoptosis and autophagy
Several studies have shown that modulation of TIGAR levels in cell lines is associated with changes in the cellular redox status, apoptosis and autophagy [4,6,7,34–42]. TIGAR-knockout mice show evidence of increased oxidative stress in the myocardium leading to increased mitophagy [7,34,43]. Furthermore, loss of TIGAR reduces regeneration and tumour formation in the intestinal epithelium, and seems to play a role in a zebrafish model of Parkinson's disease [4,43,44].
Some of the cellular changes caused by TIGAR have been explained by a decreased glycolytic flux with compensatory increased flux through the pentose phosphate pathway and an increase in the ratio of NADPH to NADP+ . Indeed, several phenotypes were rescued by exogenous addition of ribonucleosides or antioxidants. Several groups have suggested that the phenotypes were caused by changes in F26BP levels .
Although the enzymological data suggest that F26BP is a rather poor substrate for TIGAR, we and others have observed that knockdown and knockout of TIGAR leads to increases in F26BP levels under some conditions, albeit to a lower degree than 23BPG. Cellular F26BP levels are a result of the cellular PFK2 and F26BPase activity. Both activities are present in bifunctional proteins encoded by the PFKFB1–PFKFB4 genes . The ratio between PFK2 and F26BPase activity, the cell-type specific expression and post-translational regulation differ markedly between these proteins. As such, it is expected that the contribution of TIGAR to the regulation of F26BP will vary depending on the expression levels of PFKFB1–PFKFB4. This has several implications for the interpretation of the results of the present study. First, this might suggest that TIGAR only plays a role in regulating F26BP levels in tissues with very low levels of other F26BPase activities. Secondly, increases in F26BP levels upon loss of TIGAR expression could be compensated by changes in expression levels or activity of PFKFB1–PFKFB4. We have knocked down TIGAR by lentiviral transduction of shRNAs, whereas Bensaad et al.  used transient transfection of siRNAs. The increased time after transduction (>4 days) in the present study might have allowed for compensatory changes in cellular F26BPase and PFK2 activities. This might have obscured changes in F26BP levels in the present study. However, in a limited number of experiments with doxycycline-inducible shRNAs, we were not able to find conditions where loss of TIGAR was associated with higher increases in F26BP than of 23BPG (results not shown).
Loss of TIGAR expression has been shown to lead to functional consequences in several experimental systems. Loss of TIGAR consistently led to changes in 23BPG levels in all experimental systems tested, whereas changes in F26BP were less pronounced or absent. This could indicate that the effect of TIGAR on F26BP levels is indirect, possibly via the regulation of 23BPG levels. 23BPG plays a well-defined function in erythrocytes, where concentrations reach 10 mM . Very little is known about its function and abundance in other tissues. Owing to the ‘contamination’ of tissues with red blood cells, where concentrations are very high, the physiological concentration of 23BPG in normal tissues is still unknown. However, it seems reasonable to assume that levels would be similar to those observed in cultured cells (i.e. between 50 and 200 pg/mg of protein) . Functionally, 23BPG is an essential cofactor of the glycolytic enzyme PGAM. Increases in TIGAR protein could lead to a reduction in 23BPG. In turn, this might contribute to the reduced glycolytic flux induced by TIGAR. However, the saturating concentration of PGAM for 23BPG is <0.1 μM . This means that changes in cellular 23BPG probably have little effect on PGAM activity, unless the majority of cellular 23BPG is present in a bound form and thus unavailable for PGAM.
The effect of 23BPG on many other enzymes has been evaluated, but activation or inhibition in most cases is observed at concentrations that are at least one order of magnitude above the assumed physiological concentrations of 23BPG (reviewed in ). Understanding how changes in 23BPG levels might lead to changes in glycolytic flux, cellular redox status and/or F26BP levels will be the subject of future studies.
Does TIGAR activity represent a relevant glycolytic shunt?
TIGAR has been shown to decrease glycolytic flux . Curiously, the phosphatase activity towards 23BPG, PEP and 2PG could directly increase flux through the terminal part of glycolysis. At first sight these observations seem difficult to reconcile. However, in the studies of Bensaad et al. , glycolytic flux was assessed using D-[5-3H]glucose in an assay that only quantifies flux from glucose to triose phosphates. A small relative increase in terminal glycolytic flux might have been masked by an overall reduction in glycolytic flux due to changes in F26BP levels.
The question of whether TIGAR is responsible for a significant glycolytic shunt activity is nevertheless important in light of the observation that cancer cells with expression of low-activity pyruvate kinase isoform PKM2 might show a glycolytic shunt activity that bypasses the pyruvate kinase step . Avoiding the production of ATP at this step was suggested to allow cells to maintain the high glycolytic fluxes required for biosynthetic activity . The precise molecular mechanism of this shunt (which apparently involves phosphorylation of PGAM1 by PEP) and its overall contribution to the glycolytic flux are still unknown. The PEP phosphatase activity of TIGAR might represent this shunt activity. Furthermore, the 2PG phosphatase activity of TIGAR might also help to avoid ATP production during glycolysis, since the resulting glycerate could then be reintroduced into glycolysis via glycerate kinase at the expense of ATP [46,47].
Activities calculated for other glycolytic intermediates (PEP and 2PG) in mouse muscle indicate that TIGAR is able to shunt glycolysis at a rate representing, at most, 0.1 μmol/min per g of muscle (assuming saturating concentration of substrates), which represents only a minor fraction of the glycolytic capacity. TIGAR would therefore not be responsible for a large fraction of the glycolytic flux in mouse muscle, the tissue with the highest TIGAR expression. However, we cannot exclude that this activity would become relevant in some cancer cell lines with very low pyruvate kinase activity. It therefore remains possible that part of the cellular phenotype upon TIGAR loss is caused by changes in this shunt activity.
The kinetic properties and the structural similarity of the best substrates of TIGAR make it unlikely that TIGAR modulates cellular F26BP levels directly. The identification of 23BPG as an approximately 400-fold-better substrate (with regard to the kcat/Km value) suggests that the mechanism of action of TIGAR needs to be re-evaluated. It also prompts a reassessment of the role of 23BPG in cellular metabolism outside of red blood cells and downstream of p53 signalling. This adds a novel layer of complexity to the function of TIGAR that should be taken into account in future studies.
Previous studies have suggested that TIGAR might play a role in animal models of colon cancer, myocardial infarction and Parkinson's disease . This indicates that inhibition of TIGAR might represent a potential therapeutic or preventive strategy. The identification of 23BPG as the best substrate will greatly facilitate the identification of inhibitors of TIGAR in assays that are easily adaptable for high-throughput screening .
bacterial artificial chromosome
fructose-2,6-bisphosphatase, PEP, phosphoenolpyruvate
TP53 (tumour protein 53)-induced glycolysis and apoptosis regulator
Isabelle Gerin, Jennifer Bolsée, Emile Van Schaftingen and Guido Bommer designed the study. All authors performed experiments and contributed to data analysis and final revision of the paper, which was written by Isabelle Gerin, Emile Van Schaftingen and Guido Bommer.
We thank Professor Eric Fearon for sharing human tissue RNAs, Philippe Soriano for the plasmid pGK-DTA and Francois Collard for comments on the paper.
This work was supported by TELEVIE (to E.V.S. and G.T.B.), the Fondation Contre le Cancer (to E.V.S. and G.T.B.), the Fonds National de la Recherche Scientifique (FNRS) (to E.V.S. and G.T.B.), WELBIO (to E.V.S.) and the Fonds spéciaux de recherche of the Université Catholique de Louvain (to G.T.B.). G.T.B. is a chercheur qualifié of the Fonds National de la Recherche Scientifique and J.B. is a recipient of a Fonds pour la formation à la recherche dans l'industrie et dans l'agriculture (FRIA) fellowship of the Fonds National de la Recherche Scientifique.