Two isoforms of TALDO1 generated by alternative translational initiation show differential nucleocytoplasmic distribution to regulate the global metabolic network

Transaldolase 1 (TALDO1) is a rate-limiting enzyme involved in the pentose phosphate pathway, which is traditionally thought to occur in the cytoplasm. In this study, we found that the gene TALDO1 has two translational initiation sites, generating two isoforms that differ by the presence of the first 10 N-terminal amino acids. Notably, the long and short isoforms were differentially localised to the cell nucleus and cytoplasm, respectively. Pull-down and in vitro transport assays showed that the long isoform, unlike the short one, binds to importin α and is actively transported into the nucleus in an importin α/β-dependent manner, demonstrating that the 10 N-terminal amino acids are essential for its nuclear localisation. Additionally, we found that these two isoforms can form homo- and/or hetero-dimers with different localisation dynamics. A metabolite analysis revealed that the subcellular localisation of TALDO1 is not crucial for its activity in the pentose phosphate pathway. However, the expression of these two isoforms differentially affected the levels of various metabolites, including components of the tricarboxylic acid cycle, nucleotides, and sugars. These results demonstrate that the nucleocytoplasmic distribution of TALDO1, modulated via alternative translational initiation and dimer formation, plays an important role in a wide range of metabolic networks.


Results
TALDO1 has two translation initiation sites. When we performed western blotting using an anti-TALDO1 antibody, we noticed that the antibody detected two bands around 37 kDa (the estimated molecular weight of TALDO1 is 37.39 kDa) in cell lysates prepared from mouse and human cell lines (Fig. 1A). The expression levels and the ratio of the two bands varied among the cell lines. To confirm whether these two bands were TALDO1 proteins, we produced TALDO1 knockout (KO) mouse NIH/3T3 cells using CRISPR/Cas9-mediated genome engineering (Supplementary Figure 1). As shown in Fig. 1A, lane 2, neither of these bands were detected in the lysates of TALDO1 KO cells, confirming that they were TALDO1 proteins. Additionally, immunofluorescence analysis revealed the loss of nuclear staining of TALDO1 in the TALDO1 KO cells.
Notably, two potential in-frame translation initiation sites denoted as sites A and B exist in mouse Taldo1 mRNA (Fig. 1B), and these sites are highly conserved in higher eukaryotes (Fig. 1C). It is well known that the sequence flanking the AUG initiation codon (Kozak sequence) plays an important role in the enhancement of translational initiation by eukaryotic ribosomes 16 . Furthermore, the initiation of alternative translation is, in some cases, dependent on the strength of the codon-anticodon interaction 17 . Indeed, we found that the Kozak sequence within site A of TALDO1 contains a U instead of a G at position +4, suggesting that it is not optimal for efficient initiation. In contrast, site B contains a perfect Kozak consensus sequence. Furthermore, TALDO1 was listed as a protein that potentially contains an alternative translational initiation site in previous reports 18,19 . Taken together, these observations suggest that an alternative translation of TALDO1 starts from the second methionine at amino acid (A.A.) position 11, causing the generation of a short isoform of TALDO1 lacking the first 10 N-terminal amino acids.
To corroborate this hypothesis, a C-terminal triple haemagglutinin (3×HA)-tagged TALDO1 containing an intact 5ʹ untranslated region (UTR) (native 5ʹ UTR TALDO1-3×HA) was expressed in wild-type NIH/3T3 cells, and immunoblotting was performed using an anti-HA antibody. As shown in Fig. 1E (first lane), two bands, which likely resulted from the alternative translational initiation, were observed. Next, when the N-terminal 3×HA-tagged full-length TALDO1 was expressed; it was detected as a single protein (Fig. 1E, right panel), demonstrating that TALDO1 is not truncated by post-translational modification or proteolysis. Furthermore, the mutant lacking the first 10 amino acids of TALDO1  was also detected as a single protein. These results strongly suggest that TALDO1 has two translational initiation sites, generating two distinct isoforms.
To confirm this theory, we constructed three single alanine substituted mutants, which contained the substitution at the first methionine codon (M1A), the second methionine codon (M11A), or the optimised first Kozak sequence (S2A) (Fig. 1D), each containing a C-terminal 3×HA tag, and the expression of each mutant was analysed by immunoblotting. As shown in Fig. 1E (left panel), the M1A mutation diminished the upper band and enhanced the expression of the lower band, while the M11A mutation abolished the lower band. The optimised first Kozak sequence (S2A) enhanced the expression of the upper band and nearly abolished the lower band. Together, these data indicate that the expression of the two different TALDO1 isoforms, a longer isoform, which we named TALDO1L, and a shorter one lacking the first 10 amino acids, named TALDO1S, is mediated by alternative translational initiation.
The first 10 N-terminal amino acids of TALDO1 are essential for its nuclear localisation. Next, we determined the subcellular localisation of the two isoforms, 3×HA-TALDO1L and 3×HA-TALDO1S. To exclude the effect of the endogenous TALDO1, we transfected the isoform-specific expression vectors into TALDO1 KO cells. Surprisingly, these two isoforms showed different subcellular localisations; while 3×HA-TALDO1L was predominantly localised in the nucleus, the majority of 3×HA-TALDO1S was localised in the cytoplasm (Fig. 2A). The nuclear import of proteins is generally dependent on the presence of a specific signal sequence, the NLS, which is composed of basic amino acids. To determine whether the basic amino acids within the first 10 amino acids of TALDO1 are essential for its nuclear import, we constructed an AAA mutant, in which Scientific RepoRts | 6:34648 | DOI: 10.1038/srep34648 three basic amino acids within the 10 N-terminal amino acids were converted into alanines (K7A/R8A/R10A), and we transiently expressed it in TALDO1 KO cells. As expected, the AAA mutant was mainly distributed in  (A) Immunofluorescence images of TALDO1 isoforms. TALDO1 KO cells were transiently transfected with vectors expressing N-terminally HA-tagged TALDO1 long isoform (TALDO1L, 1-337 amino acids), TALDO1 short isoform (TALDO1S, 11-337 amino acids), or TALDO1 K7A, R8A, and R10A mutant (TALDO1 AAA mut) and stained with anti-HA antibodies. Quantification of the nuclear-cytoplasmic (N/C) ratio of TALDO1 isoforms and mutants. The ratio was normalised to that of TALDO1L (mean ± S.E.M., n = 23 cells). (B) Importin α subtypes recognize TALDO1L but not TALDO1S. GST pull-down assays were performed using GST, GSTimportin α1 (GST-α1), GST-importin α2 (GST-α2), or GST-importin α4 (GST-α4) with TALDO1-GFP isoforms or GFP (as the negative control). The bound proteins were analysed by western blotting with an anti-GFP antibody. Used GST-fusion proteins (1 μg) were analysed by Coomassie Brilliant Blue staining. Asterisks indicate the full-length GST-fusion proteins. (C) In vitro transport assay revealed that GFP-TALDO1L protein, but not GFP-TALDO1S protein, was transported into the nucleus in the presence of importin α, importin β, Ran-GDP, and an ATP regenerating system. GST-SV40T NLS-GFP was used as a positive control. (D) Immunofluorescence images of TALDO1 KO cells expressing either native 5ʹ UTR-TALDO1-3×HA or M1A-3×HA. The transfected cells were stained as described above. the cytoplasm, similarly to 3×HA-TALDO1S. These results suggest that the first 10 amino acids of TALDO1 are essential for its nuclear localisation.
Our previous study showed that TALDO1L binds to mouse importin α1 (KPNA1) 3 . Therefore, we used a glutathione-S-transferase (GST) pull-down assay to test whether these TALDO1 isoforms are recognised by multiple importin α subtypes. As shown in Fig. 2B, we found that GST-importin α1 and GST-importin α4 (KPNA4) interacted with green fluorescent protein (GFP)-TALDO1L but not with GFP-TALDO1S. In contrast, GST-importin α2 (KPNA2) did not bind to either TALDO1 isoform. To examine whether these protein interactions reflect the nuclear localisation of TALDO1, we performed an in vitro transport assay. A GST-SV40T NLS-GFP fusion protein (GST-SV40T NLS-GFP) containing the NLS of the simian virus 40 large T-antigen was used as a positive control, and was efficiently transported into the nucleus in the presence of importin α and importin β. As shown in Fig. 2C, the nuclear import of GFP-TALDO1L, but not GFP-TALDO1S, was observed in the presence of TALDO1-interacting importin α isoforms together with importin β. Thus, these results indicate that importin α recognises TALDO1L through its 10 N-terminal amino acids, and actively transports it into the nucleus in an importin α/β-dependent manner.
We also analysed the subcellular localisation of native 5ʹ UTR TALDO1-3×HA and M1A-3×HA in TALDO1 KO cells (Fig. 2D). M1A-3×HA was predominantly localised in the cytoplasm, most likely corresponding to the distribution of TALDO1S ( Fig. 2A). In contrast, the subcellular localisation of native 5ʹ UTR TALDO1-3×HA proteins was mostly nuclear. Collectively, these results indicate that the subcellular localisation of TALDO1 is determined by the translation initiation site used.

TALDO1 forms both homo-and hetero-dimers.
We noticed that overexpressed TALDO1S showed different subcellular distributions when expressed in either wild-type NIH/3T3 cells or TALDO1 KO cells (Fig. 3A); 3×HA-TALDO1S was predominantly localised in the cytoplasm in TALDO1 KO cells (as shown in Fig. 2A), whereas it was distributed throughout the nucleus and cytoplasm in wild-type NIH/3T3 cells. In contrast, the localisation of 3×HA-TALDO1L in wild-type NIH/3T3 cells was similar to that in TALDO1 KO cells. Since TALDO1 contains a conserved interface domain to form a dimer 20 , we speculated that endogenous TALDO1, majorly present as the TALDO1L isoform in NIH/3T3 cells, may form a heterodimer with exogenous TALDO1S, altering its subcellular distribution.
Therefore, to address the dimer-forming ability of TALDO1, two isoforms of TALDO1 harbouring different epitope tags (FLAG and HA) were transiently expressed in TALDO1 KO cells, and their interaction was analysed by co-immunoprecipitation using anti-FLAG antibodies. As shown in Fig. 3B, the immunoreactive band of 3×HA-TALDO1L was detected only in the presence, but not in the absence, of triple FLAG (3×FLAG)-tagged TALDO1L, demonstrating dimer formation between 3×HA-TALDO1L and 3×FLAG-TALDO1L (Fig. 3B, lane 6). Moreover, we found that the deletion of the first 10 amino acids did not interfere with the dimerisation potential of TALDO1 (Fig. 3B, lanes 7 and 8). These results suggest that the dimers containing the long isoform, namely TALDO1L homodimer and TALDO1L/TALDO1S heterodimer, may be actively imported into the nucleus.
Our previous study also demonstrated that full-length TALDO1 (TALDO1L) is exported into the cytoplasm by CRM1 (exportin 1) 3 . Using leptomycin B (LMB), a specific inhibitor of CRM1-dependent nuclear export 21 , we examined whether the short isoform of TALDO1 (TALDO1S-3×HA) is exported in a CRM1-dependent manner in wild-type NIH/3T3 cells or TALDO1 KO cells. The endogenous Ran binding protein RanBP1, which is exported in a CRM1-dependent manner, accumulated in nuclei after LMB treatment, as reported previously 22 . As shown in Fig. 3C, LMB treatment caused a drastic nuclear accumulation of TALDO1S-3×HA in wild-type NIH/3T3 cells (Fig. 3C), clearly demonstrating that its nuclear export was mediated by CRM1. In contrast, the cytoplasmic localisation of TALDO1S-3×HA in TALDO1 KO cells was unchanged after LMB treatment, suggesting that TALDO1 dimers do not migrate into the nucleus by passive diffusion. Since TALDO1S-3×HA potentially forms a heterodimer (TALDO1S-3×HA/endogenous TALDO1L) in wild-type NIH/3T3 cells, but not in TALDO1 KO cells, these results suggest that TALDO1 dimers containing the long isoform (TALDO1L/ TALDO1L or TALDO1L/TALDO1S) can shuttle between the nucleus and the cytoplasm, while the homodimer of the short isoforms (TALDO1S/TALDO1S) is localised only in the cytoplasm. TALDO1 has different functions in the nucleus and the cytoplasm. Next, to assess how the differentially localised TALDO1 affects global metabolite levels, we prepared cell lysates from control wild-type NIH/3T3 cells (infected with empty retroviral vector) and TALDO1 KO cells infected with empty, wild-type TALDO1, or TALDO1S expression retroviral vectors (Supplementary Figure 2) and performed a metabolic analysis using high performance liquid chromatography with mass spectrometry (HPLC/MS). As shown in Fig. 4A and Supplementary Table 1, the lack of TALDO1 caused an approximately 6-fold upregulation of S7P, as previously reported 23 , verifying that our CRISPR/Cas9 procedure disrupted the function of TALDO1. The levels of GAP, a substrate of the forward reaction of TALDO1 as S7P, and the surrounding metabolic intermediates (FBP, DHAP, BPG, and 3PGA) were also significantly increased in the TALDO1 KO cells, indicating important roles of TALDO1 in the regulation of these metabolic intermediates of the pentose phosphate pathway. Notably, we found that the introduction of wild-type TALDO1 or TALDO1S induced the reduction of S7P, GAP, and the surrounding metabolic intermediates to a similar level as that observed in wild-type NIH/3T3 cells. Additionally, the level of 6PGA, which increased in TALDO1 KO cells, as described in a previous report 7 , decreased upon the expression of either wild-type TALDO1 or TALDO1S. These findings indicate that TALDO1 could function as a pentose phosphate pathway enzyme in both the nucleus and the cytoplasm.
However, unexpectedly, the analysis of additional metabolites revealed that the subcellular localisation of TALDO1 significantly influenced the levels of a broad range of metabolites in other pathways. In the purine metabolism pathway, IMP, GMP, AMP, inosine, and ADP significantly increased upon the expression of cytoplasm-localised TALDO1S, but not with wild-type TALDO1. Notably, the expression of cytoplasmic TALDO1S also caused an increase in the levels of a broad range of metabolites in the TCA cycle, including acetyl-CoA, succinyl-CoA, succinate, fumarate, malate, glutamate, glutamine, and aspartate. It should be mentioned that the reintroduction of TALDO1 into TALDO1 KO cells only partially rescued the levels of some metabolites (e.g. malate, fumarate, and 2-oxoglutarate), which could be due to the relatively short time-course of the experiment (24 h after infection) or, alternatively, the excess amount of exogenous TALDO1 isoforms as compared with the endogenous one (data not shown). These results strongly suggest that TALDO1 not only functions as a pentose phosphate pathway enzyme, but also affects other metabolite pathways by altering its subcellular localisation between the nucleus and the cytoplasm.

Discussion
In this study, we have demonstrated that two isoforms of TALDO1 proteins, which are generated by alternative translational initiation, differ in their nucleocytoplasmic localisations due to the availability of a functional NLS, and have significantly different effects on a broad range of metabolic pathways including not only the pentose phosphate pathway, but also glycolysis, the TCA cycle, and purine metabolism.
Alternative translational initiation has been reported to affect the nucleocytoplasmic distribution of proteins (e.g. progesterone receptor 24 , FGF2 25 , and the phosphoprotein of rabies virus 26 ) or their biological roles (e.g. c-MYC 27 , BAG-1 28 , SCL 29 , and Orai1 17 ). However, TALDO1 is, to our knowledge, the first example of a protein in which alternative translational initiation affects the levels of such a broad range of metabolites. The two in-frame translation initiation sites of the Taldo1 gene, which are composed of the first AUG (containing a sub-optimal Kozak sequence) and the second AUG (containing a perfect Kozak sequence), are conserved among human, mouse, chicken, and zebrafish, but not in fly, nematode, or yeast (Fig. 1C). This implies that the regulation of the alternative translational initiation of Taldo1 might be important for physiological function in vertebrates. Notably, the two isoforms of the TALDO1 protein were detected in several human cell lines, which showed different ratios of the long and short isoforms (Fig. 1A). Thus, the alternative translational initiation of this gene may be regulated in a cell type-specific manner.
Our metabolic analysis demonstrated that, surprisingly, both nuclear TALDO1L (long isoform) and cytoplasmic TALDO1S (short isoform) function as pentose phosphate pathway enzymes in a similar fashion. Notably, cytoplasmic TALDO1 affected the levels of metabolites involved in glycolysis, purine metabolism, and the TCA cycle to a greater extent than nuclear TALDO1 did (Fig. 4). Although the mechanism by which the cytoplasmic TALDO1 more efficiently affects metabolite levels is unknown, we speculate that cytoplasmic TALDO1 might directly bind to metabolic enzyme(s) or regulator(s) of other metabolic pathways to affect their function. In the future, it will be important to understand the molecular mechanism by which the altered subcellular distribution of the metabolic enzymes affects such a broad range of cellular metabolic pathways.
Additionally, we examined the subcellular localisation of other GFP-fused enzymes of the pentose phosphate pathway (Supplementary Figure 3). As expected, most of the pentose phosphate pathway enzymes were either localised in the cytoplasm (G6PDH, 6PGDH, and RPI) or distributed throughout the cell (6PGL and RPE), except for transketolase (TKT), which was predominantly localised in the nucleus. Thus, TKT could have differential functions depending on its subcellular localisation, like TALDO1.
Consistently, it has been shown that subsets of metabolic enzymes play differential roles depending on their subcellular localisation. For example, while the isoform of fumarase functions as a metabolic enzyme within the TCA cycle in mitochondria 30,31 , it also acts as a DNA damage response protein when recruited into the nucleus, following DNA damage 32 . Another example is the hexokinase (HK) family of enzymes, which catalyse the conversion of glucose to glucose 6-phosphate, the first step of glycolysis. Two HK isoforms, HKI and HKII, show different subcellular localisations 33 . It has been shown that the functional differences between these HK isoforms are mainly attributed to their subcellular localisation; while HKs associated with mitochondria promote glycolysis, those situated in the cytoplasm promote glycogen synthesis 34 . Additionally, aconitase participates in two different metabolic pathways depending on its subcellular localisation 35 . Moreover, it has been reported that pyruvate kinase M2 (PKM2), an isoform of pyruvate kinase involved in aerobic glycolysis 36,37 , is transported into the nucleus when phosphorylated 38 , and functions as a transcriptional coactivator that promotes glycolysis in cancer cells [39][40][41] . These results, together with our observations, indicate the functional importance of the subcellular localisation of metabolic enzymes. In summary, our study has demonstrated that the function of TALDO1, which has critical roles in a broad range of metabolic pathways, is finely regulated by several mechanisms including alternative translational initiation, active nucleocytoplasmic transport, and dimer formation.
For the cloning of TALDO1 KO cell lines using the CRISPR-Cas9 system, target sequences were designed according to the CRISPR Design Tool (http://crispr.mit.edu/, Supplementary Table 2) and the corresponding annealed oligonucleotides were subcloned into the BbsI (BpiI) sites of pSpCas9 (BB)-2A-Puro (Addgene, Cambridge, MA). Retroviral expression vectors for TALDO1 and TALDO1 short isoform were constructed by subcloning the fragments of TALDO1 and TALDO1 short isoform, respectively, into the BamHI and XhoI sites of pMXs (Addgene).
Expression and purification of recombinant proteins. GST  In vitro transport assay. In vitro transport assay was performed as described previously 42 . Briefly, HeLa cells were cultured in a 100-mm dish with an 8-well slide (MP Biomedicals, Inc., Santa Ana, CA). The next day, after washing with ice-cold TB, the cells were permeabilised with 40 μg/mL digitonin (Nacalai Tesque) in TB on ice for 5 min. Then, after a further wash with ice-cold TB, the cells were immersed in ice-cold TB for 10 min. The slide was then immersed in TB containing TALDO1S-GFP, TALDO1L-GFP, or GST-SV40NLS-GFP, and incubated at 30 °C for 30 min. Finally, the cells were rinsed with ice-cold TB and then fixed with 3.7% formaldehyde in TB for 15 min. GFP-tagged proteins were observed with a Zeiss Axioplan 2 fluorescence microscope (Carl Zeiss).

Generation of TALDO1 KO cells using the CRISPR-CAS9 system. TALDO1 target sequence
(64-ACCACCGTGGTGGCCGACAC-83), guide sequence, sgRNA scaffold, and Cas9 nuclease derived from Streptococcus pyogenes were expressed in NIH/3T3 cells by transfecting them with the constructed pSpCas9(BB)-2A-Puro using Lipofectamine ® 2000. Puromycin selection was carried out at 24 h post-transfection. NIH/3T3 cells were cultured in DMEM containing 3 μg/mL puromycin for 3 days, and re-cultured in fresh DMEM until the formation of colonies was observed. Genomic changes were identified using genomic PCR against off-target genes (Wbscr16, Ikzf2, Thsd1, and Ttc28; the primers are listed in Supplementary Table 2), and using DNA sequence against Taldo1, which was subcloned into a mammalian expression vector, to detect two genomic changes in a clonal cell line. We confirmed the KO of TALDO1 using western blotting and immunofluorescence.

Retroviral infection.
Plat-E cells were plated at a density of 3.0 × 10 5 cells in a 100-mm dish and cultured for 24 h. Using Lipofectamine ® 2000, the cells were transfected with pMXs, pMXs-TALDO1, or pMXs-TALDO1 short isoform. The viral supernatants were harvested at 48 h post-transfection and filtered through a 0.45-μm membrane, followed by an incubation with 8 mg/mL hexadimethrine bromide. The mixed media were transferred to the NIH/3T3 or TALDO1 KO cells, which were plated at a density of 1 × 10 6 cells in a 100-mm dish and cultured for 24 h.
Sample preparation for HPLC/MS. The infected NIH/3T3 or TALDO1 KO cells were washed twice with PBS and subsequently harvested using a cell lifter (Corning Inc., Corning, NY), before transferring them into a 2 mL safe-lock tube (Eppendorf, Hamburg, Germany). We immediately centrifuged the tubes at 3000 × g for 3 min at 4 °C and discarded the supernatant. The cell pellets were frozen with liquid nitrogen and stored at −80 °C until lyophilisation.
Scientific RepoRts | 6:34648 | DOI: 10.1038/srep34648 The lyophilised cell pellets were weighed for normalisation and homogenised using a MM 301 mixer mill (Retsch GmbH, Haan, Germany) with zirconium balls at 20 Hz for 1 min. The cell homogenate was resuspended in 1 mL ice-cold extraction buffer (methanol:water:chloroform, 5:2:2) and shaken at 1300 rpm for 30 min at 4 °C. Four hundred microlitres of H 2 O were then added to the cell extract and the mixture was centrifuged at 13000 rpm for 5 min at 4 °C. We filtered the supernatant with a Millex ® -LG 0.2-μm PTFE filter (Millipore, Billerica, MA), followed by concentration to remove the organic solvent. The sample was lyophilised and redissolved in 50 μL distilled water.
HPLC/MS analysis. HPLC/MS was performed on a LCMS-8030+ (Shimadzu, Kyoto, Japan) with an L-column2 ODS (2.1 × 150 mm, 3 μm; Chemicals Evaluation and Research Institute, Oita, Japan). In chromatographic analysis, tributylamine was used as an ion pair reagent and the sample was injected at a 0.3 mL/min flow rate, with a 40 °C column temperature and a total injection volume of 3 μL. The chromatographic gradient separation was achieved by mixing mobile phase A (15 mM tributylamine and 10 mM acetic acid) and mobile phase B (methanol) as follows; 0% B in A, 0-0.5 min; 25% B in A, 0.5-7.5 min; 100% B in A, 7.5-12 min; 100% B in A, 12-13 min; 0% B in A, 13-14 min; and 0% B in A, 14-16 min. Electrospray ionisation was carried out in negative ion mode and mass spectrometric analysis was performed under the following conditions: probe position, +1.5 mm; desolvation lines temperature, 250 °C; nebuliser gas flow, 2 L/min; heat block temperature, 400 °C. Other conditions were determined by the instrument's auto-tuning program.
Using LabSolutions LC/MS (Shimadzu), we detected mass spectral peaks in multiple-reaction-monitoring mode and identified the metabolites based on the retention times of standard samples. The area values of each metabolite were divided by the area value of piperazine-1,4-bis(2-ethanesulphonic acid), which was used as an internal control for baseline correction, and then divided by the weight of the lyophilised cell pellet for normalisation. Multivariate analysis was performed using the Umetrics SIMCA-P+ 12 software program (MKS Data Analysis Solutions, Umeå, Sweden).