The human transketolase-like proteins TKTL1 and TKTL2 are bona fide transketolases
Three transketolase genes have been identified in the human genome to date: transketolase (TKT), transketolase-like 1 (TKTL1) and transketolase-like 2 (TKTL2). Altered TKT functionality is strongly implicated in the development of diabetes and various cancers, thus offering possible therapeutic utility. It will be of great value to know whether TKTL1 and TKTL2 are, similarly, potential therapeutic targets. However, it remains unclear whether TKTL1 and TKTL2 are functional transketolases.
Homology modelling of TKTL1 and TKTL2 using TKT as template, revealed that both TKTL1 and TKTL2 could assume a folded structure like TKT. TKTL1/2 presented a cleft of suitable dimensions between the homodimer surfaces that could accommodate the co-factor-substrate. An appropriate cavity and a hydrophobic nodule were also present in TKTL1/2, into which the diphosphate group fitted, and that was implicated in aminopyrimidine and thiazole ring binding in TKT, respectively. The presence of several identical residues at structurally equivalent positions in TKTL1/2 and TKT identified a network of interactions between the protein and co-factor-substrate, suggesting the functional fidelity of TKTL1/2 as transketolases.
Our data support the hypothesis that TKTL1 and TKTL2 are functional transketolases and represent novel therapeutic targets for diabetes and cancer.
KeywordsTransketolase Transketolase-like 1 Transketolase-like 2 Thiamine Pentose phosphate pathway Glycolysis
Nicotinamide adenine dinucleotide phosphate
Non-oxidative glucose pathway
Pentose phosphate pathway
Three transketolase genes have been identified in the human genome to date. Transketolase (EC 126.96.36.199; TKT) is a crucial enzyme that links the pentose phosphate pathway (PPP), a non-oxidative glucose pathway (NOGP), to glycolysis. The TKT gene is located on chromosome 3 at position 3p21.1 and is responsible for generating sugar phosphates for intracellular nucleotide metabolism as well as for the production of nicotinamide adenine dinucleotide phosphate (NADPH), a reducing agent and anti-oxidant . TKT is a homodimer and known thiamine diphosphate (TDP)-dependent enzyme that possesses two active sites located at the monomer contact surfaces [2, 3]. In addition to TKT, transketolase-like 1 (TKTL1) and transketolase-like 2 (TKTL2) loci are found on the X chromosome at Xq28 and on chromosome 4 at 4q32.2, respectively .
An altered function of transketolase is linked to various pathophysiologic complications, suggesting TKT as a potential therapeutic target. TKT is the best studied transketolase, and previous studies reported an altered activity in patients suffering from diabetes and various cancers [3, 4, 5]. For example, hyperglycemic individuals display lowered TKT activity that may be ameliorated by thiamine treatment, thus offering potential as a novel treatment for type 2 diabetes . TKT pathways are also intricately involved in cancer progression and metastasis. For example, PPP activation satisfies the high demand by cancer cells for nucleotides (to ensure proliferation) by the terminal conversion of glucose to ribose and conversion to lactate. Moreover, TKT inhibition by oxythiamine (a thiamine antagonist) substantially decreased pancreatic cancer cell growth . Taken together, these studies strongly implicate perturbed TKT function in the development of diabetes and various cancers .
What about TKTL1 and TKTL2? There is limited information regarding the role of TKTL2; more is known about TKTL1. TKTL1, previously postulated to be a pseudogene, encodes a transketolase like-protein that is also linked to cancer . For example, TKTL1 mutations are associated with cancers, making it a promising target for anti-cancer treatments [6, 10]. TKTL1 silencing in colon cancer cells attenuated cell proliferation, and significantly decreased TKT activity, suggesting an interplay between the two transketolases . However, some researchers found that this relationship was experimentally variable .
Human TKT and TKTL1 differ in both primary structure and in amino acid composition , and several groups have raised doubt whether TKTL1 indeed was an actual transketolase [14, 15]. There are no experimental data confirming the enzymatic activity of TKTL1 compared to TKT. TKTL1 has 38 amino acids less in the active site compared to TKT, and this may affect its enzymatic activity. TKTL1 also lacks two vital histidine residues that are otherwise conserved in transketolases, and are required for catalytic processes . There is also a substitution mutation (W124S) in the TKTL1 amino acid sequence. Moreover, the functional identity of TKTL1 and TKTL2 as possible transketolases was based only on an analysis of the one-dimensional sequence alignment of these proteins with the structure of human TKT . A significant limitation of such an approach is that it could miss the identification of alternative residues and possible interactions that may be involved in co-factor and substrate binding.
In light of conflicting findings and uncertainties, the current study investigated the possible function of TKTL1 and TKTL2 as putative transketolases. We employed homology modelling and structural analysis of TKTL1 and TKTL2 using the 0.97 Å crystal structure of TKT in complex with a TDP co-factor-substrate as the template .
Materials and methods
We performed a PHI-BLAST search and identified human TKT (UniProtKB P29401) as an appropriate template for modelling (Fig. 1). A co-crystal structure of this protein in complex with ketose D-xylulose-5-phosphate (X5P) is available at a resolution of 0.97 Å (PDB 4KXW) . We generated multiple homology models of TKTL1 and TKTL2 using this crystal structure and the Robetta  and SwissProt servers . Although all models were energy minimized, we did not proceed with molecular dynamics simulations, thinking that an additional predictive refinement would contribute little to our primary structural analysis.
We verified the TKTL1 and TKTL2 model qualities using MolProbity  and Verify3D  available on the SAVE server (services.mbi.ucla.edu/SAVES). The Verify3D method calculates a 3D-1D score for the presence of a given residue within its environment in the calculated model and reports the average score in a sliding 21-residue window. For the TKTL1 model, 96% of the residues have a 3D-1D score > 0.2, where Verify3D defines a score of > 0.2 for more than 80% of the residues as acceptable. The Ramachandran plot showed that 98.7% (585/593) of all residues were in allowed regions, 99.8% of all rotamers were favored and no bad rotamers were present in the model. In the TKTL2 model, 93% of residues had a 3D-1D score > 0.2, 99.4% (625/626) of residues had ϕ and ψ angles within the allowed regions, no bad rotamers were present and 98.2% of the rotamers were favored. By comparison, in the crystal structure of TKT , 92% of residues had a 3D-1D score > 0.2, 99.7% of residues exhibited favored ϕ and ψ angles, 94% were favored rotamers and 3% bad rotamers were present, emphasizing the quality of our proposed models.
Fitting the TDP substrate ligand in the TKTL1 and TKTL2 models
Homology models of TKTL1 and TKTL2 show substrate clefts
The binding of the TDP substrate and substrate adduct in the active site of TKT has been pursued by a number of previous studies [13, 16, 17]. For the current study, TDP fits in a channel formed at the interface surfaces of the homodimer subunits (Fig. 2c). When viewing the position of the aligned TDP ligand in the structure of the modeled TKTL1, the interface surface between the homodimer subunits similarly provides a crevice into which the TDP fits (Fig. 2d; Figure S1B shows a similar crevice in TKTL2). TKTL1 provides a surface architecture, distribution of polar/hydrophobic regions and spatial placement of residue sidechains very similar to that found in the TKT crystal structure. This allows for the stabilization of the bound TDP in a similar orientation relative to key surface elements, and the two identified active site residues.
Identical and similar hydrogen-bonds stabilize the TDP diphosphate in TKTL1/2
Interactions with the aminopyrimidine and thiazole rings
Binding of the substrate in TKTL1/2
We have constructed homology models for TKTL1 and TKTL2 using the crystal structure of human TKT in complex with a reaction intermediate of the TDP co-factor as a structural template. This allows a more rigorous assessment of the possible enzyme-co-factor and enzyme-substrate interactions compared to a simple alignment of sequences. When employing superposition of models with the TKT crystal structure to position the co-factor adduct in the models, we have shown that a similar co-factor and substrate crevice is formed at the interface between the two sub-units of the homodimer in the TKTL1 and TKTL2 models (Fig. 2c and d and Figure S1B). This location of the co-factor adduct further allowed the required insertion of the TDP diphosphate tail into a cavity, where numerous conserved and new interactions were possible that anchored TDP to the protein (Fig. 2e and f and Figure S1B, Fig. 5 and Figure S2A). The environment and spatial distribution of residues similarly allowed mostly conserved interactions in TKTL1 and TKTL2 with the aminopyrimidine and thiazole rings (Fig. 4 and Figure S2B). The residues that may be involved in interactions with the TDP co-factor and X5PA intermediate in TKT  and the equivalent residues identified here in the TKTL1 and TKTL2 models are summarized in Fig. 6.
However, will the absence of H110 in TKTL1 abolish its transketolase activity? It appears that alternative active residue patterns are possible in transketolases, suggesting some structural and functional degeneracy. For example, yeast transketolase utilizes H481 and H103 to bind to a water molecule, through which the thiazole C2 proton is indirectly abstracted. Mutation of the H481 residue abolished enzyme activity . Intriguingly, in human TKT a glutamine residue (Q428) is present at the position equivalent to yeast H481 (Fig. 1). Mutation of Q428 resulted in a 3- to 4-fold reduction, but not a loss of TKT activity . However, the mutation of H110 in human TKT (equivalent to yeast H103) resulted in a 50-fold reduction in activity, effectively abolishing enzyme function . Thus, for human TKT the proposal is made that H110 and H77 pair functioned like the yeast H481 and H103 pair, where H110 and H481, respectively, acted as ultimate acceptors of the thiazole C2 proton.
Because of its value in biocatalysis, there has been a number of mutational and directed evolution studies of transketolases . Although several interesting mutations were identified that resulted in relaxed substrate specificity and enhanced activity , there was no clear identification of a residue that could act as alternative proton acceptor to H110 or H481. However, the possible mechanistic impact of the absence of H110 and H77 in TKTL1 on possible catalytic activity is far from resolved in the literature. For example, Titmann and colleagues reported that deletion of a 38 residue region in TKT mimicked the observed deletion found in TKTL1 (containing H110 and H77) and abolished transketolase activity . However, the recombinant deletion protein was not properly folded and it did not dimerize efficiently. Since dimerization is a requirement for the formation of a functional active site, where both sub-units contribute essential residues, it is not clear that this study provides any insights into TKTL1 functionality. A related study also reported the absence of transketolase activity for a recombinant human transketolase (38 residues deleted). However, other than noting that the protein was recovered from inclusion bodies in the bacterial hosts, this study did not assess folding and dimerization of the overexpressed protein .
Transketolase catalyzes the transfer of a 2-carbon ketose from a donor such as X5P to an acceptor such as ribose-5-P to produce sedoheptulose-7-P and glyceraldehyde-3-P in a classic ping-pong-bi-bi two-substrate reaction. An alternative reaction is the one-substrate reaction, where the ketose derived from the X5P donor dissociates from the active site to produce glycolaldehyde, which then acts as acceptor for another ketose to produce erythrose . Some reported that the two-substrate reaction was decreased by two-fold in a H103A mutant yeast protein . Of note, the one-substrate X5P reaction was enhanced by ~ two-fold to a level similar to the two-substrate reaction in the H103A mutant protein . However, in yeast the H481 proton sink is still present in the H103 mutant. This is unlike the human enzyme where the structural homolog of H103 is H110, the functional equivalent of yeast H481.
Studies that expressed recombinant TKTL1 (versus TKT deletion mutants to mimic TKTL1) or that assayed endogenous TKTL1, reported transketolase activity [6, 32]. Here the recombinant TKTL1 protein exhibited both classic two-substrate activity, utilizing X5P and ribose-5-phosphate, as well as one-substrate specificity, utilizing only X5P . Furthermore, others demonstrated that siRNA knockdown of TKTL1 in cultured human leukemia cells caused a significant decrease in TKT activity that was measured by glyceraldehyde-3-P production in the two-substrate X5P and ribose-5-P reaction . No direct data on possible non-specific knockdown of TKT was provided, but the authors concluded that TKTL1 possesses TKT activity .
In light of these two studies demonstrating TKT activity for recombinant  and endogenous  TKTL1, and that the latter apparently lacks the equivalent of H110, the question naturally arises which residue acts as terminal acceptor to the thiazole C2 proton? An evaluation of the distribution of histidine residues in the region of the thiazole C2 and aminopyrimidine amino groups, shows that H231 in TKTL1 is structurally equivalent to H258 in TKT (Fig. 4a). The τ-nitrogen of H231 (like H258) is also within hydrogen-bond distance of the thiazole sulphur. It is therefore likely that H231 may act as acceptor to the C2 proton to activate the thiazole ring. For TKT the H110 residue is the preferred acceptor and may in addition stabilize the co-factor-substrate intermediate (Fig. 5a). It is thus possible that the absence of H110 in TKTL1 weakens the interaction with the ketose reaction intermediate, allowing for the dissociation of glycolaldehyde and enhancing the one-substrate reaction.
Together these results support our hypothesis that TKTL1 and TKTL2 are functional transketolases and thus opens up the possibility that altered activity/function may be implicated in diseases such as diabetes and cancer. Additional physiological and biochemical studies should therefore be pursued to assess how modulation of such enzymes may be exploited as putative therapeutic interventions for diabetes and cancer.
Availability of data and materials
The homology models for TKTL1 and TKTL2 in PDB format are available as supplementary material.
HGP, GD and MFE designed the research, HGP performed the research and analyzed the data, and HGP, GD and MFE wrote the paper. All authors read and approved the final version of this manuscript.
Ethics approval and consent to participate
Consent for publication
“The authors declare that they have no competing interests.”
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
- 25.Di Tommaso P, Moretti S, Xenarios I, Orobitg M, Montanyola A, Chang JM, Taly JF, Notredame C. T-coffee: a web server for the multiple sequence alignment of protein and RNA sequences using structural information and homology extension. Nucleic Acids Res. 2011;39(Web Server issue):W13–7.CrossRefGoogle Scholar
- 32.Diaz-Moralli S, Aguilar E, Marin S, Coy JF, Dewerchin M, Antoniewicz MR, Meca-Cortes O, Notebaert L, Ghesquiere B, Eelen G, et al. A key role for transketolase-like 1 in tumor metabolic reprogramming. Oncotarget. 2016.Google Scholar
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.