Tumours ferment glucose to lactate even in the presence of oxygen (aerobic glycolysis; Warburg effect). The pentose phosphate pathway (PPP) allows glucose conversion to ribose for nucleic acid synthesis and glucose degradation to lactate. The nonoxidative part of the PPP is controlled by transketolase enzyme reactions. We have detected upregulation of a mutated transketolase transcript (TKTL1) in human malignancies, whereas transketolase (TKT) and transketolase-like-2 (TKTL2) transcripts were not upregulated. Strong TKTL1 protein expression was correlated to invasive colon and urothelial tumours and to poor patients outcome. TKTL1 encodes a transketolase with unusual enzymatic properties, which are likely to be caused by the internal deletion of conserved residues. We propose that TKTL1 upregulation in tumours leads to enhanced, oxygen-independent glucose usage and a lactatebased matrix degradation. As inhibition of transketolase enzyme reactions suppresses tumour growth and metastasis, TKTL1 could be the relevant target for novel anti-transketolase cancer therapies. We suggest an individualised cancer therapy based on the determination of metabolic changes in tumours that might enable the targeted inhibition of invasion and metastasis.

Cancer is now viewed as a disease resulting from cancer-causing genes that deregulate cellular proliferation, differentiation, and death. Genetic alterations acquired by tumours also modify their biochemical pathways, resulting in abnormal metabolism. Warburg proposed a model of tumorigenesis involving altered energy production in tumours. He identified a particular metabolic pathway in carcinomas characterised by the anaerobic degradation of glucose even in the presence of oxygen (aerobic glycolysis) that leads to the production of large amounts of lactate (known as the Warburg effect; Warburg et al, 1924). The relevance of aerobic glycolysis to cancer cell biology remains controversial (Garber, 2004; Gatenby and Gillies, 2004). However, the widespread clinical use of positron-emission tomography (PET) for the detection of aerobic glycolysis in tumours and recent findings have rekindled interest in Warburg’s theory. Studies on the physiological changes in malignant conversion provided a metabolic signature for the different stages of tumorigenesis (Ramanathan et al, 2005); during tumorigenesis, an increase in glucose uptake and lactate production have been detected. The fully transformed state is most dependent on aerobic glycolysis and least dependent on the mitochondrial machinery for ATP synthesis (Ramanathan et al, 2005). Other important links between cancer-causing genes and glucose metabolism have been already identified. Activation of the oncogenic kinase Akt has been shown to stimulate glucose uptake and metabolism in cancer cells and renders these cells susceptible to death in response to glucose withdrawal (Elstrom et al, 2004).

 Such tumour cells have been shown to be dependent on glucose because the ability to induce fatty acid oxidation in response to glucose deprivation is impaired by activated Akt (Buzzai et al, 2005). In addition, AMP-activated protein kinase (AMPK) has been identified as a link between glucose metabolism and the cell cycle, thereby implicating p53 as an essential component of metabolic cell-cycle control (Jones et al, 2005). These findings suggest that tumorigenesis requires derangements in known cancer-causing genes and altered energy production. Despite this appreciation, the reason enhanced anaerobic glucose degradation occurs in tumours remains elusive. If carcinogenesis occurs by somatic evolution, then common components of the cancer phenotype result from active selection, and must, therefore, confer a significant growth advantage (Gatenby and Gillies, 2004). A prerequisite for the understanding of the altered glucose metabolism in tumours and its predicted selective growth advantage is the detailed analysis of glucose degrading pathways. Two main pathways of glucose degradation have been identified. The observation in the 1930s that muscle extracts can catalyse the glycolysis of glucose to lactate led to the identification of the Embden-Meyerhof pathway. In this pathway, fructose-1,6-diphosphate is cleaved leading to pyruvate, which is reduced to lactate in the absence of oxygen. In addition to this pathway, glucose is also degraded by the pentose phosphate pathway (PPP). The nonoxidative part of the PPP is controlled by thiamine- (vitamin B1) dependent transketolase enzyme reactions. 

Transketolase enzyme reactions of the nonoxidative part of the PPP enable oxygen-independent glucose degradation, and play a crucial role in nucleic acid ribose synthesis utilising glucose carbons in tumour cells. More than 85% of ribose recovered from nucleic acids of certain tumour cells is generated directly or indirectly from the nonoxidative pathway of the PPP (Boros et al, 1997). The importance of transketolases for tumour cell metabolism is underlined by the fact that the application of specific transketolase inhibitors to tumours induces a dramatic reduction in tumour cell proliferation (Rais et al, 1999). In addition, the activation of transketolases by application of thiamine stimulates tumour growth (Comin-Anduix et al, 2001). Furthermore, several natural products have been reported to inhibit transketolase enzyme activity in vitro, and also to inhibit cell proliferation or suppress tumour growth in mouse models or cancer patients as a result of reduction of transketolase activity (Hidvegi et al, 1999; Boros et al, 2001a, b; Comin-Anduix et al, 2002; Jakab et al, 2003). To establish a transketolase-inhibitory anticancer therapy, a high throughput screening of library compounds using recombinant human transketolase (TKT) has been performed. This approach resulted in the identification of two novel small-molecule inhibitors, which inhibit human TKT and suppress proliferation of cancer cell lines (Du et al, 2004). So far, three human transketolase genes have been recognised, and the relative contributions of TKT, transketolase-like-1 (TKTL1), and transketolase-like-2 (TKTL2) to tumour-specific transketolase metabolism have not been investigated.

At the time of surgery, 14 % (8 out of 59) of the patients had lymph node metastases. Overall, 28 tumours were classified as nonmuscle-invasive (pTa, pT1 and carcinoma in situ) and 31 were classified as muscle-invasive (XpT2). Overall, 42% of the tumours showed no or only weak staining (staining score 0 or 1), 10% showed some staining (score 2), and 48% showed strong staining (score 3). Patient samples Surgical resection specimens were obtained from the Department of Surgery, University Hospital Mannheim, Faculty of Clinical Medicine of Ruprecht-Karls-University Heidelberg, Germany (approval by the local Ethics Committee), and the University of Palermo, Department of Surgical and Oncological Science, Surgery Pathophysiology Section. None of the patients received neoadjuvant radiotherapy or chemotherapy. From each patient, cancerous and normal tissue was available. For RNA extraction, the specimens taken during the operation were immediately snap-frozen in liquid nitrogen and subsequently stored at
801C until use. For immunohistochemistry (IHC), the specimens were fixed in 3.4% buffered formalin for 24 h and embedded in paraffin. Histological diagnosis was performed by three independent, experienced pathologists (A.z.H.; R.G.; G.S.). 

Immunohistochemical staining Three to 5 mm thick paraffin sections were analysed by IHC. Dewaxed sections were heated for antigen unmasking in 10 mM sodium citrate (pH 6.0) in a microwave oven for 1 min at 450 W followed by 5 min at 100 W. After rinsing in dH2O, inhibition of endogenous peroxidase was performed with a 5 min incubation with 3% H2O2. Endogenous avidin-biotin was blocked by the use of a commercial biotin blocking system (DAKO) for 10 min. After two washes in Tris/saline buffer (TBS), slides were incubated with 1% goat serum for 30 min to block unspecific staining. Sections were subsequently exposed to mouse anti-TKTL1 (clone JFC12T10; mouse IgG2b) antibody (15 mg ml
1 ) or anti-Ser473 phospho-Akt (587F11; mouse IgG2b; Cell Signaling Technology) overnight at 41C. The monoclonal anti-TKTL1 antibody JFC12T10 has been described previously (Coy et al, 2005). Slides were washed in TBS and incubated with biotinylated anti-mouse immunoglobulins for 30 min at room temperature and treated with streptavidin-peroxidase (DAKO). Staining was revealed using 3-amino9-ethylcarbazole (AEC) substrate and counter-stained with haematoxylin (Figure 2M– T and Figure 3). Alternatively, primary antibodies were visualised with avidinbiotinylated horseradish peroxidase complex (ABC) and diaminobenzidine tetrahydrochloride (DAB) (Elite kit; Vector Laboratories), and counter-stained with Mayer’s haematoxylin (Figure 2A–L).

 For scoring of TKTL1 expression, a scale from 0 to 3 was defined as: score 0 indicates 0 –20%, score 1 indicates 21– 50%, score 2 indicates 51–80%, and score 3 indicates 480% of the tumour cells were stained for TKTL1. RESULTS TKTL1, but not TKT or TKTL2, mRNA is overexpressed in carcinomas Identification of genes selectively expressed or overexpressed in tumours is a crucial prerequisite for molecular diagnosis and such targets, we used the real-time PCR technique. When comparing the transcript levels from five colon cancer tissues to nontumour samples from the same patients, we initially detected a 35-fold overexpression of TKTL1 in one colon carcinoma sample. As TKTL1 is one of three highly similar transketolases encoded by three separate genes (TKT, TKTL1, TKTL2; Coy et al, 1996, 2005), we designed primers to specifically discriminate expression of the three transketolase genes in human carcinomas. Using these primers, a 79-fold overexpression of the TKTL1 gene was identified in one colon carcinoma tissue, whereas none of the tested colon carcinomas showed an overexpression of the TKT transcript. In contrast, the TKTL2 transcript was downregulated more than 10-fold in three out of five colon carcinomas (Table 1). To test whether overexpression of the TKTL1 transcript occurs in other tumour types, cDNA from five gastric and five lung adenocarcinomas and their corresponding normal tissues were analysed using the real-time PCR technique. Two of five gastric carcinomas and two of five lung adenocarcinomas had greater than 10-fold overexpression of TKTL1 (Figure 1A), whereas TKT expression was unchanged in all tested carcinoma tissues (not shown). 

Similar to its downregulation in colon carcinomas, TKTL2 expression was downregulated more than 10-fold in two of five lung adenocarcinomas. In the colon carcinomas depicted in Table 1, the total amount of transketolase transcripts in four of five carcinoma tissues (Table 1; T2– T5) was lower than that in normal tissue, even in the colon carcinoma tissue with an overexpression of TKTL1 (Table 1; T5). Overexpression of TKT was not detected in any of the 54 carcinoma tissue samples tested. The only transketolase gene overexpressed in carcinoma tissue was the TKTL1 gene. TKTL1 protein is overexpressed in human carcinomas In order to determine TKTL1 protein expression levels in human carcinomas, we performed IHC on 1030 human carcinomas derived from 16 different epithelial tumour entities using a monoclonal antibody (JFC12T10) that specifically detects TKTL1 (Coy et al, 2005). A gastric carcinoma specimen that we found to have 1000-fold overexpression of TKTL1 mRNA when compared to the corresponding normal tissue showed a strong overexpression of the TKTL1 protein on Western blot level (Figure 1B) as well as a strong TKTL1 immunoreactivity on paraffin sections (Figure 2C –G). Mainly cytoplasmic expression was detected. TKTL1 expression was restricted to tumour cells, and the surrounding stromal tissue showed no staining. In the corresponding normal tissue, no staining was detected (Figure 2A and B). Immunohistochemical analysis of two gastric carcinoma samples without overexpression of TKTL1 transcript did not reveal TKTL1 protein expression. These results demonstrate a strong correlation of TKTL1 mRNA and protein expression. In some cases of undifferentiated gastric carcinoma, strong nuclear expression was observed (Figure 2H and I). 

Analysis of bladder carcinomas showed absence of TKTL1 reactivity in superficial, nonmuscle-invading tumours (Figure 2J), whereas invasive tumours showed immunoreactivity (Figure 2K and L). Non-small-cell lung carcinomas (NSCLC) (Figure 2M), breast carcinomas (Figure 2N), follicular thyroid carcinomas (FTC) (Figure 2O), papillary thyroid carcinomas (PTC) (Figure 2P), prostate carcinomas (Figure 2Q), pancreas carcinomas (Figure 2R; and undifferentiated thyroid (UTC), ovarian, cervix, rectal, and kidney carcinomas (not shown) also showed strong upregulation of TKTL1. Similar to bladder carcinomas, no or weak reactivity for TKTL1 was observed in noninvasive colon carcinomas (Figure 2S), whereas in invasive tumours, strong TKTL1 staining was detected (Figure 2T). All histological variants of thyroid cancer (FTC, Figure 2O; PTC, Figure 2P; UTC, not shown) revealed abundant TKTL1 expression within the cytoplasm. Interestingly, the majority of nuclei were stained in FTC (Figure 2O), whereas in PTC (Figure 2P), and UTC, (not shown) only few nuclei were TKTL1- positive. Non-small-cell lung cancer cells were strongly positive for TKTL1 within the cytoplasm, whereas nuclear staining was absent (Figure 2M).Immunohistochemical localisation of TKTL1 protein and activated Akt Recent studies have demonstrated that activated Akt exerts a direct influence on glucose metabolism leading to a dose-dependent stimulation of aerobic glycolysis (Elstrom et al, 2004), and to an inhibition of b-oxidation of fatty acids (Buzzai et al, 2005). Therefore, we investigated whether tumours that overexpress the TKTL1 protein also have activated Akt (phospho-Akt; p-Akt) indicative of an activated glucose metabolism and an inhibited b-oxidation of fatty acids.

 We performed IHC to detect p-Akt in thyroid, lung, colon, bladder, and prostate cancer specimens (Figure 3). All the different types of cancer examined showed strong staining for p-Akt in the cytoplasm, nucleus, and both the cytoplasm and nucleus, whereas normal tissues showed no or very weak staining (Figure 3A). In lung, colon, and prostate carcinomas, mainly nuclear localisation of p-Akt was seen (Figure 3B, F and H; yellow arrowheads), whereas in all the histological variants of thyroid cancer, and bladder cancer, strong cytoplasmic staining was seen (Figure 3C– E and G). Only few nuclei in PTC and FTC samples were positive for p-Akt (Figure 3C and D, yellow arrowheads). Activated Akt was detected in 69% of carcinomas examined, whereas 83% of tested carcinoma specimens showed overexpression of TKTL1.TKTL1 protein is overexpressed in invasive colon carcinomas and is associated with poor patient survival In order to study the correlation of TKTL1 expression on clinical– pathological parameters, a retrospective survey of surgical samples from 70 patients (55 men and 15 women, median age of 60715 years) with colon adenocarcinoma was performed. Non-neoplastic colon tissues examined did not reveal TKTL1 expression, and noninvasive colon cancer specimens were negative or barely positive for TKTL1 staining (Figure 2S). In contrast, all invasive colon carcinomas showed TKTL1 expression (Figure 2T), in particular, 20% were classified as weakly positive for TKTL1 (score 1 þ ), 25% as positive (score 2 þ ), and 55% as strongly positive (score 3 þ ). A significant correlation between TKTL1 expression and survival was found (Figure 4A).