The hypoxia‐responsiveness of the glycolytic machinery may allow pretreatment identification of hypoxic tumors from HIF‐1 targets (e.g., Glut‐1) or [18F]‐fluorodeoxyglucose positron emission tomography but results have been mixed. We hypothesized that this discrepancy is an inevitable consequence of elevated aerobic glycolysis in tumors (Warburg effect) as energetics in predominantly glycolytic cells is little affected by hypoxia. Accordingly, we characterized glycolytic and mitochondrial ATP generation in normoxic and anoxic cell lines. Measurements demonstrated that most cancer cells rely largely on aerobic glycolysis as it accounts for 56–63% of their ATP budget, but in the cervical carcinoma SiHa, ATP synthesis was mainly mitochondrial. Moreover, the stimulatory effect of anoxia on glycolytic flux was inversely correlated to the relative reliance on aerobic glycolysis. Next, tumor cells representing a Warburg or a nonglycolytic phenotype were grown in mice and spatial patterns of hypoxia (pimonidazole‐stained), Glut‐1 expression and ¹⁸F‐FDG uptake were analysed on sectioned tumors. Only in SiHa tumors did foci of elevated glucose metabolism consistently colocalize with regions of hypoxia and elevated Glut‐1 expression. In contrast, spatial patterns of Glut‐1 and pimonidazole staining correlated reasonably well in all tumors. In conclusion, Glut‐1's value as a hypoxia marker is not severely restricted by aerobic glycolysis. In contrast, the specificity of ¹⁸F‐FDG uptake and Glut‐1 expression as markers of regional hypoxia and glucose metabolism, respectively, scales inversely with the intensity of the Warburg effect. This linkage suggests that multi‐tracer imaging combining FDG and hypoxia‐specific markers may provide therapeutically relevant information on tumor energetic phenotypes.

Tumor hypoxia adversely affects treatment outcome, especially in squamous cell carcinomas (SCCs). Image guided radiotherapy (IGRT) based on PET-generated tumor hypoxia maps allows dose boosting to hypoxic sub-volumes and has received considerable interest. However, the combination of slow oxygenation-dependent tracer retention, slow clearance of unbound tracer from non-hypoxic tissue and the necessity to average signal over large non-homogenous tissue areas due to the low PET resolution remains problematic. To assess pitfalls inherent to low-resolution imaging we have analyzed the fine-scale distribution of a PET hypoxia tracer (autoradiograms) and tissue architecture (immunofluorescence microscopy) in sectioned experimental SCCs, and compared the results to those obtained when applying macroscopic averaging mimicking the resolution in clinical PET scanners. We show that tumor areas that would be classified as non-hypoxic based on simple PET threshold identification, often contains foci of hypoxic cells, in particular in tumors where necrosis and severely hypoxic cells are intermixed. In contrast, in a non-necrotic tumor model we found that the risk of missing hypoxic cells was greatly reduced, however, its patchy hypoxic pattern made a clear delineation of a target to boost unfeasible. We discuss the implications of these and other complicating factors in PET hypoxia-imaging and outline future strategies to overcome or circumvent them.