Comparison of tumour hypoxia measured by FMISO-PET and gene signatures for patients with HNSCC

Purpose: Tumour hypoxia is well known to increase radio-resistance of tumours. In a recent prospective biomarker imaging trial, hypoxia has been measured by [18F]fluoromisonidazole positron emission tomography (FMISO-PET) scans [1,2]. Here, we compared hypoxia imaging with the expression of hypoxia-associated gene signatures for patients with locally advanced head and neck squamous cell carcinoma (HNSCC) treated by primary radiochemotherapy (RCHT).

Material and methods: FMISO-PET imaging and gene expression analyses were performed on the cohort of 50 HNSCC patients [1,2]. For this study, the FMISO-PET parameters tumour-to-background ratio (TBPpeak) and hypoxic tumour volume (HV1.6) analysed before RCHT were considered. Expressions of 15-, 26- and 30-gene hypoxia-associated signatures [3-5] were analysed from formalin-fixed paraffin-embedded (FFPE) tumour biopsies obtained before RCHT using the GeneChip® Human Transcriptome Array 2.0 (Affymetrix) and nanoString analysis. Gene expressions were compared between the two methods using the Pearson correlation coefficient. Linear regression was applied to relate TBRpeak and HV1.6 to the mean expression of the gene signatures, including the interaction with tumour volume which was assessed on the planning CT by an experienced radiation oncologist. The association of FMISO-PET parameters and gene expressions to loco-regional control (LRC) and progression-free survival (PFS) was assessed by Cox regression.

Results: The mean expressions of all hypoxia-associated gene signatures were highly correlated between Affymetrix and nanoString analyses (R>0.5). While TBRpeak and HV1.6 were weakly correlated with the expression of hypoxia-associated genes alone, significant correlations were observed if the interaction term of gene expression and tumour volume was included (R>0.5). Both FMISO-PET parameters were significantly correlated with LRC and PFS (p<0.01), while the combination of hypoxia-associated gene expressions and their interaction with tumour volume showed a significant but weaker correlation for the 30-gene signature to LRC and for the 15- and 30-gene signature to PFS (p<0.05). The figure shows patient stratifications using HV1.6 (p=0.02), the 30-gene signature (p=0.07) and their combination (p<0.01).

Conclusion: Hypoxia imaging correlates to the expression of hypoxia-associated genes if the interaction of gene expression and tumour volume is included. Interestingly, both methods may complement each other, which may be of advantage for identifying patients who are at high risk of treatment failure and may benefit from dose escalation. While FMISO-PET directly measures hypoxia, the gene signatures are also associated with other radiobiologic phenomena such stemness of cancer cells.