A regression model to predict tumor cellularity in patients with prostate cancer and low FDG uptake

Objectives To explore underlying biological processes of low FDG uptake in prostate cancer using gene expression, measured by array chips (over 20,000 transcripts and variants).

Methods We analyzed a U133A array dataset published by YP Wang et al. (GSE8218) that includes 136 prostate samples with documenting percentages of tumor, stroma, hypertrophy, and atrophy. Since glucose transporters (SLC2A), hexokinase (HK), and glucose-6-phosphatase (G6PC) are directly involved in FDG retention in cells, we used tumor percentage as dependent variable.

Results Five of 17 SLC2A transcripts were significantly associated with tumor percentage. The regression model is y_T=Tumor percentage= -96.01 + 0.174*(SLC2A1) + 0.170*(SLC2A8) +0.035*(SLC2A10) -0.016*(SLC2A3) - 0.068*(SLC2A5) with p-value < 0.05 for all 5 genes in the model. The regression model had adjusted R square=0.553. Two of 6 HK transcripts were significantly associated with tumor percentage. In regression model y_T= 25.038 -0.011*(HK1) + 0.144*(HK2), both HK1 and HK2 had p-value<0.002. HK1 and HK2 abundances in this dataset were 2462.01 ± 613.5 and 149.50 ± 58.34. One of 4 G6PC transcripts was found significantly associated with tumor percentage (p<0.001). The regression model y_T= -35.466 + 0.053*. Since the current FDG oncology protocol is approximately 60 min incubation post injection of FDG, we combined three steps related to FDG uptake and retention. In the new model, 4 transcripts were significantly associated with tumor percentage (all p-values < 0.0001). The final regression model y_T= -114.706 + 0.224*(SLC2A1) + 0.026*(SLC2A10) - 0.013*(HK1) + 0.046*(G6PC3) yielded adjusted R square=0.478.

Conclusions 55.3% of the increase in tumor percentage in 136 samples was explained by SLC2A family. SLC2A1 (Glut1) weighs more than other variable. For the combination, 47.8% of the variation is explained by SLC2A1 and SLC2A10 increases, HK1 decrease, and G6PC3 increase, suggesting that FDG increases in prostate cancer cell without enough phosphorylated FDG in prostate cancer cells for the imaging 1 hour post FDG injection.

Research Support T32EB006351