Reciprocal change in Glucose metabolism of Cancer and Immune Cells mediated by different Glucose Transporters predicts Immunotherapy response

The metabolic properties of tumor microenvironment (TME) are dynamically dysregulated to achieve immune escape and promote cancer cell survival. However, in vivo properties of glucose metabolism in cancer and immune cells are poorly understood and their clinical application to development of a biomarker reflecting immune functionality is still lacking. Methods: We analyzed RNA-seq and fluorodeoxyglucose (FDG) positron emission tomography profiles of 63 lung squamous cell carcinoma (LUSC) specimens to correlate FDG uptake, expression of glucose transporters (GLUT) by RNA-seq and immune cell enrichment score (ImmuneScore). Single cell RNA-seq analysis in five lung cancer specimens was performed. We tested the GLUT3/GLUT1 ratio, the GLUT-ratio, as a surrogate representing immune metabolic functionality by investigating the association with immunotherapy response in two melanoma cohorts. Results: ImmuneScore showed a negative correlation with GLUT1 (r = -0.70, p < 0.01) and a positive correlation with GLUT3 (r = 0.39, p < 0.01) in LUSC. Single-cell RNA-seq showed GLUT1 and GLUT3 were mostly expressed in cancer and immune cells, respectively. In immune-poor LUSC, FDG uptake was positively correlated with GLUT1 (r = 0.27, p = 0.04) and negatively correlated with ImmuneScore (r = -0.28, p = 0.04). In immune-rich LUSC, FDG uptake was positively correlated with both GLUT3 (r = 0.78, p = 0.01) and ImmuneScore (r = 0.58, p = 0.10). The GLUT-ratio was higher in anti-PD1 responders than nonresponders (p = 0.08 for baseline; p = 0.02 for on-treatment) and associated with a progression-free survival in melanoma patients who treated with anti-CTLA4 (p = 0.04). Conclusions: Competitive uptake of glucose by cancer and immune cells in TME could be mediated by differential GLUT expression in these cells.


FDG PET imaging
Patients fasted for at least 6 hours before the image acquisition to have blood glucose level less than < 140 mg/dL. FDG (5.18 MBq/kg) was intravenously injected to patients and the emission scan was performed 60 minutes after the injection using dedicated PET/CT scanners (Biograph mCT40 and mCT64, Siemens Healthcare, Germany). The emission scan was obtained from the skull base to the proximal thigh and a CT scan was consecutively obtained for attenuation correction. PET images were reconstructed using an iterative algorithm (ordered-subset expectation maximization) with image matrix size of 256 × 256, iteration number 2 and 21 subsets.
FDG-PET/CT data provided by The Cancer Imaging Archive (TCIA) were also used [1]. Twenty-two patients who had both RNA-seq and FDG PET data were used. FDG-PET/CT images were acquired according to the standard imaging protocol of multiple institutes. PET data were reconstructed by different methods and parameters based on iterative algorithms according to institutes.

Preprocessing of TCGA RNA-sequencing data and concordant FDG PET from TCIA
We used mRNA transcriptome data of LUSC from The Cancer Genome Atlas projects (TCGA) and concordant FDG PET image from TCIA for validation purpose [1]. Using 'TCGABiolinks' R package [2], we downloaded the level three RNA sequence data of lung squamous cell carcinoma from TCGA data portal (https://portal.gdc.cancer.gov/) obtained FDG PET data of TCIA were analyzed with the same manner described in the previous section.

Estimation of glycolysis enrichment score
To examine the overall glycolytic activity of tumor microenvironment, we used Reactome to select genes of glycolysis pathway [3]. Single sample gene set enrichment analysis was applied against the curated gene sets of the Reactome glycolysis pathways to define metabolic profiles of each cancer samples. We implemented single sample gene set enrichment analysis [3] using the curated gene sets from canonical pathways (MSigDB C2, Broad Institute; version 3.0) with GSVA R/Bioconductor package [4,5]. The enrichment scores of Reactome glycolysis pathway was normalized by z-score across all samples.

Glucose metabolism profiles of single cell data
Glucose metabolism profiles of single cell were evaluated by the scaled gene expression data. The gene expression values of glucose transporter families were obtained.
All cells were plotted by GLUT1 and GLUT3 expression. The scaled gene expression of GLUTs with z-score > 0 used as a threshold to determine GLUT1-and GLUT3-positive cells.
The proportion of four different groups according to GLUT1 positivity and GLUT3 positivity was estimated. The glycolysis enrichment and oxidative phosphorylation of each cell was estimated by AddModuleScore function of Seurat package [6]. The gene sets related to glycolysis and oxidative phosphorylation were obtained by Reactome [3].