Background: Dysregulated cellular metabolism is a hallmark of breast cancer, and targeting it has promising implications for improving care and patient outcomes. Specifically, heterogeneity in tumor metabolism is thought to play a role in determining chemotherapy response, the development of resistance, and promoting metastastasis. Despite this, metabolic tumor heterogeneity for individual breast cancer patients has not been characterized completely.
Methods: In this study, we used state-of-the-art techniques to characterize metabolic heterogeneity within individual patient tumors by integrating single cell RNA-seq data with genome-scale metabolic modeling. Using SimBioSys’ TumorScope – a commercially available biophysical modeling platform, we compared intra-tumoral metabolic heterogeneity from experimental single cell RNA-seq data to simulated intra tumoral heterogeneity.
Results: Using single cell RNA-seq data, we found that intra-tumoral gradients in nutrient availability are widely present within patient tumors (for a single luminal A patient, glucose import flux ranged from 0.19 – 1.25 g/gDW/day, while glutathione import ranged from 0.004 – 0.054 g/gDW/day). We also found that these gradients lead to cellular growth rate gradients within individual tumors (for our representative patient, median SGR = 0.62 %/day +/- 0.33 %/day stdev). Using TumorScope, we found this same gradient behavior within patient tumors. Selecting a similarly growing luminal A patient from our TumorScope simulations resulted in gradients in glucose import (range = 0.17 – 1.26 g/gDW/day), glutathione import (range = 0.024 – 0.058 g/gDW/day), and tumor SGR (median = 0.40 %/day, stdev = 0.42 %/day), which closely match metabolism from single cells (comparing maximum-scaled SGR distributions between single cells and TumorScope yielded a p-value = 0.10). We next examined which nutrients govern heterogeneity in tumor SGR. We found that glucose availability with the tumor microenvironment is more limiting to cell growth than oxygen availability, and this result was consistent between metabolic profiles from both single cell RNA-seq data and TumorScope simulations. TumorScope’s spatially resolved simulations offered the additional insight that gradients in nutrient availability are caused by heterogeneity in the distribution of macro- and micro-vasculature and the composition of the tumor microenvironment. We then used data reduction techniques to compare populations of single cells with differing metabolic phenotypes to identify molecular behavior at the single cells in higher molecular resolution. We found that single cells collected from the clinic co-cluster with single cells from TumorScope simulations, suggesting that a significant amount of intra-tumoral metabolic heterogeneity observed in patients is captured by TumorScope simulations.
Conclusion: Accessing tumor heterogeneity has traditionally required specialized equipment, analytic expertise, and invasive procedures, largely limiting its study to large, academic hospitals. Currently, metabolic heterogeneity is only understood in 2D and for few markers (using pathology slides) or in relatively few cells with little or no spatial resolution (for single cell RNA-seq). TumorScope provides a novel approach to simulate metabolic heterogeneity at the single cell scale in 3D across a whole tumor. TumorScope democratizes the study of tumor heterogeneity by making it accessible to clinicians and researchers from MRI data alone. TumorScope has the capability to capture tumor metabolic heterogeneity at a higher scale than previously achievable which will allow for a dramatic increase in our understanding of tumor biology and ultimately improve clinical decision making.
