Background

Tumor dependence on adequate blood supply has provided a rationale for anti-angiogenic cancer therapy (Carmeliet and Jain, 2000; Vasudev et al., 2013). Traditionally attributed to provision of oxygen and other blood-borne substances, the roles of the tumor vasculature in tumor biology were recently extended to include perfusion-independent, paracrine inputs exerted by secreted endothelial-produced angiocrine factors (Butler et al., 2010; Beck et al., 2011; Lu et al., 2013), as well as via direct tumor-BV contacts mediated by reciprocal Notch signaling (Cao et al., 2014; Wieland et al., 2017). Because tumor neovascularization often fails to fully match increasing perfusion demands imposed by continual tumor growth, the relative distance of a tumor cell from the nearest perfused vessel is likely to greatly impact the nature and magnitude of vascular inputs operating through all indicated modes of tumor-BV communication. Correspondingly, differential exposures to vascular signals might facilitate emergence of certain tumor phenotypes and generate intra-tumor cell heterogeneity. To determine how vascular cells may impact tumor cells, tumor-endothelial cell (EC) co-cultures in both 2D and 3D configurations have been used (Khodarev et al., 2003; Timmins et al., 2004). However, because ex-vivo systems do not reveal perfusion-dependent effects and are also unable to recapitulate the complex interactions prevailing in the tumor microenvironment, in vivo approaches are clearly needed. The latter has thus far been limited to using EC-specific manipulations of selected candidate genes (Beck et al., 2011; Cao et al., 2014 and 2017). The need still existed for an integrative, high-throughput approach for unraveling the whole spectrum of genes, pathways and phenotypes affected by the vasculature and, moreover, a general approach to determine spatial ranges of vascular influences per each of these phenotypes.

The methodology described herein allows retrospective sorting of successive GBM tumor cells subpopulations residing at progressively increasing distances from the nearest BV and in a form suitable for subsequent transcriptomic and functional analyses. This methodology was used to understand how vasculature impacts cancer cell transcriptome and metabolome (Kumar et al., 2019). This methodology is not limited to the above two omics, but can be further extended to any other omics such as to study epigenome or methylome. If integrated with single cell RNA sequencing, this assay can help in dissecting the characteristics of tumor cells as a function of blood vessels to a very fine resolution. Of particular relevance is the existence of diverse metabolic niches within a tumor due to differential availability of oxygen and other blood-borne substances. Using this methodology we showed that the classical 'Warburg Effect', i.e., the use of aerobic glycolysis for energy production instead of the more efficient mitochondrial respiration, in fact, reflects overall summation of different metabolic niches, and that perivascular cells still actively rely on mitochondrial respiration. While there is a growing awareness of intratumor metabolic compartmentalization, and even of an interplay between different compartments, it has thus far been dealt with in a dichotomous manner of distinguishing 'normoxic' vs. 'hypoxic' microenvironments (Allen et al., 2016). Considering, however, existence of intermediate states shaped by the perfusion gradient and by differential exposure to angiocrine factors, our study showed that there is a continuum of diverse microenvironments. Ability to isolate such tumor subpopulations should aid in addressing open issues pertinent to the highly complex BV-tumor interplay. We believe this methodology can also be extended to other solid tumors too, as there exist an unveven distribution of vascular supply in most of the different types of solid tumors.