Adipose Tissue Functions As a Reservoir for Leukemia Stem Cells and Confers Chemo-Resistance

Adipose tissue (AT) serves as a storage site for lipids as well as an endocrine organ. In the context of cancer, AT can function to facilitate the progression of tumors. Interestingly, recent studies have shown that AT acts as an extramedullary reservoir of hematopoietic stem cells (HSCs), suggesting the presence of a HSC niche in AT. The fact that leukemia stem cells (LSCs) co-opt the HSC marrow and the supportive effect of AT on cancer cells led us to hypothesize that AT functions as a reservoir for LSCs.

  To test this hypothesis, we first examined whether there were LSCs residing in AT. Using a murine model of primary blast crisis CML, we found enrichment of phenotypically primitive leukemia cells (PLCs) in the gonadal adipose tissue (GAT) relative to bone marrow, spleen and peripheral blood (Figure 1). The high percentage of PLCs in GAT led us to postulate that PLCs are preferentially reliant on fatty acids as an energy source since GAT was found to be lipolytic in leukemic mice. Indeed, we observed that PLCs had a higher rate of fatty acid oxidation (FAO) compared to lineage+ leukemia cells and their non-leukemic counterparts. Additionally, conditioned medium from adipocytes selectively increased the FAO rate in PLCs suggesting a FAO regulatory role of AT. Further, we showed that CD36, a fatty acid transporter, was highly expressed by a subset of PLCs and selectively regulated FAO in PLCs. Interestingly, CD36+ PLCs were strikingly enriched in GAT. Together, these results suggested that GAT functions as a reservoir for CD36+ PLCs and facilitates use of FAO for energy metabolism.

  Next we examined the characteristics of CD36+ PLCs. CD36+ PLCs differed metabolically from CD36- PLCs with regard to fatty acid metabolism. Specifically, we found CD36+ PLCs had a higher FAO rate and were more dependent on the transportation function of CD36 for FAO (Figure 2). We also compared the cell cycle status between these two populations and found CD36+ PLCs were more quiescent. Interestingly, both CD36+ and CD36- PLCs contained LSCs and were able to reconstitute the whole leukemic BM when transplanted into recipients. Collectively, these findings indicated that at least two metabolically distinct types of leukemia-initiating cells exist in our blast crisis model, and that the major forms of energy metabolism can differ as a function of anatomical location and expression of CD36.

  Since front line agents commonly used in cancer generally target actively cycling cells, we speculated that CD36+ PLCs might be more drug resistant due to their increased quiescence. Indeed, we found enrichment of CD36+ PLCs in BM after applying a 5-day chemotherapy regimen to leukemic mice, while CD36- PLCs were not protected. More importantly, we observed that CD36+ PLCs were highly enriched in GAT after chemotherapy suggesting GAT conferred chemo-resistance to resident CD36+ PLCs (Figure 3). Taken together, our observations imply that the heterogeneity found in PLCs is translated into drug sensitivities in different leukemic sub-fractions. Specifically, CD36+ PLCs represent a relatively drug resistant subpopulation and GAT serves as a reservoir for resident CD36+ PLCs.

  Lastly, we investigated whether these findings in the murine model could be recapitulated in primary human bcCML cells. We found that within the CD34+ subpopulation of primary human bcCML cells, CD36+ subset had a high FAO rate compared to CD36- subset. Furthermore, this CD36+ subset was more quiescent and drug resistant. These data suggest that a similar heterogeneity can be found in primary human bcCML cells based on the expression of CD36.

  Collectively, our findings suggested that GAT in leukemia mice functions as a reservoir for LSCs and confers chemo-resistance to resident leukemia cells, implying a potential role of GAT in the pathogenesis of leukemia and relative efficacy of therapeutic challenge. Furthermore, our data indicate metabolic heterogeneity within LSC populations, where pathways controlling energy consumption can differ. We propose that metabolic heterogeneity in LSCs may contribute to the challenge in effectively eradicating such cells.