Leukemic Stem Cells Evade Chemotherapy by Metabolic Adaptation to an Adipose Tissue Niche

Abstract

Adipose tissue (AT) has previously been identified as an extra-medullary reservoir for normal hematopoietic stem cells (HSCs) and may promote tumor development. Here, we show that a subpopulation of leukemic stem cells (LSCs) can utilize gonadal adipose tissue (GAT) as a niche to support their metabolism and evade chemotherapy. In a mouse model of blast crisis chronic myeloid leukemia (CML), adipose-resident LSCs exhibit a pro-inflammatory phenotype and induce lipolysis in GAT. GAT lipolysis fuels fatty acid oxidation in LSCs, especially within a subpopulation expressing the fatty acid transporter CD36. CD36(+) LSCs have unique metabolic properties, are strikingly enriched in AT, and are protected from chemotherapy by the GAT microenvironment. CD36 also marks a fraction of human blast crisis CML and acute myeloid leukemia (AML) cells with similar biological properties. These findings suggest striking interplay between leukemic cells and AT to create a unique microenvironment that supports the metabolic demands and survival of a distinct LSC subpopulation.

Figure 1. Adipose Tissue Functions as A Reservoir for LSCs

(A) Leukemia cell (GFP+/YFP+) frequency in in multiple tissues (CD45+ cells) – bone marrow (BM), gonadal adipose tissue (GAT), inguinal adipose tissue (IAT), spleen and peripheral blood (PB). Error bars show means ± S. D. n=5, ** P<0.005. (B) Histological analyses of leukemia infiltration of adipose tissue. Hematoxylin and eosin (H&E) labeling of tissue morphology of GAT (20X magnification). Orange arrow indicates a blood vessel identified by CD31 labeling and morphology. Immunofluorescent labeling of GFP (indicated in red) shows extensive infiltration of leukemia cells throughout the adipose tissue. Labeling with CD31 and Plin indicate endothelial cells and adipocytes respectively. DAPI indicates nuclei of cells. Overlay of all fields (bottom row) shows the position of numerous leukemia cells in direct contact with adipocytes. Scale bar = 50 microns. (C) Frequency of immunophenotypically-defined LSCs in multiple tissues. Error bars show means ± S. D. n=3, * P<0.05. (D) Limiting-dilution analysis of LSC frequency in GAT vs. BM. (E) Homing assays to determine the degree of bulk leukemia cell and LSC migration to GAT and BM. Error bars show means ± S. D. n=4, ** P<0.005.

Figure 2. Leukemia Cells in Gonadal Adipose Tissue are Pro-inflammatory

(A) Heat map of genes differentially expressed by LSCs and their non-leukemic BM counterparts (NBM). RNA from LSCs in each tissue and NBM was isolated and subjected to RNA-seq. Three cohorts for each type of LSCs were analyzed. Each cohort consisted of pooled LSCs from ten mice. Genes that are only significantly differentially expressed in all three cohorts were chosen to make the heat map. (Red=up-regulated, Blue=down-regulated). See also Figure S2A. (B) Expression of pro-inflammatory cytokines/chemokines genes in BM leukemia cells and GAT leukemia cells. Error bars show means ± S. D. from triplicates. ** P<0.005. (C) GAT SVF from normal mice and GAT non-leukemic (GFP−/YFP−) SVF from leukemia mice were isolated by flow cytometric sorting to determine the expression of pro-inflammatory cytokines. Error bars show means ± S. D. from triplicates. ** P<0.005. (D) Atrophy of GAT is found in leukemic mice. GAT from control (transplanted with normal BM hematopoietic cells) and leukemic mice was weighted at day 12 after transplantation. Arrows indicate GAT. Error bars show means ± S. D. n=8, ** P<0.005.

Figure 3. Leukemic Adipose Tissue is Lipolytic

(A) GAT explants (300mg) from control (transplanted with normal BM hematopoietic cells) and leukemic mice were cultured in DMEM containing 1% BSA for 24h and FFA and FABP4 levels in culture medium were determined. Error bars show means ± S. D. from triplicates. ** P<0.005. (B) Serum FFA level in control and leukemic mice. Error bars show means ± S. D. n=8, ** P<0.005. (C) mRNA expression of lipolysis related genes in control and leukemic GAT. Error bars show means ± S. D. from triplicates. ** P<0.005. (D) Protein levels of ATGL, LPL and CIDEA in control and leukemic GAT. Each lane represents GAT from two mice. (E and F) Normal GAT explants were cultured in DMEM containing 1% BSA and indicated cytokines (5ng/ml for TNF-α and 10ng/ml for others) for 24h. FFA level in culture medium (E) and expression of lipolysis related genes (F) were determined. Error bars show means ± S. D. from triplicates. ** P<0.005. (G) Conditioned medium (CM) from control and leukemic GAT was subjected to cytokine arrays. Array results as well as quantification for IL-1α, IL-1β, TNF-α and CSF2 were shown. Error bars show means ± S. D. from the duplicate dots for each cytokines. ** P<0.005. (H) Leukemia and normal BM cells were treated with BSA or palmitate (PA) (20 and 100μm) for 24h and expression of pro-inflammatory cytokines/chemokines genes was determined. Error bars show means ± S. D. from triplicates. ** P<0.005. (I) Leukemia cells (1 million/ml) were treated with BSA, PA (100μm) or oleic acid (OA) (100μm) for 24h, and ELISA was performed to determine the concentration of IL-1α in media. Error bars show means ± S. D. from triplicates. ** P<0.005.

Figure 4. CD36, A Fatty Acid Transporter, Modulates Energy Metabolism in Leukemia Cells

(A) Leukemia cells and non-leukemia cells (GFP−/YFP−) from the same sample were sorted and FAO assays monitoring the release of tritiated water (³H₂O) from tritium labeled palmitate were performed to determine their FAO rates. Error bars show means ± S. D. from triplicates. ** P<0.005. (B) FAO rates of LSCs, Lin+ leukemia cells and their counterparts from normal BM were determined with or without the presence of adipocytes conditioned medium (CM) or the CPT1 inhibitor etomoxir (Ex) (100uM). Error bars show means ± S. D. from triplicates. * P<0.05, ** P<0.005. (C) Expression of CD36 mRNA in LSCs, Lin+ leukemia cells and their counterparts. Error bars show means ± S. D. from triplicates. ** P<0.005. (D) FAO rates of LSCs and Lin+ leukemia cells and their counterparts from normal BM were determined with or without the presence of the CD36 inhibitor SSO (50μM). Error bars show means ± S. D. from triplicates. ** P<0.005. (E) FAO rate in CD36+ vs CD36− LSCs in the presence or absence of SSO (50μM). Error bars show means ± S. D. from triplicates. ** P<0.005.

Figure 5. CD36 Expression Segregates LSCs into Two Distinct Populations

(A) Frequency of CD36+ LSCs in BM, GAT, PB, and Spleen. Error bars show means ± S. D. n=3, ** P<0.005. (B) Homing ability of CD36+ and CD36− LSCs to GAT. Error bars show means ± S. D. n=3, **P<0.005. (C) Functional analysis of LSC potential in CD36+ vs. CD36− LSCs. Equal numbers of CD36+ and CD36− LSCs were injected into recipient mice (1000 cells/mouse). Two weeks later, BM cells from recipients were collected to examine the composition of leukemia cells. (D) Limiting-dilution analysis of leukemia-initiating cell frequency in CD36+ and CD36− LSCs. (E) Cell cycle status of CD36+ vs CD36− LSCs as determined by in vivo BrdU labeling. Error bars show means ± S. D. n=3, ** P<0.005. (F) Chemotherapy prolongs survival of leukemic mice. Leukemic mice were treated with a chemotherapeutic regimen as described in experimental procedures. Survival of control (vehicle treated) and chemotherapy treated leukemic mice was monitored. n=5, ** P<0.005. (G) Preferential survival of CD36+ LSCs in GAT after chemotherapy. BM and GAT from control (vehicle treated) and chemotherapy treated leukemic mice were harvested and the composition of residual leukemia cells was examined. Error bars show means ± S. D. n=5, * P<0.05, ** P<0.005.

Figure 6. Loss of CD36 decreases leukemic burden in gonadal adipose tissue and sensitizes LSCs to chemotherapy

(A) Leukemia burden in the GAT and BM of wild type (WT) vs CD36 knock-out (KO) leukemia mice. Error bars show means ± S. D. n=5, ** P<0.005. (B) Percentage of LSCs in BM and GAT resident leukemia cells from WT and CD36 KO leukemia mice. Error bars show means ± S. D. n=5. (C) GAT weight in WT and KO leukemia mice. Error bars show means ± S. D. n=5, * P<0.05. (D) Serum FFA level in WT and KO leukemia mice. Error bars show means ± S. D. n=5, * P<0.05. (E) Secretion of IL-1α in WT and CD36KO leukemia cells. Leukemia cells (1 million/ml) were treated with BSA or OA (100μM) for 24h, and ELISA was performed to determine the concentration of IL-1α in media. Error bars show means ± S. D. from triplicates. ** P<0.005. (F) Homing ability of WT and CD36 KO LSCs to GAT. Homing ability of LSCs was determined by the ratio of the percentage of LSCs in the leukemia cells localized to GAT to the percentage of LSCs in leukemia cells before injection. Error bars show means ± S. D. n=4, * P<0.05. (G) FAO rate in WT and CD36KO LSCs. FAO rate was measured with or without the CD36 inhibitor SSO (50 μM). Error bars show means ± S. D. from triplicates. ** P<0.005. (H and I) Percentage of residual leukemia cells (H) as well as LSCs (I) was determined in WT and CD36 KO leukemia mice after chemotherapy. Error bars show means ± S. D. n=5, * P<0.05.

Figure 7. CD36 Expression Segregates Primary Human blast crisis CML and AML cells to Two Functionally Distinct Populations

(A) Expression of CD36 in primary human blast crisis CML (bcCML) cells. Red text indicates specimens with sufficient CD34+/CD36+ cells for subsequent analyses. (B) Fatty acid uptake in bcCML cells. Leukemia cells were serum starved and pre-treated with or without SSO (50μM) before incubation with BODIPY-Dodecanoic acid (1μM). Error bars show means ± S. D. from triplicates. ** P<0.005. See also Figure S7A. (C) FAO rate in CD36+/CD34+ vs. CD36−/CD34+ cells in the presence or absence of CD36 inhibitor SSO (50 μM). Error bars show means ± S. D. from triplicates. ** P<0.005. See also Figure S7B. (D) Leukemia burden in BM, GAT and PB in immune deficient NSG mice transplanted with human bcCML cells. Error bars show means ± S. D. n=5, **P<0.005. See also Figure S7C. (E) CD36+/CD34+ cells are enriched in GAT relative to BM. Error bars show means ± S. D. n=5, ** P<0.005. See also Figure S7D. (F) Homing ability of CD36+/CD34+ cells to BM and GAT. Leukemia cells were pre-treated with or without SSO (50 μM) for 1h before injection. Error bars show means ± S. D. n=5, ** P<0.005. See also Figure S7E. (G) Serum FFA level in normal NSG mice and NSG mice engrafted with bcCML cells. Error bars show means ± S. D. n=5, * P<0.05, ** P<0.005. (H) NSG mice transplanted with bcCML cells were treated with Ara-C (100mg/kg/day) or saline (control) for 3 days. Composition of BM residual leukemia cells was examined to evaluate relative drug resistance of CD36+/CD34+ vs. CD36−/CD34+ cells. Error bars show means ± S. D. n=4, * p<0.05, ** P<0.005. See also Figure S7H. (I) Expression of CD36 vs CD34 in primary human AML samples. Red text indicates specimens with sufficient CD34+/CD36+ cells for subsequent analyses. (J) Fatty acid uptake in AML cells. Leukemia cells were serum starved and pre-treated with or without SSO (50μM) before incubation with BODIPY-Dodecanoic acid (1μM). Error bars show means ± S. D. from triplicates. ** P<0.005. See also Figure S7I.