The ‘omics of obesity in B-cell acute lymphoblastic leukemia

Abstract

The obesity pandemic currently affects more than 70 million Americans and more than 650 million individuals worldwide. In addition to increasing susceptibility to pathogenic infections (eg, SARS-CoV-2), obesity promotes the development of many cancer subtypes and increases mortality rates in most cases. We and others have demonstrated that, in the context of B-cell acute lymphoblastic leukemia (B-ALL), adipocytes promote multidrug chemoresistance. Furthermore, others have demonstrated that B-ALL cells exposed to the adipocyte secretome alter their metabolic states to circumvent chemotherapy-mediated cytotoxicity. To better understand how adipocytes impact the function of human B-ALL cells, we used a multi-omic RNA-sequencing (single-cell and bulk transcriptomic) and mass spectroscopy (metabolomic and proteomic) approaches to define adipocyte-induced changes in normal and malignant B cells. These analyses revealed that the adipocyte secretome directly modulates programs in human B-ALL cells associated with metabolism, protection from oxidative stress, increased survival, B-cell development, and drivers of chemoresistance. Single-cell RNA sequencing analysis of mice on low- and high-fat diets revealed that obesity suppresses an immunologically active B-cell subpopulation and that the loss of this transcriptomic signature in patients with B-ALL is associated with poor survival outcomes. Analyses of sera and plasma samples from healthy donors and those with B-ALL revealed that obesity is associated with higher circulating levels of immunoglobulin-associated proteins, which support observations in obese mice of altered immunological homeostasis. In all, our multi-omics approach increases our understanding of pathways that may promote chemoresistance in human B-ALL and highlight a novel B-cell–specific signature in patients associated with survival outcomes.

The obesity pandemic shows no signs of slowing as it is estimated that more than 50% of the adult population in the United States, or roughly 130 million people, will present with overweight or obesity within the next 8 years (1,2). Obesity also impacts 14.7 million American children and adolescents (those aged 2-19 years) (1,2). The increasing number of individuals living with obesity in the United States and globally is concerning, given that obesity is associated with higher mortality rates in many pathological settings including in patients with pathogenic infections (3-7) and those with many forms of cancer (8).

Understanding the relationship between obesity and cancer has been and continues to be a growing area of interest. Although it has been appreciated for decades that obesity increases the risk of developing specific classes of solid cancers (9,10) and mortality rates associated with some diseases are higher compared with those observed in lean patients (8), our understanding of how increased adiposity impacts hematological malignancies is growing rapidly (11-19). Notably, several groups, including ours, have demonstrated that adipocytes directly promote chemoresistance in human B-cell acute lymphoblastic leukemia (B-ALL) cells to various chemotherapies including L-asparaginase, methotrexate (MTX), doxorubicin, daunorubicin, and vincristine (18,20-23). Furthermore, this protective effect is multifactorial and can be mediated through adipocyte-derived lipids and glutamine, which increase β-oxidation and glutaminolysis, respectively, in B-ALL cells, fueling pro-proliferative and anti-apoptotic growth of leukemia cells (20,22). Additionally, we have recently demonstrated that the adipocyte secretome induces the upregulation of galectin-9 on human B-ALL cells, protecting them from chemotherapy-mediated cytotoxicity by inducing a senescence-like state (18). Despite the important insight into B-ALL–adipocyte interactions yielded by these studies, we decided to expand our knowledge of how adipocytes impact human B-ALL cells in the presence and absence of the chemotherapy MTX, which is used in all treatment phases for this disease (24,25), using a multi-omics approach to account for the complexity of the adipocyte secretome and its multifactorial impact on leukemia cells.

To gain a comprehensive understanding of this relationship, we performed single-cell RNA-sequencing (scRNA-seq) of circulating blood cells collected from lean and obese mice to determine the impact of adiposity on the population frequency and gene expression patterns of nonmalignant B cells. We also performed mass spectrometry on adipocyte-conditioned media (ACM) to identify potential adipocyte-induced drivers of observed changes. We found that increased adiposity suppresses specific B-cell subpopulations in the adipose tissue of mice and downregulates a gene expression signature, which is associated with higher survival rates in patients with B-ALL. Analysis of circulating factors in healthy patients and those with B-ALL, grouped by body mass index, revealed circulating factors associated with better cardiovascular function and lipid metabolism are higher in lean individuals, whereas those involved with inflammatory responses (particularly immunoglobulin-associated proteins) are higher in individuals with obesity. Given these observations, we subsequently used mass spectrometry and RNA-seq analyses to define the impact of ACM on human B-ALL cells in the presence and absence of MTX. Collectively, these studies revealed that relative to responses observed in unconditioned media (UCM) or bone marrow stromal cell–conditioned media (SCM) cultures, human B-ALL cells exposed to the adipocyte secretome upregulate Wnt16-mediated pathways, those involved in increased glutaminolyisis and β-oxidation, B-cell development, and chemoresistance. Notable downregulated pathways included those involved in tumor suppression (p53) and apoptosis. Interestingly, pathways induced by adipocytes alone were further enhanced, in the presence of MTX treatment, where additional pathways known to promote chemoresistance in solid tumors were also induced (DNAH8, MTHFD1L, and S100A11). Given our previously published results demonstrating adipocyte-mediated chemoresistance to MTX (18), our multi-omics approach to understanding this relationship highlights previously unknown pathways, which may contribute to more aggressive B-ALL disease in adipose-rich microenvironments.

Methods

Cell lines

Human B-ALL cell lines (REH and RCH-AcV) were a generous gift from Dr Christopher Porter at Emory University. REH and RCH-AcV were maintained in RPMI1640 (cat#10-041-CV, Corning, Glendale, AZ, USA), supplemented with 20% fetal bovine serum (FBS; cat#S11550, BioTechne, Minneapolis, MN, USA), USA. OP-9 bone marrow stromal cells were maintained in α-Minimum Essental Medium (MEM) (cat#15-012-CV, Corning) supplemented with 20% FBS. For adipocyte differentiation, 10⁵ OP-9 cells were plated in 6-well plates in Dulbecco’s Modified Eagle Medium (DMEM; cat#10-017-CV, Corning) supplemented with 10% FBS as previously described (18,26). After 24 hours of culture, the media was removed and switched to differentiation media, which is composed of α-MEM supplemented with 1.8 mM oleate (cat#O7501, Sigma, St. Louis, MO, USA) bound to bovine serum albumin (cat#A6003, Sigma) with molar ratio 5.5:1 along with 175 nM insulin (cat#I6634, Sigma) and 0.2% FBS. ACM was collected after 3 days of differentiation and used in the experiments described in this study. For SCM, OP-9 cells were plated in DMEM supplemented with 10% FBS, and conditioned media were collected on day 3 of culture.

Human samples

De-identified frozen plasma (n = 3 from disease-free lean and donors with obesity; average age of 23.5 years for all samples) and serum (n = 7 obtained at diagnosis from lean patients with B-ALL and those with B-ALL and obesity; average age of 13.9) samples were obtained through the Children’s Healthcare of Atlanta and Emory University’s Children’s Clinical and Translational Discovery Core (protocol number IRB00089506) and the Aflac Cancer and Blood Disorders Center Leukemia and Lymphoma Biorepository (protocol number IRB00034535), respectively. Informed consent for the inclusion of samples in the biorepositories is required before banking, distribution, and usage. The use of human samples for this study received an institutional review board exemption from the Emory University institutional review board given that our usage of de-identified samples is not considered research with “human subjects” or “clinical investigation” as outlined in Emory policies and procedures and federal rules.

Murine adipose immune cell study

Female C57BL/6J mice were purchased from Jackson Laboratories at 6 weeks of age (stock #000664). Mice were acclimated for 1 week and then given either low fat, low sucrose diet (LFLS; Research Diets D11092101) or high fat, high sucrose diet (HFHS; Research Diets D15031601) ad libitum. After 2 weeks in standard ventilated housing, mice were moved to warm water blankets set at 42°C, to maintain an internal cage temperature of approximately 30°C, as we have previously described (27,28). At 14 weeks of age, mature mice were ovariectomized under isoflurane anesthesia and immediately supplemented with 17β-estradiol (E₂), provided in the drinking water at a final concentration of 0.5 µM. E₂ supplementation continued for 2 weeks, then mice were randomized based on body fat percentage by quantitative magnetic resonance Echo (ECHO MRI) within diet groups for 7 weeks and then fasted for 4-6 hours prior to euthanasia according to the Association of Assessment and Accreditation of Laboratory Animal Care guidelines. CD45+ cells were harvested from subcutaneous adipose and analyzed using scRNA-seq analysis. All animal studies were approved by the University of Colorado Denver institutional animal care and use committee.

Adipose tissue fluorescence-activated cell sorting

Whole inguinal subcutaneous adipose depots were excised, minced briefly (5 minutes) with dissecting scissors, and digested for 75 minutes at 37°C in a collagenase solution (Collagenase type II, Worthington LS004177 and HBSS with 3% bovine serum albumin, 0.8 mM ZnCl2, Mg, Ca, 0.8 mg/ml collagenase). Stromal pellets were incubated in red blood cell lysis buffer (Sigma Aldrich), washed, and stained with fluorescent-conjugated antibodies specific for CD45 (Biolegend cat#103108), CD31 (Biolegend cat#102418), and CD29 (Biolegend cat#102218). CD45+ cells were retained via cell sorting for further analysis.

Results

HFHS diet reduces the frequency of B-cell populations in murine adipose tissue and downregulates gene programs associated with survival in patients with B-ALL

To determine the impact of obesity on human B-ALL cells, we first assessed how obesity affected nonmalignant B cells using a murine model of diet-induced obesity. Mice were maintained on either LFLS diets (lean mice) or HFHS diets (obese mice) for 7 weeks, and immune cells (CD45⁺) were harvested from subcutaneous adipose tissue and analyzed using scRNA-seq. Given the adipocytes attract and protect B-ALL cells from the cytotoxic effects of chemotherapy (29), we hypothesized that this microenvironment may also impact the function or frequency of normal B cells.

After quality control, normalization, and filtering analysis, the scRNA-seq data of 16 246 cells from the lean (12 357 cells) and obese (3709 cells) groups were merged and clustered together. An unsupervised analysis identified 21 clusters based on their transcriptomic profiles. Cells were annotated based on known lineage markers (eg, T cells [Cd3d⁺], B cells [Cd79a⁺], and macrophages [Cd14⁺]). This approach resulted in the identification of 9077 Cd79a⁺ B cells from lean mice and 2753 Cd79a⁺ B cells from obese mice. The lower number of B cells found in the adipose tissue of obese relative to lean mice may reflect the differential migration of these lymphocytes to adipose tissue because of the documented changes in the adipocyte secretome induced by obesity (30-32) or the altered proliferative capacity of nonmalignant B cells when they home to adipose tissue (33).

In our split uniform manifold approximation and projection plot, which depicts the distribution of 11 subclusters in lean and obese mice, we observed that the frequency of B-cell populations in adipose tissue was modulated by HFHS diets, with a notable reduction in B cells found in cluster numbers 3 and 8 (Figure 1, A). Because the lean sample had a higher proportion of captured cells, we determined the percentage contribution in each cluster of the 2 groups, which further highlighted an enrichment of B cells in lean relative to obese mice in these clusters (Figure 1, B). Additionally, we quantified the shift in abundance of cell types between lean and obese mice by applying the MiloR package, which also identified a cluster of neighborhoods in the region of cluster 3 with a log-fold change greater than 3 (Figure 1, C). Given that cluster 3 showed the largest reduction in the frequency of B cells in adipose tissue, we focused our analysis on this group for further studies.

Figure 1.

Diet-induced obesity results in a loss of B-cell populations in adipose tissue, and this signature is associated with poor survival outcomes in patients with B-ALL. A) Split UMAP plot showing B-cell clusters in the dataset for lean and obese mice (n = 3-4 pooled mice per diet type). The clusters are numbered and colored as shown in the legend on the right labeled with the top 2 markers from each cluster. The cluster(s) showing the relative differences in abundance between lean and obese mice are circled. B) Histograms showing relative differences in the percentage of cells from lean (orange—left bar) and obese (teal—right bar) samples across clusters 3 and 8. C) Differential enrichment of cells from lean (red) and obese (blue) samples plotted using the MiloR tool. The color bar in the legend indicates enrichment in obese (blue) or lean (red) groups. D) Heatmap of the top differentially expressed genes in cluster 3 vs all other clusters. Red denotes upregulated and green denotes downregulated genes. E) Statistically significantly upregulated (red) and downregulated (blue) pathways/GO and transcriptional factors based on the gene signature derived from cluster 3 relative to the other B cells. F) Overall survival probability plot for the expression of cluster 3–derived top 10 genes (Figure 1, D) in the TARGET-ALL-P2-B- cell dataset using the survival genie platform. Statistical significance in Figure 1, B, as determined by the student t test, is denoted by **** P < .0001. B-ALL = B-cell acute lymphoblastic leukemia; UMAP = uniform manifold approximation and projection.

To evaluate the markers statistically significantly expressed in cluster 3, we performed differential gene expression analysis between cluster 3 and all other B-cell clusters. Figure 1, D, depicts a heatmap of the top 10 differentially expressed genes in cluster 3 compared with the remaining B-cell clusters sorted by the average log-fold change (P < .001). Gene expression analysis of this group revealed that B cells in this cluster more highly express genes that encode heat shock proteins (Hspa1a, Hspa1b, Hsph1, Dnajb1), adaptor proteins that promote endocytosis via clathrin-coated pits [Fchsd2 (34)], proteins that enable cadherin-binding activity and regulate actin dynamics [Swap70 (35)], proteins that augment signal transduction and increase susceptibility to B-cell transformation if de-regulated [Jun and Gem (36,37)], and proteins that suppress aberrant B-cell activation and apoptosis in lymphocytes [Klf2 (38-41)]. Using this signature, we analyzed the pathways both upregulated and downregulated in this cluster using Metascape. This analysis revealed that pathways upregulated in B cells found in cluster number 3 are associated with lymphocyte activation, leukocyte activation, immune response–related signaling, and B-cell receptor activation (the top 4 pathways) (Figure 1, E). These upregulated pathways in B cells found in cluster number 3 were mainly associated with high gene expression levels of heat shock transcription factor 1 (Figure 1, E). This transcription factor mitigates replicative and genotoxic stress in lymphocytes (B cells and T cells), without inducing cell cycle arrest (42,43). These properties are conferred through the induction of BCL6 (the second most highly upregulated gene associated with a transcription factor in B cells found in cluster number 3; Figure 1, E), which represses genes involved in DNA damage sensing and checkpoint activation (42,43). Pathways downregulated in B cells found in cluster number 3 include those involved in mRNA metabolic processes, cellular responses to stimuli, and chromatin organization (the second through fourth most downregulated pathways) (Figure 1, E). These changes were associated with lower expression of genes encoding those involved in B-cell differentiation, calcium signaling, DNA synthesis, proliferation, and chromatin structure [MYC and ING1 (44,45)] (Figure 1, E). A closer look into genes associated with B-cell development in clusters 3 and 8 revealed that Runx1 was more highly expressed in B cells in cluster 8 (Supplementary Figure 1, available online). Additionally, we found that Btg, an antiproliferative gene in B cells (46), was more highly expressed in B cells found in cluster 3 (Supplementary Figure 2, available online), whereas we observed an elevated, yet statistically insignificant, increase in genes associated with oncogenic transformation in both clusters (Myc and Kras; Supplementary Figure 3, available online). In all, these results indicate that obesity may compromise the function of normal B cells in multiple ways and prime these lymphocytes for oncogenic transformation.

Given the HFHS diet reduces the frequency of B cells in cluster number 3 in adipose tissue and alters their transcriptional programs, we wanted to determine if this genomic signature was related to B-cell malignancies in humans. We mined the Therapeutically Applicable Research to Generate Effective Treatments (TARGET)-ALL-P2-B-cell dataset using the survival genie platform to determine if survival in patients segregated based on the expression of the top 10 differentially regulated genes found in B cells that are found in cluster 3. Notably, we observed that a statistically significant survival advantage was conferred to patients with B-ALL expressing high gene expression levels of cluster 3 B-cell–associated genes (>50% survival over 8 years; Figure 1, F), which are upregulated in nonmalignant B cells isolated from mice fed LFLS diets. In contrast, a low expression of cluster 3 B-cell–associated genes in patients with B-ALL was associated with a statistically significant reduction in survival (hazard ratio = 0.54; P = .0042) (Figure 1, F). Overall, these results demonstrate that increased adiposity is associated with altered B-cell homeostasis, and diet-induced gene signatures in nonmalignant B cells found in adipose tissue are predictive of survival outcomes in patients with B-ALL.

Changes in circulating factors are associated with obesity in nonleukemic donors and patients with B-ALL

Our murine data demonstrated that obesity substantially impacts nonmalignant B cells and induces patterns in normal B cells, which may be predictive of outcomes in patients with B-ALL. We therefore wanted to determine how obesity impacted systemic factors in healthy donors and those with B-ALL.

We performed a small study in which the sera from lean adults (n = 3) and those with obesity (n = 3) were analyzed via mass spectrometry (see Supplementary Table 1, available online for donor characteristics). Notably, proteins that regulate satiety, lipid metabolism, and those that protect against various diseases (atherosclerosis, diabetes, and neurotoxicity) were higher in the sera of lean adults [apolipoprotein A4 (APOA4), apolipoprotein D (APOD), adiponectin (ADIPOQ) (47-50)] (Figure 2). High-density lipoprotein (HDL) is known as good cholesterol because it absorbs cholesterol in circulation and carries it back to the liver, which eliminates it from the body (51). We observed that phospholipid transfer protein was higher in the sera of lean relative to adult donors with obesity (Figure 2), which is notable because this protein transfers phospholipids from triglyceride-rich lipoproteins to HDL, and its deregulation can lead to atherosclerosis (52,53). Interestingly, histidine-rich glycoprotein was also expressed at higher levels in sera from lean donors relative to those with obesity (Figure 2). This protein is thought to work as an adaptor molecule given that it can interact with various ligands found in the blood simultaneously, and it has been noted to regulate various biological processes including immunity, pathogen clearance, coagulation, angiogenesis, fibrinolysis, necrosis, and cellular adhesion (54). Notably, immune-related proteins were elevated in the sera of donors with obesity relative to lean donors (Figure 2). These proteins were mostly those associated with B-cell or plasma cell activation and included several immunoglobulin-associated proteins (IGLV3-27, IGHD, IGLV5-45, IGKV6D-21, IGLV2-18, IGKV3D-20, IGHV6-1, and IGHG3), as well as those associated with complement activation (complement factor B) (Figure 2). Notably, CDL5, a protein expressed mostly by macrophages in lymphoid and inflamed tissue (55), was elevated in sera from donors with obesity relative to lean donors (Figure 2). This protein is also considered a biomarker for tuberculosis and liver cirrhosis (55). Despite the small size, these data revealed that sera profiles of lean donors are characterized by proteins that regulate lipid homeostasis (including HDL), immunity, and normal blood function (coagulation and angiogenesis), whereas those profiled from donors with obesity exhibited more inflammation and immune-associated proteins including those associated with immunoglobulin, complement activation, and pattern recognition receptors (CD5L). The top 50 differentially modulated proteins in our sera studies are presented in Supplementary Table 2 (available online).

Figure 2.

The sera of healthy adults is characterized by proteins involved in satiety, lipid metabolism, coagulation, and angiogenesis, whereas those associated with obesity contain high levels of proteins associated with immunity and inflammation. The sera of 6 adult donors (3 in the recommended body mass index range and 3 with obesity) were profiled using mass spectrometry. Proteins statistically significantly increased in healthy donors and those with obesity are indicated in red as determined by analysis using Spectronaut.

In addition to profiling sera from nonleukemic adult donors, we also analyzed plasma samples collected at diagnosis from pediatric patients with B-ALL. We used mass spectrometry to profile these samples based on their collection from lean patients (n = 7) and patients with obesity (n = 7) (Figure 3). The patient demographics are presented in Supplementary Table 3 (available online), and the top 50 differentially expressed proteins in our plasma studies are presented in Supplementary Table 4 (available online). Strikingly, our volcano plot analysis of plasma proteomic profiles revealed the most differentially expressed proteins were largely upregulated in lean patients, whereas only 4 proteins were more highly expressed in pediatric patients with obesity and B-ALL (Figure 3). Of these 4 proteins, 2 (50%) proteins were associated with immunoglobulin proteins, which is consistent with observations made in adult sera analyzed from donors with obesity relative to lean donors (Figure 2). Proteins found at higher circulating levels in the plasma of lean pediatric patients with B-ALL relative to those with obesity were similar to those found in lean mice (Figures 1, C-F); heat shock proteins (HSPA8, HSP90B1, HSPA8) were exclusively observed in the plasma of lean pediatric patients with B-ALL (Figure 3). Interestingly, cytoskeletal proteins (TUBA1B, TUBB, ACTG1, ACTN4, ACTA2, ACTG2, and ACTB) were also detected exclusively in the plasma of lean patients. Soluble cytoskeletal proteins are emerging as biomarkers for many diseases including solid cancers (56-58), and our results suggest that similar consideration should be given for B-ALL. Notably, circulating cytoskeletal proteins were absent in pediatric patients with B-ALL and obesity. Given that obesity decreases survival outcomes in pediatric patients with this leukemia subtype (59-61), the systemic absence of these proteins in patients may indicate a more aggressive disease and a poor prognosis. Hemoglobin-associated proteins (HBB and HBA2) were also higher in the plasma of lean relative to patients with obesity and B-ALL, which suggests that increased oxygen availability in pediatric patients with B-ALL may be influenced by adiposity (Figure 3). Other indicators of altered circulatory function dictated by adiposity and B-ALL included the identification of CALM1, CALM2, and CALM3 in the plasma of lean pediatric patients with B-ALL (Figure 3). This complex encodes for calmodulin (62) and, when perturbed, is associated with long QT syndrome, which is a heart signaling disorder that causes fast, chaotic heartbeats or arrhythmias (62). In all, these results highlight a striking dichotomy between the plasma of lean patients and patients with obesity and B-ALL. Notably, our results suggest that heat shock proteins may be predictive of disease severity and overall survival outcomes, that cytoskeletal proteins could be reliable biomarkers for a state of disease, and that cardiovascular health in patients with B-ALL could be assessed relative to circulating levels of hemoglobin-associated or CALM1, CALM2, and CALM3 proteins (63,64).

Figure 3.

The plasma of lean pediatric patients with B-ALL is exclusively characterized by the presence of heat shock proteins, cytoskeletal proteins, and proteins that promote cardiovascular health. The plasma of 14 pediatric patients with B-ALL (7 in the recommended body mass index range and 7 with obesity) was profiled using mass spectrometry. Proteins statistically significantly increased in healthy donors and those with obesity are indicated in red as determined by analysis using Spectronaut. B-ALL = B-cell acute lymphoblastic leukemia.

The adipocyte secretome contains an abundance of metabolites that promote cancer growth

Given the changes in serum and plasma profiles associated with obesity in nondiseased donors and patients with B-ALL, we next performed mass spectrometry analysis to examine metabolites directly secreted by adipocytes. For these experiments, we compared the adipocyte and bone marrow stromal cell secretomes relative to UCM (control). We chose to analyze bone marrow SCM because of stromal cells’ reported role in promoting B-cell malignancies (65); however, we have recently reported that factors secreted by these cells do not confer chemoresistance to human B-ALL cells (18).

This analysis yielded the high-confidence identification of 168 metabolites detected in the samples tested (Supplementary Table 5, available online). A heatmap was generated of metabolites differentially expressed in ACM, bone marrow SCM, and UCM or DMEM, which highlighted notable changes in metabolites (present and absent) in the adipocyte secretome relative to the other conditions tested (Supplementary Figure 4, available online). Principal component analysis corroborated the distinctive metabolomic profile of the adipocyte secretome relative to SCM and DMEM (Figure 4, A). We next generated a heatmap of the top 25 differentially secreted metabolites from the conditions tested (Figure 4, B). Notably, asparagine levels were highly elevated in ACM (Figure 4, B). L-asparaginase (ASNase) is a first-line therapy for ALL, and obesity in mice substantially impairs the efficacy of this drug (20). Despite not being elevated in the plasma of pediatric patients with obesity who were treated for high-risk ALL (20), concentrated local levels of this metabolite in adipose-rich tissue where residual disease is common (eg, the bone marrow) could impact minimal residual disease levels after treatment. In other words, adipocyte-secreted asparagine could increase resistance to ASNase treatment and allow for increased minimal residual disease levels after treatment in patients with obesity. Proline levels were also increased in ACM relative to SCM and UCM (Figure 4, B). When administered exogenously to cancer cells, this nonessential amino acid is documented to play a major role in tumorigenesis by alleviating endoplasmic reticulum (ER) stress, augmenting clonogenicity, and increasing tumorigenic potential (66). Mechanistically, proline promotes mechanistic target of rapamycin complex 1 activity, which in turn augments translation through the induction of the eukaryotic translation initiation factor 4E–binding protein 1 (66). In addition to asparagine and proline, aspartate is also found at high levels in the adipocyte secretome (Figure 4, B). Recently, aspartate has been shown to be imported into cancer cells through the SLC1A3 receptor (67). This uptake allows cancer cells to maintain a high proliferative capacity under hypoxic conditions, and high surface expression of SLC1A3 is associated with cancer cells having a competitive advantage in low oxygen conditions and in tumor xenograft experiments (67). Adipocytes also secreted metabolites that are highly cytotoxic to T cells. Notably, 3-hydroxyanthranilic acid, which is generated during tryptophan metabolism, induces T-cell death by inhibiting pyruvate dehydrogenase kinase 1 activity, which subsequently suppresses NF-κB activation in antigen-stimulated T cells (68).

Figure 4.

Adipocytes secrete metabolites, which may promote the growth of cancer cells. A) Principal component (PC) analysis was performed on all high-confidence metabolites identified in Dulbecco’s Modified Eagle Medium (DMEM; green), bone marrow stromal cell-conditioned media (SCM; blue), or adipocyte-conditioned media (ACM; red). n = 5 supernatant cultures per condition. The 2-dimensional plot of component 1 vs component 2 is shown with 95% confidence intervals for the groups. B) A heatmap was generated of the top 25 differentiating metabolites, from a list of statistically significantly different metabolites by analysis of variance. C) Pathway enrichment analyses were performed by Metaboanalyst on metabolites identified by Tukey post hoc analysis to differ between ACM vs DMEM (black), SCM s DMEM (grey), and ACM vs SCM (white). Only pathways with any P value less than .05 were included.

In addition to assessing the potential impact of notable adipocyte-secreted metabolites on tumorigenesis, we performed pathway enrichment analyses using Metaboanalyst on metabolites identified by Tukey post hoc analyses (Figure 4, C). Compared with metabolites found in UCM (DMEM) and SCM, adipocytes had a differential expression of metabolites associated with alanine, aspartate, and glutamate metabolism; valine, leucine, and isoleucine biosynthesis; and arginine and proline metabolism, among others (Figure 4, C). In all, these results demonstrate that ACM contained differential amounts of essential and nonessential amino acids and other metabolites, which can fuel cancer progression directly and indirectly by impacting the function of cancer cells and immune cells.

Adipocyte-secreted factors statistically significantly alter the metabolic state of human B-ALL cells in the absence and presence of MTX

Given the extensive metabolic differences observed between the secretomes of UCM (RPMI-1640), SCM, and ACM, we hypothesized that exposing human B-ALL cells to cell type–specific secreted factors would elicit distinctive metabolic programs in leukemia cells. Indeed, exposing REH and RCH-AcV cells (human B-ALL cell lines) to each condition for 1 day was sufficient to differentially alter the metabolome of leukemia cells in the absence and presence of MTX (Figure 5; Supplementary Figures 5 and 6, available online). The impact of the adipocyte secretome on the metabolic state of human B-ALL cells was notably different as evidenced by our principal component analyses (PCAs), which demonstrated that in the absence (Figure 5, A; Supplementary Figure 5, A, available online) and presence (Figure 5, B; Supplementary Figure 5, B, available online) of MTX, human B-ALL cells adopted a distinctive metabolic state. To gain a better understanding of adipocyte-induced metabolic changes in human B-ALL cells, we generated a heatmap of the top 25 differentially expressed metabolites in human cells regulated by each condition. When human B-ALL cells were exposed to the adipocyte secretome in the absence of MTX, notably, asparagine levels were high compared with SCM-exposed cells, whereas glutamine and most phosphatidylcholine species were low (Figure 5, C and E; Supplementary Figure 5, C and E, available online). Alterations in the asparagine-glutamine axis, with glutamine being consumed to maintain high levels of asparagine in leukemia cells, are consistent with the protective effect of these amino acids against ASNase treatment in vitro and in obese mice (20). We performed a similar analysis of human B-ALL cells treated with MTX in each condition. Of the top 25 metabolites changed in B-ALL by ACM treatment, we observed that the adipocyte secretome still promoted the intracellular accumulation of asparagine in both human B-ALL cell lines tested (Figure 5, D; Supplementary Figure 5, D, available online). Phosphatidylcholine species in ACM-cultured human B-ALL cells were low in the absence and presence of MTX treatment, except for phosphatidylcholine (36:2), which increased with MTX exposure (Figure 5, D and F; Supplementary Figure 5, D and F, available online). Despite this finding, the clinical significance of adipocyte-mediated lipid remodeling in B-ALL cells is unknown at this time. Additionally, an in-depth analysis of metabolic changes in human B-ALL cells induced by the secretomes of adipocytes and bone marrow stromal cells in the absence (Supplementary Figure 6, A and C, available online) and presence of MTX (Supplementary Figure 6, B and D, available online) revealed that adipocyte-secreted factors increase choline in leukemia cells treated with this chemotherapy (Supplementary Figure 6, B and D, available online). This adipocyte-mediated induction is notable, given that increased choline enhances the proliferative activity of neuroepithelial brain tumors (69) and promotes drug resistance in human breast cancer cells (70). Therefore, we speculate that elevated levels of this metabolite in adipocyte-exposed, MTX-treated human B-ALL cells may also contribute to adipocyte-induced chemoresistance in leukemia cells that we described in our previous study (18).

Figure 5.

Adipocytes induce metabolic changes within B-ALL cells in both the absence and presence of chemotherapy. Adipocyte-conditioned media (ACM) treatment is associated with increased glutaminolysis, β-oxidation, and glutathione metabolism, among others. Principal component analysis was performed on all high-confidence metabolites identified in REH cells (human B-ALL cells) treated with RPMI 1640 Medium (RPMI; green), bone marrow stromal cell- conditioned media (SCM; blue), or ACM (red) either (A) without or (B) with MTX. n = 3 biological replicates per group. The 2-dimensional plot of component 1 vs component 2 is shown with 95% confidence intervals for the groups. A heatmap was generated of the top 25 differentiating metabolites, from a list of significantly different metabolites by analysis of variance, in REH cells treated with the conditioned medias either (C) without or (D) with MTX. Pie charts are shown that indicate the number of phosphatidylcholine species (PCs) that differ between 2 conditioned media groups according to a Tukey post hoc analysis in REH cells treated with the conditioned medias either (E) without or (F) with MTX. Additional results, and from RCH-AcV cells, are shown in Supplementary Figures 2 and 3 (available online). A list of all differentially detected intracellular metabolites in REH cells can be found in Supplementary Tables 6 and 7 (available online). B-ALL = B-cell acute lymphoblastic leukemia; DMEM = Dulbecco’s Modified Eagle Medium; MTX = methotrexate

To gain a more comprehensive understanding of the relationship between differentially expressed metabolites, we performed pathway enrichment analyses using Metaboanalyst on metabolites identified by Tukey post hoc analyses (Supplementary Figure 6, E-G, available online). In the absence of MTX treatment, the adipocyte secretome statistically significantly affected metabolic pathways in human B-ALL cells associated with glycerophospholipid metabolism and, to a lesser extent, alanine, aspartate, and glutamine metabolism (Supplementary Figure 6, F, available online). In the presence of MTX, the adipocyte secretome affected metabolic processes in human B-ALL cells, which are associated with arginine biosynthesis; alanine, aspartate, and glutamate metabolism; nitrogen metabolism; D-glutamine and D-glutamate metabolism; and glutathione metabolism, among others (Supplementary Figure 6, E and G, available online). The alteration of these pathways in human B-ALL cells exposed to the adipocyte secretome suggests that leukemia cells may activate metabolic programs, which allow for the usage of alternative energy sources and those that buffer against oxidative stress (glutathione metabolism). In all, the results demonstrate that the adipocyte secretome induces specific metabolic programs in human B-ALL, which may promote more aggressive disease in the context of chemotherapy treatment.

Adipocyte-secreted factors statistically significantly upregulate genes associated with chemoresistance in human B-ALL cells

To determine if transcriptomic changes mimicked metabolic changes in human B-ALL cells exposed to the adipocyte secretome, we performed RNA-seq analysis of cells treated as previously described in the absence and presence of MTX.

We first analyzed the impact of UCM, SCM, and ACM on human B-ALL cells using PCA analysis, which revealed cell line–specific changes in gene expression programs in the conditions tested (Figure 6, A). To identify differentially expressed genes, we compared gene expression profiles of human B-ALL cells cultured in SCM relative to RPMI (control) and ACM relative to RPMI (Figure 6, B). The adipocyte secretome had the greatest impact on gene expression programs in human B-ALL cells with a greater than 1.5-fold change in either direction observed (Figure 6, B). Although expression profiles differed between RCH and REH (as shown by PCA; Figure 6, A), we identified 2 common cancer-promoting genes in the top upregulated set of genes (Figure 6, B). Notably, the Wnt family member wingless-type mouse mammary tumor virus (MMTV) integration site family member 16 (WNT16), which is associated with cellular stemness, proliferation, and chemoresistance in prostate cancer (71), is upregulated in human B-ALL cells exposed to the adipocyte secretome. In the prostate cancer tumor microenvironment, WNT16B is upregulated in response to chemotherapy-induced genotoxic stress, secreted into the niche, and protects cancer cells from cell death by activating canonical Wnt signaling (71). In addition to paracrine signaling, autocrine Wnt signaling maintains a population of leukemia-initiating cells in acute myeloid leukemia (72-74). Tribbles homolog 2 (TRIB2) was also upregulated in both human B-ALL cell lines when cultured in ACM (Figure 6, B). This protein kinase signals through canonical mitogen-activated protein kinase (MAPK) and Ak strain transforming (AKT) pathways, which promote survival and chemoresistance when highly expressed in malignant cells (75,76).

Figure 6.

The adipocyte secretome induces transcriptional changes in human B-ALL cells REH and RCH, which are associated with chemoresistance prior to treatment with MTX. A comparison of RNA-seq expression profiles of leukemia cells cultured in adipocyte-conditioned media (ACM) or bone marrow stromal cell-conditioned media (SCM). Plots show normalized values, generated by RNA-sequencing analysis software DEseq, for 3112 genes with expression levels greater than zero in at least 1 condition. A) Principal component analysis (PCA) of all conditions for both cell lines. B) Pairwise comparison of ACM or SCM vs RPMI for RCH cells (left) and REH cells (right). Red: fold change ≥1.5; Blue: fold change ≤−1.5. Genes with the highest or lowest fold change values for ACM vs RPMI are listed below each pair of plots. *Top upregulated genes shared between ACM-treated RCH and REH samples. The ranked log results are provided in Supplementary Table 10 (available online). RNA-seq = RNA sequencing.

A similar analysis was performed on MTX-treated human B-ALL cells cultured in UCM, SCM, and ACM. A PCA of MTX-treated B-ALL cells revealed that expression profiles still cluster cell type rather than by culturing conditions (Figure 7, A); however, the profiles of human B-ALL cells cultured in ACM are distinctive from the SCM and RPMI clusters, which exhibited similar transcriptomes. These results suggest that MTX treatment induces additional gene expression changes in human B-ALL cells exposed to the adipocyte secretome (Figures 6, A, and 7, A). A pairwise comparison of each conditioned media treatment group relative to profiles observed in human B-ALL cells cultured in RPMI (control) revealed that a larger number of genes become downregulated in leukemia in MTX-treated (Figure 7, B) relative to untreated cells (Figure 6, B). Notably, the top upregulated genes that were induced in B-ALL cells by ACM alone (Figure 6, B) remain upregulated with MTX treatment (Figure 7, B). This observation highlights that specific transcriptional programs are activated in human B-ALL by the adipocyte secretome, which primes cells before MTX treatment. Furthermore, it demonstrates that some of these changes cannot be reversed by MTX treatment. Notably, WNT16 and TRIB2 remain upregulated in ACM-cultured, MTX-treated human B-ALL cells (Figure 7, B), which adds further support for the involvement of genes in adipocyte-mediated chemoresistance to MTX and perhaps other chemotherapies. The top downregulated genes, which were noted in both MTX-treated cell lines tested in the presence of ACM, included activating transcription factor 3 (ATF3), sestrin-2 (SESN2), and serpin family E member 1 (SERPINE1). Downregulation of ATF3 suggests that human B-ALL may become more inflammatory when exposed to the adipocyte secretome and treated with MTX, given that this transcription factor is a negative regulator of inflammation (77). Furthermore, the downregulation of ATF3 may contribute to metabolic shifts observed in human B-ALL cells exposed to MTX in ACM (Figures 5; Supplementary Figures 5 and 6, available online), given its noted role as a master regulator of metabolic homeostasis, including glucose metabolism, in solid cancers (77). The downregulation of SESN2 may reduce proliferation in ACM-treated B-ALL cells and reduce their response to genotoxic stress (78), which would mitigate the cytotoxic effects of MTX. Additionally, high SERPINE1 levels drive increased cancer cell proliferation (79); therefore, downregulation of this gene in ACM-treated human B-ALL cells may also contribute to reduced proliferation and the acquisition of chemoresistance to MTX. The downregulation of these genes supports our recently published observations that human B-ALL cells exposed to the adipocyte secretome are capable of withstanding elevated levels of genomic stress and undergo a senescence-like, reduced proliferative state in response to adipocyte-secreted factors (18).

Figure 7.

The adipocyte secretome induces transcriptional changes in human B-ALL cells (REH and RCH), which indicate cellular quiescence, reduced tumor suppression, increased survival, and chemoresistance when treated with MTX. RNA-seq expression analysis of adipocyte-conditioned media (ACM) and bone marrow stromal cell-conditioned media (SCM) cultured leukemia cell responses to MTX. Plots show normalized values generated by the RNA-sequencing analysis software DEseq for 3112 genes with expression levels greater than zero in at least 1 condition. A) Principal component analysis (PCA) of all conditions for both cell lines. B) Pairwise comparison of ACM or SCM vs RPMI 1640 media (RPMI) for two human B-ALL cell lines, RCH cells (left) and REH cells (right). Red: fold change ≥1.5; Blue: fold change ≤−1.5. Genes with the highest or lowest fold change values for ACM vs RPMI are listed below each pair of plots. *Top upregulated genes shared between ACM-treated RCH and REH samples. **Top downregulated genes shared between ACM-treated RCH and REH samples. C) The Venn diagram compares genes exclusively changed in ACM-cultured cells before and after MTX treatment. The ranked log results are provided in Supplementary Table 10 (available online). B-ALL = B-cell acute lymphoblastic leukemia; MTX = methotrexate; RNA-seq = RNA sequencing; PC = principal component; DEG = differntially expressed genes; GO = gene ontology.

Pathway analysis of 130 genes commonly upregulated (n = 11) or downregulated (n = 119) in ACM-cultured, MTX-treated human B-ALL cells (Figure 7, C) revealed the upregulation of pro-cancer pathways including Wnt signaling (MYC) and fatty acid synthesis (Figure 7, D). Downregulated pathways included those involved in tumor suppression (p53), apoptosis (tumor necrosis factor receptor [TNFR] signaling/first apoptotic signal [FAS]), integrin signaling (laminin subunit alpha-3 [LAMA3], actin alpha 2 [ACTA2]), and interleukin signaling (interleukin-23A [IL-23A], cyclin dependent kinase inhibitor 1A [CDKN1A]) (Figure 7, D). Statistical analysis using gene ontology (GO) term enrichment of the 119 downregulated genes further highlighted that suppression of TNF-related apoptosis-inducing ligand (TRAIL)-activated apoptotic signaling, intrinsic apoptotic signaling, the DNA damage response, the integrated stress response, and ER stress contribute to chemoresistance to MTX in B-ALL cells exposed to the adipocyte secretome (Figure 7, D). In all, these analyses highlight novel genomic contributors of adipocyte-mediated chemoresistance in human B-ALL cells.

Adipocyte-secreted factors statistically significantly upregulate proteins associated with glutaminolysis, β-oxidation, B-progenitor cell development, and chemoresistance in the absence and presence of chemotherapy treatment

To determine the relationship between adipocyte-induced transcriptomic changes and alterations in the proteome of human B-ALL cells in the absence and presence of MTX treatment, we performed mass spectrometry on leukemia cell lysates from each condition. In the absence of chemotherapy treatment, notable ACM-induced proteins in human B-ALL included upregulated proteins in ACM-cultured B-ALL cells including glutamate-ammonia ligase (GLUL), dynein axonemal heavy chain 8 (DNAH8), carnitine palmitoyltransferase 1A (CPT1A), S100 calcium-binding protein A11 (S100A11), and methylenetetrahydrofolate dehydrogenase (MTHFD1L), among others (Figure 8, A). Notably, only GLUL and DNAH8 proteins were increased in both human B-ALL cell lines cultured in ACM (Figure 8, A; Supplementary Figure 7, A, available online), suggesting that the induction of these proteins is specific to the adipocyte microenvironment. High levels of DNAH8 and MTHFD1L are associated with poor survival outcomes for several solid cancer types because of their ability to help cells mitigate chemotherapy-induced oxidative stress (80-83). Furthermore, increased protein levels of S100A11 are also associated with chemoresistance in solid cancers (84). GLUL and CPT1A activate metabolic programs in solid tumors, which promote tumor development, by increasing glutaminolysis (85,86) and fatty acid oxidation (87), respectively. Notable downregulated proteins in ACM-cultured human B-ALL cells included spleen tyrosine kinase (SYK) and karyopherin alpha 2 (KPNA2), which may reduce proliferation in B-ALL cells (88,89). Furthermore, caldesmon (CALD1) and lymphoid-restricted membrane protein (LRMP) were also downregulated, which may reduce the immunogenicity of human B-ALL when exposed to adipocyte-secreted factors (90,91). Given the cell specificity observed in the proteomic analysis (Figure 8; Supplementary Figures 7-9, available online), it was not surprising that the only protein that was downregulated in both human B-ALL cells when cultured in ACM was late endosomal/lysosomal adaptor, MAPK, and mTOR activator 1 (LAMTOR1). This protein, when inhibited in cancer cells, also slows tumor cell proliferation (92), which would reduce the efficacy of chemotherapy treatment.

Figure 8.

Adipocyte-secreted factors statistically significantly upregulate proteins associated with glutaminolysis, β-oxidation, B-progenitor cell development, and chemoresistance in the absence and presence of chemotherapy treatment. Mass spectrometry analysis was performed on unconditioned media (UCM), bone marrow stromal cell-conditioned media (SCM), and adipocyte-conditioned media (ACM)–cultured human B-ALL cell lines RCH and REH in absence and presence of MTX treatment. Differentially expressed proteins in REH cells are shown with the other treatment conditions and results for RCH-AcV cells shown in Supplementary Figures 4-6 (available online). The MaxQuant software (Max Planck Institute of Biochemistry) was used for database searches, and the R program was used to determine statistical significance. The differentially expressed intracellular proteins for REH cells in ACM are shown in Supplementary Tables 11 and 12 (available online). B-ALL = B-cell acute lymphoblastic leukemia; MTX = methotrexate; NS = not statistically significant; FC = fold-change; nomp-value = normalized p-value.

When treated with MTX, like the observed transcriptomic results, more proteins were altered in both ACM-treated human B-ALL cell lines tested (Figure 8; Supplementary Figure 7, available online). Given that overlapping response may indicate proteins involved in the induction of chemoresistance, we focused on those that were upregulated and downregulated in both MTX-treated cell lines when cultured in ACM. Again, GLUL, which was upregulated with ACM exposure, remained highly expressed in MTX-treated cells (Figure 8; Supplementary Figure 7, available online). This observation further highlights the adipocyte-mediated metabolic switch to glutaminolysis in the presence and absence of drug treatment. The protein DNAH8 was also increased in drug-treated cells (Figure 8; Supplementary Figure 7, available online), like responses observed in ACM-cultured B-ALL not exposed to chemotherapy treatment (Figure 7; Supplementary Figure 6, available online). Other proteins upregulated in both human B-ALL cell lines cultured in ACM and treated with MTX included histone H1.3 (HIST1H1D), perilipin-2 (PLIN2), paired-box family protein 5 (PAX5), and fatty acyl-CoA reductase 1 (FAR1). Of these, an increase in PLIN2 is associated with inhibiting insulin-induced glucose uptake (93) and promoting the proper hydrolysis of lipid droplets (LD) (94). This LD protein promotes LD formation, stability, and trafficking of LDs (94). The PAX5 protein is a critical regulator of B lymphopoiesis (95,96) and, when deregulated, contributes to B-ALL formation (97). In the context of MTX treatment, the upregulation of PAX5 in ACM-treated human B-ALL cells may prevent excessive genotoxic stress, given that PAX5 acts as a checkpoint in normal B-cell maturation (97). Shared downregulated proteins in MTX-treated, ACM-cultured human B-ALL cells included flywch2 family member 2 (FLYWCH2) and ribosomal proteins (RPS15A) (Figure 8; Supplementary Figure 7, available online), with a decrease in the latter suggesting that human B-ALL cells may be downregulating translation when exposed to MTX in the adipose-rich microenvironment (which may be consistent with reduced proliferation). In all, our proteomic analysis nominated several proteins, which may contribute to adipocyte-mediated chemoresistance in human B-ALL cells.

Discussion

In many cases, obesity promotes the development of more aggressive cancers, which are more difficult to treat (98). This relationship between obesity and the progression of many solid tumors is well-established (99-101), whereas our understanding of this connection with hematological malignancies has grown considerably over the past decades (12-14,18,20,59,60,102). Our continued efforts to understand how obesity impacts cancer progression is of paramount importance given that obesity is approaching pandemic levels in the United States, where models estimate that more than 50% of the US population will present with overweight or obesity by 2030 (2,103,104), with similar trajectories predicted worldwide (2,105).

In this study, we took an ambitious multi-omics approach to understand the impact of obesity and increased adiposity on B-ALL. Importantly, we sought to identify changes occurring in human B-ALL cells exposed to the adipocyte secretome in the absence and presence of MTX, which is commonly used in all phases of treatment for this leukemia subtype (106,107). Our approach consisted of using scRNA-seq to determine how obesity alters immune cells found in the adipose tissue of mice fed LFLS (lean mice) or HFHS diets (obese mice). These studies revealed a unique signature in the B-cell population, which corresponded with a reduced frequency of immunologically active nonmalignant B cells. Furthermore, this B-cell signature observed in mice has prognostic value, where we observe that patients with B-ALL who had a reduction in this signature have statistically significantly worse survival outcomes relative to patients with high expression of the top 10 genes found in this B-cell population.

In addition to our murine diet-induced obesity studies, we also performed mass spectrometry to define how obesity impacts the circulating proteome of donors without and with B-ALL. Major limitations to these studies are the relatively small sample sizes and not having age- and sample-matched comparisons for disease-free adults and pediatric donors with B-ALL. Despite these limitations, the data suggest distinct serological signatures associated with obesity in healthy individuals and patients with B-ALL. In healthy adult sera, proteins found to be at high levels in circulation were those that modulate lipid homeostasis (apolipoprotein D, apolipoprotein A4, adiponectin, phospholipid transfer protein) and those that regulate various biological processes including immunity (histidine-rich glycoprotein). In contrast, in adults with obesity, circulating sera profiles were heavily skewed toward a signature that suggested heightened B-cell activation via the presence of immunoglobulin-associated proteins (IGLV3-27, IGHD, IGLV5-45, IGKV6D-21, etc.), those involved in T-cell activation (CD44), complement activation (CFB), and those associated with chronic inflammation (CD5L). These B-cell–mediated changes in the circulating profiles of immunoglobulins may result from metabolic changes in normal B cells induced by obesity-associated chronic inflammation or changes in fatty acid metabolism (108).

The plasma profiles of pediatric patients with B-ALL differed statistically significantly from the sera of adults. Furthermore, circulating proteins in lean pediatric patients with B-ALL were notably different from those with obesity, highlighted by the detection of a statistically significant number of proteins that were exclusively observed in lean patients. Of these, a large majority of the statistically significantly identified proteins were heat shock proteins (HSP90AB1, HSPAB HSP90B1), cytoskeletal proteins (tubulins TUBA1B, TUBB; actin ACTG1, ACTN4, ACTA2, ACTG2, and ACTB), hemoglobin-associated proteins (HBB and HBA2), and those associated with good cardiovascular health (calmodulin complex CALM1/2/3). Interestingly, heat shock protein and cytoskeletal proteins are being studied as biomarkers for various cancers (109,110); therefore, similar consideration should be given to these proteins in the context of B-ALL given weight-associated survival differences observed in pediatric patients with this leukemia subtype (12,61,98).

Given how obesity impacts nonmalignant B cells in mice and systemic factors in humans without and with B-ALL, we next determined which metabolites were found in the adipocyte secretome relative to bone marrow stromal cells and UCM. Mass spectrometry analyses of the conditioned media relative to the base media (control) revealed that the adipocyte secretome was distinctive from the other groups tested with SCM and UCM having the greatest metabolite overlap. One limitation of the conditioned media study is that some of the metabolites detected in our conditioned media could represent differences in the abundances of FBS in the base media between SCM and ACM. However, these differences are likely to be negligible because of the abundance of metabolites that were detected, which are not base components of RPMI and DMEM. Notably, adipocytes secreted abundant metabolites documented to drive cancer progression, with perturbations of pathways of alanine, aspartate, and glutamate metabolism; valine, leucine, and isoleucine biosynthesis; and arginine and proline metabolism, among others.

In addition to assessing the metabolome of adipocytes and bone marrow stromal cells, as well as UCM, we also determined how each condition impacted the metabolic state of human B-ALL cells alone and when treated with MTX. These studies revealed exposure to the adipocyte secretome alone resulted in many metabolic shifts in human B-ALL cells, relative to the other conditions tested, with notable inductions of increased β-oxidation and glutaminolysis in leukemia cells. In the presence of MTX, ACM-induced metabolic changes in human B-ALL cells were more pronounced and extensive with the alteration of additional metabolic pathways involving glutathione metabolism, nitrogen metabolism, and arginine biosynthesis, among others. These results demonstrate that metabolic programs induced in adipocyte-exposed human B-ALL cells prime leukemia cells to withstand the cytotoxic effects of chemotherapies.

To determine the impact of the adipocyte secretome on gene expression profiles in human B-ALL cells, we performed RNA-seq analyses to determine transcriptional changes in the absence and presence of MTX treatment. Similar to metabolic changes observed in B-ALL cells exposed to the adipocyte secretome, transcriptional changes were more extensive in leukemia cells cultured in ACM and treated with MTX. Notably, the genes coding for WNT16 and TRIB2 were upregulated in both human B-ALL cell lines cultured in ACM in the absence of chemotherapy treatment. Furthermore, the expression of these genes remained high with MTX treatment. Shared genes being downregulated included those coding for ATF3, SERPINE1, and SESN2. Pathway analysis of transcriptional changes modulated in ACM-cultured B-ALL cells treated with MTX revealed that Wnt signaling and fatty acid oxidation were notable pathways that increased in leukemia cells, whereas the p53 pathway, apoptotic signaling, and stress response pathways were notable ones that decreased.

Mass spectrometry was performed to analyze the proteome of human B-ALL cells in the conditions mentioned above. Notably, in 1 of the 2 human B-ALL cell lines tested (REH cells), WNT16 was upregulated in ACM-cultured leukemia cells treated with MTX, which suggests that this protein may indeed be important for the induction of chemoresistance. Furthermore, proteins involved in glutaminolysis (GLUL), fatty acid oxidation (CPT1A), glucose regulation (PLN2), B-cell development (PAX5), and chemoresistance (DNAH8, MTHFD1L, and S100A11) were induced in human B-ALL cells in ACM-cultured B-ALL treated with MTX, where only GLUL and DNAH8 proteins were increased in both human B-ALL cells in response to the adipocyte secretome alone.

In all, our multi-omics approach allowed us to identify gene signatures that predict survival outcomes in patients with B-ALL, nominate biomarkers that may predict disease severity, identify novel mechanisms that may contribute to adipocyte-mediated chemoresistance in B-ALL (WNT16, DNAH8, PLN2, MTHFD1L, etc.), and confirm ones already documented to contribute to the protection of leukemia cells (β-oxidation and glutaminolysis). Importantly, the deposition of the presented scRNA-seq, RNA-seq, and mass spectrometry (metabolomics and proteomics) datasets will serve as valuable resources for additional data mining for B-ALL and for researchers interested in comparing responses to solid cancers.

Data availability

The datasets generated during and/or analyzed in the current study will be deposited in publicly available databases without restriction if the publication criteria for this manuscript are met.

Author contributions

Curtis J Henry, PhD (Conceptualization; Data curation; Formal analysis; Funding acquisition; Writing—original draft; Writing—review & editing), Joshua D. Chandler, PhD (Writing—review & editing), Manoj Bhasin, PhD (Writing—review & editing), Christopher D. Scharer, PhD (Writing—review & editing), Ryan Walsh, PhD (Writing—review & editing), Pritha Bagchi, PhD (Writing—review & editing), Eric Weaver, PhD (Writing—review & editing), Robert English, PhD (Writing—review & editing), Melissa L. Kemp, PhD (Writing—review & editing), Anne M. Fitzpatrick, PhD (Writing—review & editing), Rich Johnston, PhD (Writing—review & editing), Jeremy M. Boss, PhD (Writing—review & editing), Ayjha C. Brown (Writing—review & editing), Matthew B. Kilgore, PhD (Writing—review & editing), Marisol Castillo-Castrejon, PhD (Writing—review & editing), Michelle Salemi, MS (Writing—review & editing), Anthony J. Ross, MD (Writing—review & editing), Jessica Hoehner, MPH (Writing—review & editing), Swati Bhasin, PhD (Writing—review & editing), William Pilcher, BS (Writing—review & editing), Maya Jaffe (Writing—review & editing), Kirsten A. Cottrill, PhD (Writing—review & editing), Miyoung Lee, PhD (Writing—review & editing), Delaney K. Geitgey, BS (Writing—original draft; Writing—review & editing), Karmella A. Haynes, PhD (Writing—review & editing), and Elizabeth Wellberg, PhD (Writing—review & editing).

Funding

Metabolomic studies were supported by an NIH R01 (Grant No. NR018666) awarded to AMF and NIH R56 (Grant No. HL150658) awarded to JDC as well as in part by CF@LANTA (a component of Emory University and Children’s Healthcare of Atlanta). RNA-sequencing studies were supported by NIH R01 (Grant No. CA241156) awarded to EAW and the COBRE funding mechanism (Grant No. 5P30GM122744-04) awarded to MCC. With the exception of RNA-sequencing studies, all results presented in this study were supported by funding from the CURE Childhood Cancer Foundation (Grant No. 001006916), Swim Across America (Grant No. 00103163), The Mark Foundation for Cancer Research (Grant No. 18-031-ASP), Emory University School of Medicine Bridge Funding (Grant No. 00098174), The American Cancer Society and Emory University Winship Cancer Institute Institutional Research Grant (Grant No. IRG-21-137-07-IRG) and the TREC Training Course (Grant No. R25CA203650) awarded to CJH. This work was also supported by the Wallace H. Coulter Department of Biomedical Engineering at the Georgia Institute of Technology seed grant awarded to KAH and CJH.

Conflicts of interest

KAC received grant funding from NIH (T32 5T32HL116271-09). AMF received grant funding from NIH T32 HL116271; NIH K24NR018866. MLK declared scientific advisory board, personal equity with Parthenon Therapeutics. JDC declared grant funding from NIH and Cystic Fibrosis Foundation. The other authors declared no conflicts of interest for this study.