Elevated lactate in AML bone marrow contributes to macrophage polarization via GPR81 signaling

Key Points

• •GPR81, the lactate receptor, is a mechanism of LAM polarization to a suppressive phenotype.
• •Targeting GPR81 reduces macrophage polarization and has therapeutic potential for AML.

Visual Abstract

Abstract

Interactions between acute myeloid leukemia (AML) and the bone marrow microenvironment (BMME) are critical to leukemia progression and chemoresistance. In the solid tumor microenvironment, altered metabolite levels contribute to cancer progression. We performed a metabolomic analysis of bone marrow serum from patients with AML, revealing increased metabolites compared to age- and sex-matched controls. The most highly elevated metabolite in the AML BMME was lactate. Lactate signaling in solid tumors induces immunosuppressive tumor-associated macrophages and correlates with poor prognosis. This has not yet been studied in the leukemic BMME. Herein, we describe the role of lactate in the polarization of leukemia-associated macrophages (LAMs). Using a murine AML model of blast crisis chronic myelogenous leukemia, we characterize the suppressive phenotype of LAMs through surface markers, transcriptomics, and cytokine profiling. Mice genetically lacking GPR81, the extracellular lactate receptor, were then used to demonstrate GPR81 signaling as a mechanism of both the polarization of LAMs and the direct support of leukemia cells. Furthermore, elevated lactate diminished the function of hematopoietic progenitors and reduced stromal support for normal hematopoiesis. We report microenvironmental lactate as a mechanism of AML-induced immunosuppression and leukemic progression, thus identifying GPR81 signaling as an exciting and novel therapeutic target for treating this devastating disease.

Introduction

Acute myeloid leukemia (AML) is the most common acute leukemia in adults. It has a nearly 90% mortality rate at 5 years after diagnosis in the most affected group of patients aged >65 years.¹ AML is a hematologic malignancy initiated by genetic mutations in immature myeloid progenitor cells. During disease progression, leukemic cells proliferate rapidly, accumulating in the bone marrow (BM) and other tissues.² Leukemic-initiated dysfunction of the BM microenvironment (BMME) leads to a loss of normal hematopoiesis. This paucity of functional blood cells increases susceptibility to infection, hemorrhage, and BM failure, all critical factors in morbidity and mortality associated with AML.^3,4 Chemotherapies initially reduce leukemic burden; however, relapse occurs in most patients due to surviving leukemia stem cells (LSCs).5, 6, 7 With ∼20 000 new cases annually in the United States (Surveillance, Epidemiology, and End Results [SEER] Program and National Institutes of Health [NIH]) and an increasing global incidence,⁸ there is a clear, unmet need for novel treatments.

Signaling within the BMME directs hematopoietic stem cells (HSCs) to produce hematopoietic stem and progenitor cells (HSPCs). Multiple “HSC niche” cells regulate HSCs, including mesenchymal stromal cells (MSCs),^9,10 osteoblasts (OBs),^11,12 endothelial cells,13, 14, 15 and macrophages.¹⁶ HSC niche cells signal by contact-mediated mechanisms and secreted factors.^9,17, 18, 19, 20, 21 The dysfunctional BMME not only loses support for normal hematopoiesis but also can gain support for leukemogenesis and leukemic progression.22, 23, 24 Furthermore, LSCs have been found to gain resistance to therapy by residing within the altered niche.^25,26 Therefore, a thorough understanding of the microenvironment is needed to improve treatment.

Altered metabolite concentrations in the solid tumor microenvironment (TME) impede immune cells while supporting cancer.^27,28 Thus far, the extracellular metabolites in the AML BMME have not been well defined. A hallmark of cancer cells is energy production by aerobic glycolysis, called the “Warburg effect,” even with functional mitochondrial oxidative phosphorylation.²⁹ In a final step of glycolysis, lactate dehydrogenase (LDH) converts pyruvate to lactate. Extended glycolysis, therefore, demands secretion of excess lactate by cancer cells. Lactate concentrations have been reported to be elevated 5- to 30-fold in the TME and correlate with poor prognosis.^30,31 Though heterogeneous within an individual, AML cells are known to upregulate both glycolysis and oxidative phosphorylation, are dependent on glycolysis for survival,^32,33 and therefore would be expected to secrete lactate. Increased lactate in AML BM has recently been reported³⁴ but is not yet well documented. We hypothesized that metabolites such as lactate accumulate in the AML BMME, driving immunosuppression and leukemic progression.

Macrophages are critical to the stimulation of T cells during an immune response. Chronic signals in the TME, such as elevated lactate, alternatively polarize tumor-associated macrophages (TAMs); TAMs instead block T cells from targeting cancer cells. The presence of TAMs correlates with poor prognosis in multiple cancer types.35, 36, 37 An immune-suppressed BM is well-known in AML, yet lactate has never been directly connected to this. Leukemia-associated macrophages (LAMs) in AML have been reported to be alternatively activated, support leukemic transformation, and correlate with poor prognosis.^38,39 Repolarization of macrophages toward a more proinflammatory phenotype affects survival time in murine models of AML.^39,40 Yet, little is known about the molecular mechanisms of how LAMs are polarized to this phenotype in AML. This research aimed to determine whether elevated lactate in the BMME during myeloid leukemias contributes to suppressive macrophages.

Methods

Human BM serum collection

Deidentified BM aspirates were collected from patients or healthy donors and immediately centrifuged to remove cells. The supernatant was quickly taken to storage at –80°C until use. Patients were eligible if diagnosed de novo for AML.

Metabolomics by liquid chromatography–mass spectrometry

Liquid chromatography–mass spectrometry was performed using LC-20AD high-performance liquid chromatography (HPLC) system (Shimadzu). Mass spectrometric analyses were performed on a TSQ Quantum Ultra triple-quadrupole mass spectrometer running in multiple reaction monitoring mode (ThermoFisher Scientific). Peak heights for metabolite chromatograms were analyzed using the Xcalibur software (RRID: SCR_014593). See supplemental Methods for details.

Murine AML model (bcCML)

Blast crisis chronic myelogenous leukemia (bcCML), which includes lentiviral insertion of both the BCR::ABL translocation conjugated to green fluorescent protein (GFP) and Nup::Hox98 fusion conjugated to yellow fluorescent protein (YFP), has been previously characterized as an AML model.^23,41, 42, 43, 44 Primary bcCML was initiated via tail vein injection of bcCML cells into 8- to 12-week-old mice after 6.5-Gy sublethal irradiation. For comparative in vivo experiments, 10 000 wild-type (wt) or Gpr81^−/− primary bcCML spleen cells per 100-μL dose were injected into age- and sex-matched wt or Gpr81^−/− mice.

Flow cytometry and FACS

Flow cytometric analyses were performed at the University of Rochester Wilmot Cancer Center on an LSRFortessa Cell Analyzer (BD Biosciences). Fluorescence-activated cell sorting (FACS) was performed at the Flow Cytometry Core at the University of Rochester Medical Center on a FACSAria II system (BD Biosciences) with an 85-micron nozzle at 4°C, using FACSDiva software (BD Biosciences; RRID: SCR_001456), and data were analyzed using FlowJo version 10 software (BD Biosciences; RRID: SCR_008520). Antibody details are listed in supplemental Methods Table 1.

RNA sequencing

The University of Rochester Genomics Research Center performed RNA sequencing and analysis. See supplemental Methods for details.

Cytokine profiling

Macrophages were sorted via FACS from the BM of bcCML or healthy mice, then cocultured on a stromal monolayer from a healthy mouse (see supplemental Methods for details) for 4 days to establish a microenvironment. Media were collected and stored immediately at −20°C until use. Proteome Profiler Mouse XL Cytokine Array (R&D Systems ARY028; Bio-Techne)) was used according to the manufacturer’s protocol and then analyzed using the ChemiDoc MP Imaging System (Bio-Rad) and Image Lab software (Bio-Rad). Images were equally adjusted for background and qualitatively assessed for cytokine presence.

BMDM production

BM-derived macrophage (BMDM) production methods were adapted from a previously published protocol.⁴⁵ Whole BM was plated in vented tissue culture–treated 75 cm² flasks (NEST 708001) overnight in complete Dulbecco’s modified Eagle medium (Corning 10-013-CV). The next day, nonadherent cells were transferred to 12-well tissue culture–treated culture dishes at 400 000 cells per well in 1-mL media with 25 ng/mL recombinant murine macrophage colony-stimulating factor (PeproTech 315-02), in which they then adhered. Media were changed on day 4, and differentiation to macrophages was complete by day 7.

Macrophage polarization experiments

BMDMs were cultured in media without fetal bovine serum, starting 6 hours before treatment. Treatments were done for 48 hours or for the time indicated. Treatments included 10 mmol/L sodium L-lactate (Sigma-Aldrich), 100 ng/mL lipopolysaccharides (LPSs) from Escherichia coli O55:B5 (L6529S; Sigma-Aldrich), and/or 5 ng/mL of recombinant murine interleukin-4 (IL-4) and IL-13 (214-14 and 210-13; PeproTech). To remove cells for analysis, 0.25% trypsin-EDTA (ThermoFisher Scientific; Gibco; 25200056) was added for 5 minutes, and then cell scrapers were used.

CFU-C assays

LSKs were sorted via FACS then added to the cultures; specific colony-forming unit (CFU) experiments are described in supplemental Methods. To plate for CFU-cultures (CFU-C), adherent and nonadherent cells were collected in a 15-mL conical tube, in which 0.25% trypsin-EDTA was used on the adherent cells for 5 minutes and then flushed with media. Cells were pelleted by centrifugation at 900g for 5 minutes and then resuspended in 0.5-mL minimum essential medium (MEM) α. Both 1:20 and 1:100 dilutions were made in 2.5-mL media, and then 0.2 mL of these were resuspended each in 2.5-mL aliquots of MethoCult (M3434; Stemcell Technologies) and immediately plated in duplicates of 1.2 mL into 35 × 10 mm sterile suspension culture dishes (430588; Corning). Dishes were placed inside a 150 × 25 mm dish (715001; Nest), with an open dish of water to prevent drying. They were incubated for 14 days (5% CO₂; 37°C), and colonies were counted.

Approval by the institutional animal care and use committee and the institutional review board was obtained for this research. All facilities and animal care comply with federal and NIH policies, accredited by the Association for Assessment and Accreditation of Laboratory Animal Care International. All patients and volunteers provided written, informed consent on protocols approved by the Research Subjects Review Board of the University of Rochester in accordance with the Declaration of Helsinki.

Results

Extracellular metabolite levels are altered in the BM of patients with AML, including elevated lactate

To profile extracellular metabolite levels in the AML BMME, we performed metabolomics on serum from BM biopsies of patients with AML at diagnosis, as well as healthy age- and sex-matched controls (Figure 1A). The AML samples included different ages, sex, and mutational subtypes (supplemental Table 1). AML BM displayed a general increase in extracellular metabolites (Figure 1B-C). Six metabolites, listed in Table 1, were significantly altered during disease. Of these, lactate was the most highly elevated. Lactate concentrations were ∼2-5-fold higher in AML BM, measured at 2 to 5 mmol/L (Figure 1D). In vivo concentrations are likely greater because peripheral blood unavoidably dilutes BM biopsies. Furthermore, we postulate that lactate concentrations are higher near AML cell–dense pockets, because the BM is spatially heterogeneous for concentrations of similar molecules and pH.⁴⁶

Figure 1.

Lactate is elevated in the BMME during AML. (A-C) Metabolomics of BM serum from patients with AML or normal controls: (A) graphical depiction of procedure; (B) heat map showing the relative abundance of detectable metabolites; (C) and scores plot of principal component analysis (n = 4; in triplicates). (D) Lactate concentration in AML and normal BM serum (n = 4). (E) Graphical depiction of bcCML cell generation and disease initiation. Significance levels were determined by unpaired t test for panels B,D and are indicated as follows: ∗P < .05; ∗∗P < .01; ∗∗∗P < .001; ∗∗∗∗P < .0001. Error bar indicates mean ± standard deviation (SD). 1,3-BPG, biphosphoglycerate; DHAP, dihydroxyacetone phosphate; G3P, glyceraldehyde-3-phosphate; LC-MS, liquid chromatography–mass spectrometry.

Table 1.

Significantly altered extracellular metabolites in human AML BM

Compound	t stat	P value	−log10(P)	FDR
Lactate	5.8818	6.42E-06	5.1922	0.000224
Citrulline	−4.0561	.00052604	3.279	0.0092
Serine	3.748	.0011128	2.9536	0.00974
DHAP	3.6597	.0013779	2.8608	0.00974
Ribose-P	3.6554	.0013923	2.8563	0.00974
Ornithine	3.4174	.0024667	2.6079	0.0143

DHAP, dihydroxyacetone phosphate; FDR, false discovery rate.

Next, we aimed to study BM lactate signaling in vivo. We performed metabolomics on BM extracellular fluid from a murine AML model, bcCML (Figure 1E).^23,44 This model displays a rapid accumulation of myeloid leukemia cells, such as de novo AML. BM metabolites, including lactate, were elevated in bcCML BM (supplemental Figure 1A; supplemental Table 2). The lactate increase was relative to AML (supplemental Figure 1B), demonstrating that this model is well suited.

LAMs are alternatively activated to a suppressive phenotype

We hypothesized that lactate is immunosuppressive in AML by inducing macrophage polarization. We profiled the activation phenotype of BM macrophages (Ly-6C⁻, Ly-6G⁻, CD45⁺, and F4/80⁺) from leukemic mice compared to nonleukemic (NL) controls (Figure 2A). Leukemic cells were distinguished from NL myeloid cells by coexpression of GFP in bcCML cells. Well-described macrophage activation markers were assayed: CD38 and major histocompatibility complex class II (MHCII) as classic/proinflammatory markers; or early growth response protein 2 and macrophage mannose receptor (CD206) as alternative/suppressive markers.47, 48, 49 The SPICE 6.1 program (NIH) was used to quantify the frequency of phenotype subsets. This unbiased approach identified a CD206⁺ subset of LAMs enriched in disease (Figure 2B), a marker of suppressive TAMs associated with poor prognosis.50, 51, 52

Figure 2.

LAMs display an alternatively activated, suppressive phenotype. (A) Gating scheme for flow cytometric analysis of polarization markers on murine macrophages. (B) SPICE analysis showing the frequency of macrophage subpopulations from healthy controls or late-stage bcCML (n = 5); the arrows indicate population enriched in disease. (C-D) NL control macrophages, LAMs, or leukemia-derived (GFP⁺) macrophages (n = 7) from bcCML BM; (C) frequency of CD206^hi and (D) expression level by MFI of CD206 on CD206⁺. (E-F) Bulk RNA sequencing of LAMs vs macrophages from NL controls; (E) heat map of differentially expressed genes (F) and PCA of the top 500 variable genes. (G) Macrophages were sorted from NL or leukemic mice and cultured for 4 days (n = 2), and then media were profiled for cytokines. In representative example, arrows indicate a qualitative difference. (H) Venn diagrams displaying the number of gene ontology pathways significantly upregulated or downregulated by murine cancer-associated macrophages compared to each study’s own healthy controls: LAMs (n = 6) or TAMs from CM (n = 5) or BC (n = 3). (I) Hallmark gene sets enriched in LAMs compared directly to TAMs, determined by gene set enrichment analysis (GSEA): Venn diagram displaying numbers of significantly enriched gene sets and representative enrichment plots of top significant sets. Significance levels were determined by 1-way analysis of variance (ANOVA) for panels C-D and are indicated as follows: ∗∗∗P < .001; ∗∗∗∗P < .0001. Error bar indicates mean ± SD. Significance for panel I was determined by GSEA as a false discovery rate q value of <0.25. CM, colorectal liver metastasis; Ctrl, control; DAPI, 4′,6-diamidino-2-phenylindole; EGR2, early growth response protein 2; FL, female left ear punch; FR, female right ear punch; FN, female no ear punch; FSC, forward scatter; MFI, mean fluorescence intensity; ML, male left ear punch; MN, male no ear punch; MR, male right ear punch; PCA, principal component analysis; SPICE, simplified presentation of incredibly complex evaluations; SSC, side scatter.

By the time of late-stage disease (≥50% leukemic cells in BM), there was an increase in the frequency of CD206^hi LAMs (Figure 2C) and expression level of CD206 (Figure 2D). This was not observed on leukemic cells (GFP⁺). Furthermore, the frequency of MHCII⁺ macrophages was lower in bcCML, whereas CD38 and early growth response protein 2 were unchanged (supplemental Figure 2A-C). High CD206 and low MHCII expression on LAMs indicate a shift to a suppressive phenotype. Increased CD206 was detectable by early-stage disease (7%-10% leukemic cells in the BM; supplemental Figure 2D-E).

The transcriptome of LAMs was distinct from NL macrophages (Figure 2E-F), and top upregulated gene ontology pathways indicate altered interactions with immune and hematopoietic cells (supplemental Figure 2F). Top downregulated pathways were related to neutrophils and cell cycle regulation (supplemental Figure 2G). In vitro, several cytokines were exclusively expressed by LAMs (Figure 2G; Table 2). These are implicated in the immunosuppressive solid tumor microenvironment (CCL12, CCL6, and proprotein convertase subtilisin/kexin type 9 [PCSK9]),^53,54 cancer cell proliferation and invasion (CXCL10),⁵⁵ immune cell chemotaxis (CCL12/monocyte chemotactic protein 5 [MCP-5] and CXCL10),^53,55 and tumor growth and metastasis (CXCL5/LIX, MMP3, and PCSK9).56, 57, 58

Table 2.

Cytokines upregulated in LAM cocultures and reported functions in cancer

Cytokine	Reported functions
CXCL10/IP-10	Immune cell chemotaxis, tumor growth, and invasion
CCL12 (MCP-5)	Immune cell chemotaxis
CCL6/C10 (MRP-1)	Immunosuppressive TME
PCSK9	Immunosuppressive TME, tumor growth, and metastasis
LIX (CXCL5)	Tumor growth and metastasis
MMP3	Tumor growth and metastasis

IP, gamma-induced protein 10; LIX, lipopolysaccharide-induced CXC chemokine; MCP-5, monocyte chemotactic protein 5; MRP-1, myeloid-related protein 1.

To our knowledge, this is the first published transcriptomic data set of murine leukemic BM macrophages. As such, we investigated the similarity of LAMs to TAMs from other cancers. LAMs were compared to murine F4/80⁺ TAMs from solid tumor models of colorectal metastasis⁵⁹ and breast cancer (BC⁶⁰; Gene Expression Omnibus accession numbers GSE206211 and GSE126268). Pathway analysis identified gene ontology elements upregulated/downregulated in LAMs or TAMs compared to their internal controls (Figure 2H). Common upregulated elements are listed in Table 3, with downregulated elements in supplemental Table 3. Two elements were common to LAMs and both types of TAMs. One was C-type lectin receptor signaling, which functions in immunomodulation by recognizing polysaccharides from pathogens⁶¹; a notable C-type lectin receptor is CD206. The other was nuclear casein kinase and cyclin-dependent kinase substrate 1 (NUCKS1), a ubiquitously expressed transcription factor that functions in the cell cycle and DNA damage response and regulates inflammation through NF-κβ–mediated cytokine expression.⁶² Although overexpression in various cancers has been reported,^63,64 NUCKS1 in macrophages has not yet been studied. Additionally, LAMs and colorectal metastasis TAMs upregulated tumor necrosis factor and transforming growth factor β signaling, 2 key pathways related to T-cell immunomodulation and the permissive cancer microenvironment,65, 66, 67 as well as interferon regulatory factor 8, which is involved in chronic inflammation, myeloid differentiation, and the activation of macrophages.⁶⁸ There were more common downregulated elements, largely involved in the cell cycle; for example, p53 signaling, which is important for the resolution of an alternative phenotype.⁶⁹

Table 3.

Common upregulated elements in LAMs and TAMs

LAMs and both types of TAMs	LAMs and CM TAMs	CM and BC TAMs
CLR signaling pathway	Apoptosis	Lysosome
NUCKS1 24931609 ChIP-Seq hepatocytes mouse	Toxoplasmosis	TYROBP causal network WP3625
	TNF signaling pathway	Microglia pathogen phagocytosis pathway WP3626
	IRF8 27001747 ChIP-Seq BMDM mouse
	TGF-β signaling pathway WP113
	Chagas disease (American trypanosomiasis)

ChIP-Seq, chromatin immunoprecipitation sequencing; CLR, C-type lectin receptor; TGF-β, tumor growth factor β; TNF, tumor necrosis factor.

Then, we directly compared LAMs to TAMs by performing gene set enrichment analyses of “Hallmark” gene sets from the Human Molecular Signatures Database, which represent pathways with homology in humans. LAMs were enriched for metabolic, cell cycle, and stress-related pathways compared to both TAM types (Figure 2I; listed in supplemental Table 4). Although pathway analysis showed that overall cell cycle control is downregulated in LAMs compared to healthy macrophages, gene set enrichment analysis indicates that this may be to a lesser degree than in TAMs. Instead, both TAM types were enriched for angiogenesis epithelial to mesenchymal transition pathways more relevant to solid tumors (supplemental Figure 2H), and BC TAMs were enriched for several inflammatory signaling pathways listed in supplemental Table 5. In all, our data demonstrate that LAMs display a CD206^hi immunosuppressive phenotype similar to TAMs, yet with key differences that highlight them as a unique population in need of further study.

GPR81 signaling is a mechanism of lactate-induced LAM polarization

To isolate the influence of lactate on polarization, we used BMDMs treated in vitro with pathologic levels of lactate. As positive controls, we applied LPS as proinflammatory/classic stimuli or IL-4 and IL-13 as alternative stimuli.70, 71, 72, 73 IL-4/-13 are well-described type 2 helper T cytokines that persist in the AML BMME.⁷⁴ Polarization to CD206^hi BMDMs increased synergistically when lactate was combined with IL-4/13 (Figure 3A-B).

Figure 3.

Lactate-GPR81 signaling contributes to LAM polarization. (A-B) Fold change expression level of CD206 on BMDMs in vitro treated with LPS, lactate, and/or IL-4/-13 (ILs) for (A) 12 hours (n = 6) or (B) 7 days (n = 3). (C) Depiction of the initiation of bcCML in a GPR81KO BMME. (D-G) Flow cytometric analysis of NL (GFP⁻) BM cells from bcCML in wt or GPR81KO mice and NL controls; frequency of macrophages in the BM relative to NL controls of the same genetic background (n = 4-7); (D) BM macrophages, (E) frequency of CD206⁺, (F) expression level of CD206 relative to NL controls of the same genetic background, and (G) frequency of MHCII⁺ (n = 4-11). (H) Expression of CD206 on wt or GPR81KO BMDMs treated with LPS, interleukins, and/or lactate (n = 3). (I) Frequency of CD206^hi wt or GPR81KO BMDMs with or without polarization by ILs and lactate. Significance levels were determined by 1-way ANOVA for panels A-B, D-I and are indicated as follows: not significant (ns); ∗P < .05; ∗∗P < .01; ∗∗∗P < .001; ∗∗∗∗P < .0001. Error bar indicates mean ± SD. Ctrl, control; MFI, mean fluorescence intensity.

Generally, the activation of macrophages to a classic/proinflammatory phenotype is marked by upregulated expression of inducible nitric oxide synthase (iNOS/Nos2), which mediates cytotoxic NO production to assist pathogen killing and phagocytosis. Instead, alternative/suppressive macrophages express Arginase 1 (Arg1), which is involved in depleting the substrate of iNOS, L-arginase,^75,76 to resolve an immune response. CD206^hi BMDMs expressed Arg1 but not iNOS (supplemental Figure 3A-B), supporting that CD206^hi LAMs are functionally suppressive.

Next, we aimed to determine the specific lactate signaling pathway involved in LAM polarization. Lactate is an extracellular ligand to G-protein–coupled receptor 81 (GPR81)/hydroxycarboxylic acid receptor 1, the cell surface “lactate sensor.”⁷⁷ GPR81 has been linked to the pathophysiology of cancer and the immune-suppressed TME. For example, GPR81 signaling inhibits antigen-presenting cells in lung cancer and BC, proinflammatory NF-κB signaling in macrophages, and inflammasome activity.78, 79, 80 Additionally, lactate import/export occurs through monocarboxylate transporter 1 (MCT1) and MCT4.^81,82 Intracellularly, lactate activates pathways such as hypoxia-inducible factor 1, which has been connected to TAM polarization,³⁵ can be converted to pyruvate via LDH, or consumed as metabolic fuel via the citric acid cycle.⁸³ Although inhibiting MCT1/4 or LDH has been studied as a therapeutic approach for targeting lactate signaling in AML cells,^32,84 the hematopoietic system also relies on these key proteins for homeostasis leading to off-target effects. However, GPR81 has not yet been studied in AML for macrophage polarization or AML cell signaling.

To investigate lactate-GPR81 signaling as a mechanism of polarization, we used mice genetically lacking GPR81 (GPR81KO).⁷⁸ wt bcCML was initiated in GPR81KO mice (Figure 3C). The increased frequency of BM macrophages during bcCML was partially reversed in GPR81KO mice (Figure 3D). Fewer LAMs in the GPR81KO BMME expressed CD206 and at lower amounts (Figure 3E-F). This indicates that GPR81-lactate signaling contributes to LAM polarization. Furthermore, the frequency of MHCII⁺ macrophages increased (Figure 3G). By late-stage disease, GPR81KO mice displayed reduced leukemic burden in the peripheral blood and spleen (supplemental Figure 3C-E). This indicates less disease progression and/or peripheralization of leukemia cells when GPR81 signaling is absent from the BMME. Next, we repeated BMDM polarization in vitro; GPR81KO BMDMs were less polarized by stimuli than wt (Figure 3H-I), and Arg1 expression was decreased relative to CD206 expression (supplemental Figure 3F). Therefore, GPR81 signaling contributes to suppressive LAMs.

Elevated lactate is harmful to the hematopoietic BMME

Unfortunately, most current treatment options for AML exacerbate alterations to the BMME and the subsequent loss of normal hematopoiesis, leading to fatal complications of the disease. Therefore, novel therapeutic targets for AML must also consider effects on the hematopoietic BMME. We considered that excess lactate in the BM may also be harmful to normal hematopoiesis. As we have previously reported, bcCML presents with an increased percentage of lineage-/sca1+/c-kit+ (LSK) HSPCs, as well as multipotent progenitors (MPPs): megakaryocyte-biased MPP2, myeloid-biased MPP3, lymphoid-primed MPP4, as well as MSCs, a key stromal cell type for the maintenance of HSPCs (supplemental Figure 4A-G; see supplemental Methods Table 2 for markers).^23,85 Still, mature blood cell populations are lost as leukemia progresses, indicating dysfunction in the progenitors. Therefore, we investigated whether high BM lactate or LAMs affect HSPC maintenance.

To determine whether lactate reduces the hematopoietic potential of HSPCs, colony-forming (CFU-C) assays were performed.^86,87 HSPCs lost colony-forming potential when treated with increasing lactate (Figure 4A). We then asked whether coculture with HSC-niche–supportive stromal cells could increase the maintenance of HSPCs in increased lactate. However, CFU-Cs were still reduced (supplemental Figure 4H-I). Because we have previously reported that aged macrophages can also alter the colony-forming potential of HSPCs,⁸⁸ we tested the addition of LAMs to HSPC cocultures rather than lactate (supplemental Figure 4J). HSPCs showed reduced CFU-Cs when cocultured with LAMs compared to healthy macrophages (Figure 4B), suggesting that LAMs provide altered hematopoietic maintenance signals.

Figure 4.

Excess lactate negatively affects the function of HSPC and support, partially regulated by GPR81. (A) HSPCs treated with lactate for 72 hours, fold change CFUs (CFU-C) relative to the 0 mmol/L lactate control group (n = 7-10). (B) HSPCs cocultured with LAMs or healthy control macrophages for 4 days: fold change CFU-Cs relative to the control group (n = 4). (C) Cell counts per cubic microliter of blood from NL wt or GPR81 adult mice (aged 8-12 weeks): WBCs, lymphocytes, monocytes, granulocytes, RBC, or platelets (n = 7-17). (D-I) Hematopoietic progenitors’ frequency in the BM, either from bcCML in wt or GPR81KO mice, relative to the NL control of the same genetic background: (D) HSPCs (HSPCs/LSK), (E) LT-HSC, (F) ST-HSC, and (G) the MPP subsets MPP2, (H) MPP3, and (I) MPP4 (n = 4). Significance levels were determined by 1-way ANOVA for panels A,D or by unpaired t tests for panels B-C and are indicated as follows: ns; ∗P < .05; ∗∗P < .01; ∗∗∗P < .001; ∗∗∗∗P < .0001. Error bar indicates mean ± SD. Ctrl, control; GRA, granulocytes; LT-HSC, long-term HSC; LYM, lymphocytes; MON, monocytes; PLT, platelets; RBC, red blood cells; ST-HSC, short-term HSC; WBC, white blood cells.

We posited that GPR81 may be largely dispensable to normal hematopoiesis because healthy adult GPR81KO mice have appropriate ratios of mature blood cells (Figure 4C). Next, we investigated whether hematopoietic progenitors are affected by GPR81 signaling during pathologic conditions. Expansion of LSKs and short term HSCs (ST-HSCs) was abrogated when bcCML was initiated in GPR81KO mice, and other progenitor populations were unchanged (Figure 4D-I). However, GPR81KO HSPCs still showed a moderate reduction of CFU-C loss when treated with lactate compared to wt (supplemental Figure 4K-L).

We also asked whether lactate alters key HSC niche stromal cells: MSCs and OBs. Lactate treatment reduced differentiation and self-renewal of stromal cultures from healthy mice (supplemental Figure 4M-O). This is consistent with a loss of functional OBs and bone volume that we have previously reported in AML.²³ These results highlight multiple damaging effects of elevated lactate on critical components of hematopoiesis in the BM. Altogether, these data show that reducing GPR81 signaling positively affects hematopoietic progenitors, without harmful effects on normal hematopoiesis or stromal support.

Lactate-GPR81 signaling drives leukemia cell growth and self-repopulation

Finally, we investigated whether GPR81 signaling affects myeloid leukemia cells, because it is known to be crucial for the survival of other cancer cell types,⁸⁹ enhancing growth/survival pathways, immune evasion by upregulating programmed death-ligand 1 (PD-L1), and chemoresistance via compound export (ATP binding cassette subfamily B member 1 [ABCB1] transporter).89, 90, 91, 92 GPR81KO bcCML cells were generated and then used to initiate disease in GPR81KO mice (double knockout; supplemental Figure 5A). Leukemic burden was largely reduced in the BM, peripheral blood, and spleen by the time point of late-stage disease compared to wt controls (Figure 5A-D). There was a delayed early expansion of engrafted cells (supplemental Figure 5B-D), and the time to progression to >50% leukemic cells in the BM was significantly increased in GPR81 double knockout mice (Figure 5E).

Figure 5.

GPR81 signaling drives leukemic cell expansion rate and LSC self-renewal. (A-C) Leukemic burden in the (A) BM, (B) peripheral blood, and (C) spleen, of wt or GPR81KO bcCML, at the time point of late-stage disease in wt (n = 5). (D) Leukemic burden in the BM over time (n = 5-7). (E) Time to progression to late-stage disease in wt or GPR81 DKO bcCML (n = 5). (F) LSC repopulation, number of colonies at each passage, and (G) survival curve (n = 3; in duplicates). Significance levels were determined by unpaired t tests for panels A-D and log-rank (Mantel-Cox) test for panels E,G and are indicated as follows: ∗P < .05; ∗∗P < .01; ∗∗∗P < .001. Error bar indicates mean ± SD. DKO, double knockout.

Because residual LSCs after chemotherapy are the primary reason for relapse in patients,^5,25 we assessed knockout of GPR81 signaling on LSC function. When serial passaging in methylcellulose-containing media, GPR81KO bcCML cells produced fewer colonies on average at passage (P) 0 and P1 and lost repopulating capacity by P2 to P3, whereas wt bcCML cells continued to repopulate colonies to at least P7 (Figure 5F-G). These data display the importance of GPR81 to the rapid growth and self-repopulation of leukemia cells.

Discussion

Current AML chemotherapies commonly lead to relapse and damage the BM. Effective immunotherapies for myeloid leukemias are still under development and pose the challenge of shared antigens with immune cells. We report that the lactate receptor GPR81 is a potential therapeutic target for both LAMs in the suppressive BMME and leukemia cells, which spares normal hematopoietic function.

We observed elevated levels of metabolites in the BMME during AML. Although future studies will focus on additional metabolites that were elevated in human AML BM, herein, we have focused on lactate. These studies focused on defining elevated lactate as a critical driver of AML-induced macrophage polarization and leukemia progression, as visually demonstrated in Figure 6. These novel and exciting findings include both murine and human data on BM lactate levels, highlighting their relevance to human disease. Due to the Warburg effect, it is likely that accumulated leukemic cells are the source of the excess lactate. Lactate has been reported to have the strongest prognostic risk value among metabolites detected in the serum from cytogenetically normal patients with AML.⁹³

Figure 6.

Effects of elevated lactate and signaling through the lactate receptor GPR81 in the AML BMME. Lactate is elevated in the AML BMME. This polarizes LAMs to an immune suppressive phenotype, characterized by increased CD206 expression, secretion of cancer-supportive cytokines, expression of Arg1, and altered metabolism. Increased BM lactate also dysregulates normal hematopoiesis both directly to HSPCs, through harm to the HSC niche such as MSCs/OBs, and altered support by LAMs. Autocrine lactate signaling supports the growth and repopulation of leukemia cells. The lactate receptor GPR81 is a mechanism of LAM polarization and is implicated in these pathologic drivers of AML progression.

Our results highlight GPR81 as a mechanism of LAM polarization. LAMs were CD206^hi, displaying transcriptome and cytokine profiles associated with immunosuppression and cancer support. Furthermore, lactate-polarized BMDMs were CD206^hi and increased Arg1 expression, a functional marker of suppressive macrophages. CD206 has been recently suggested as another prognostic factor for AML,⁹⁴ and it is induced on monocytes cocultured with AML blasts.⁷⁵ The LAM-secreted cytokines we identified herein may be further investigated for their role in AML, particularly in suppressing cytotoxic T-cell response.

A recent study found that human AML-associated macrophages (AAMs) displayed alternative polarization, decreased phagocytosis, upregulated mitochondrial function, and CD206 overexpression.³⁸ AAMs correlated with poor prognosis in a cohort of patients with myelodysplastic syndrome and influenced engraftment of patient-derived xenografts, showing the clinical importance of suppressive macrophages. The similarity of murine LAMs in our study to these AAMs highlights the translatability of our results to human AML. Other current research demonstrates that macrophage polarization can be altered through therapeutic means in AML.^95,96 Increasing the proinflammatory capabilities of macrophages by GPR81 inhibition as an immunotherapy in combination with chemotherapies or other immunotherapies that target AML blasts, such as chimeric antigen receptor (CAR) T-cell therapy, may increase efficacy. Additionally, our findings support that targeting GPR81 has therapeutic potential to simultaneously limit AML cell growth and LSC repopulation.

It is known that high lactate can contribute to the expression of programmed death-ligand 1 on tumor cells, CD8⁺ T-cell exhaustion, and suppression of MHCII^hi immune cells.^78,97 It is likely that other BM niche or immune cells may also be afflicted by a chronic increase in BM lactate. We experimentally observed reduced MSC function, which may contribute to a decrease in healthy OBs, a symptom of AML. Interestingly, regulatory T cells (T_Regs) are more resistant to lactate-mediated inhibition than other T-cell types⁹⁸ and are increased in patients with AML.⁹⁹ Furthermore, alternative macrophages induce chemotaxis/differentiation of T_Regs to the TME,^54,100,101 and T_Regs provide an immune-suppressed niche in which LSCs may escape immune attack.¹⁰² The impact of GPR81 signaling on T-cell subsets or other BM cell types in AML is yet to be studied.

In NL mice, experimental knockout of GPR81 did not cause negative effects on hematopoiesis or stromal cell support, indicating GPR81 as a potentially safe therapeutic target for the hematopoietic system. Furthermore, leukemic mice without GPR81 exhibited a reduced expansion of HSPC populations that may lead to their exhaustion. Additional research will help elucidate the role of GPR81 in a human environment and solidify the potential of GPR81 as a therapeutic target. Additionally, because lactate production is a hallmark of cancer, findings on the mechanisms of lactate signaling within the BMME are potentially applicable to multiple malignancies with BM involvement, including additional types of leukemia and bone metastases of solid tumors.

Our findings demonstrate the lactate receptor GPR81 as a mechanism for the polarization of suppressive LAMs and as a driver of AML cell growth. This research suggests that targeting GPR81 signaling in the BMME during AML may be a selective and well-tolerated therapeutic option to prevent LSC repopulation and rescue microenvironmental dysfunction.

Conflict-of-interest disclosure: The authors declare no competing financial interests.