Subversion of Serotonin Receptor Signaling in Osteoblasts by Kynurenine Drives Acute Myeloid Leukemia

Abstract

Remodeling of the microenvironment by tumor cells can activate pathways that favor cancer growth. Molecular delineation and targeting of such malignant-cell nonautonomous pathways may help overcome resistance to targeted therapies. Herein we leverage genetic mouse models, patient-derived xenografts, and patient samples to show that acute myeloid leukemia (AML) exploits peripheral serotonin signaling to remodel the endosteal niche to its advantage. AML progression requires the presence of serotonin receptor 1B (HTR1B) in osteoblasts and is driven by AML-secreted kynurenine, which acts as an oncometabolite and HTR1B ligand. AML cells utilize kynurenine to induce a proinflammatory state in osteoblasts that, through the acute-phase protein serum amyloid A (SAA), acts in a positive feedback loop on leukemia cells by increasing expression of IDO1-the rate-limiting enzyme for kynurenine synthesis-thereby enabling AML progression. This leukemia-osteoblast cross-talk, conferred by the kynurenine-HTR1B-SAA-IDO1 axis, could be exploited as a niche-focused therapeutic approach against AML, opening new avenues for cancer treatment.

Significance: AML remains recalcitrant to treatments due to the emergence of resistant clones. We show a leukemia-cell nonautonomous progression mechanism that involves activation of a kynurenine-HTR1B-SAA-IDO1 axis between AML cells and osteoblasts. Targeting the niche by interrupting this axis can be pharmacologically harnessed to hamper AML progression and overcome therapy resistance. This article is highlighted in the In This Issue feature, p. 873.

Figure 1.

Ablation of Htr1b in osteoblasts prevents AML progression. A, Survival curve of WT mice treated with vehicle (n = 4) or PTH (n = 7) and injected with MLL/AF9 AML cells. B–E, Survival curves of WT MLL/AF9-injected mice, their spleen weights, and representative epifluorescence images (radiance p/sec/cm²/sr) of leukemia progression 14 days after MLL/AF9 injection in Htr1b^−/− (n = 29) and Htr1b^+/+ littermates (n = 13; B); Htr1b^fl/fl; LepR-Cre: Htr1b_Lep-R^−/− (n = 8) and Htr1b_Lep-R^+/+ littermates (n = 6; C); Htr1b^fl/fl; Col1a1-Cre: Htr1b_c-osb^−/− (n = 11) and Htr1b_c-osb^+/+ littermates (n = 12; B)—the 4 Htr1b_c-osb^−/− mice that developed leukemia are represented with red stars in the histogram of spleen weight and excluded from the statistical analysis; Htr1b^fl/fl; OCN-Cre: Htr1b_d-osb^−/− (n = 5) and Htr1b_d-osb^+/+ littermates (n = 10; E). Orange arrow indicates the systematic genetic interrogation approach followed. F, Survival curve of Htr1b^fl/fl; Osx-Cre: Htr1b_Osx^−/− (DOX removed 24 hours after MLL/AF9 injection; n = 9) and Htr1b_Osx^+/+ (kept on DOX, n = 6). G, Leukemia burden quantification (total flux, photons/sec) at day 12 after MLL/AF9 injection, Htr1b_Osx^+/+ (DOX, n = 6), Htr1b_Osx^−/− (no DOX, n = 9). H, Survival curve of WT mice injected with MLL/AF9 cells and treated with either vehicle (n = 10) or the HTR1B antagonist SB224289 (SB9; n = 10). All survival curves shown are Kaplan–Meier curves with the P value of log-rank (Mantel–Cox) test between the indicated groups. All data are represented as mean ± SEM; statistical analysis done with an unpaired t test. *, P ≤ 0.05; **, P ≤ 0.01; ****, P ≤ 0.0001. See also Supplementary Fig. S1. DOX, doxycycline.

Figure 2.

Kynurenine (Kyn) is an oncometabolite increased in the BM niche of patients with MDS and AML that binds to HTR1B. A and B, Volcano plots for metabolites with CV <30% comparing OCI-AML3 cells untreated (AML) and human osteoblasts (hOsb; A) or AML cells untreated versus cocultures (24 hours). In B, arrows point to Kyn. C, Tryptophan (Trp) catabolism scheme. D, Relative abundance of Trp and its catabolic metabolites: Kyn, serotonin (5-HT), and 5-hydroxytryptophan (5-HTP) in the indicated supernatants at 24 hours (n = 6); two-way ANOVA. E, Heat map of the first 30 metabolites with CV <15% and histograms of fold induction (FI) of AML versus hOsb (gray) or AML versus coculture (blue). F, Violin plots of Kyn/Trp ratio levels in serum circulating levels of control-injected (n = 19) versus MLL/AF9-injected (n = 28) mice; unpaired t test. G, Violin plots of Kyn/Trp ratio levels in BM plasma from healthy donors (n = 27), MDS (n = 30), and AML (n = 24) patients; one-way ANOVA. H, Kyn/Trp levels in paired BM plasma samples at MDS stage and its corresponding transformed AML stage (n = 6); paired t test. I, RNA-sequencing analysis of BM-MNCs from patients with MDS (n = 30) and AML (n = 30) patients (TPM) for TPH1 and IDO1; two-way ANOVA. J,IDO1/TPH1 mRNA ratio in BM-MNCs from healthy donors (n = 32), patients with MDS (n = 10), and patients with AML (n = 20); one-way ANOVA. K, Concentration dependence of the Kyn-mediated competition of [³H]-5-HT (25 nmol/L, 41.3Ci/mmol) binding by HEK293T membranes overexpressing the mouse (n = 4 experiments) or the human receptor (n = 2 experiments), yielding an IC₅₀ of 54.1 μmol/L and 24.4 μmol/L, respectively (see Table 1 for details). L, G_i/o-mediated cAMP inhibition assays (n = 14). M, Binding of [³H]-5-HT (25 nmol/L, 41.3 Ci/mmol) or [³H]-Kyn (50 μmol/L, 0.125 Ci/mmol) was measured with Htr1b-overexpressing HEK293T membranes in the presence of increasing concentrations of SB9 (n = 4). Nonlinear regression fitting was used to fit the isotherms, and the best-fit values and statistics of the fit are shown in Table 1. All data are expressed as mean ± SEM. *, P ≤ 0.05; **, P ≤ 0.01; ***, P ≤ 0.001; ****, P ≤ 0.0001. See also Supplementary Fig. S2 and Table 1.

Figure 3.

Genetic inhibition of kynurenine (Kyn) production hinders AML progression. A, Representative epifluorescence images of leukemia progression in WT mice injected with MLL/AF9-CRISPR/Cas9-edited cells (sgRNAs: #146, #196 and #203; Ctrl: no leukemia). B, Survival curve of mice injected with the indicated sgRNAs MLL/AF9-edited or Cas9-only-MLL/AF9 control cells (n = 3 all groups). C, Representative epifluorescence images of leukemia progression in WT mice injected with MLL/AF9-CRISPR/Cas9–edited cells (sgRNAs: #610) and Ido1 mRNA levels of MLL/AF9-sgRNA#610-edited cells before injection (n = 4); unpaired t test. D, Survival curve of WT mice injected with MLL/AF9-sgRNA#610-edited cells (n = 11) or Cas9 only control (n = 9). Mice showing > 60% of unedited (WT) sequence in their BM after harvesting are depicted as sgRNA#610_{editing lost} (green; n = 5). E,IDO1 mRNA levels in OCI-AML3 cells nucleofected with Cas9 and sgRN#610 used in transplant experiment. F,IDO1 mRNA levels in OCI-AML3 cells exposed to IFNγ (overnight, 50 ng/mL, n = 3); two-way ANOVA. G, Outline of transplantation assay with OCI-AML3 CRISPR/Cas9-IDO1–targeted cells in NSG mice. H, AML burden in BM, SP, and spleen weight (mg)—referred to as total body weight (g)—of NSG mice 3 weeks after injection of OCI-AML3 cells (n = 8, Cas9; n = 10, #126+170). I, Proliferation of OCI-AML3 cells upon 72 hours of coculture with primary human osteoblasts (n = 7). Survival curves are Kaplan–Meier with P value of log-rank (Mantel–Cox) test between the indicated groups. All data are expressed as mean ± SEM. Statistical analysis done with unpaired t test unless otherwise stated. **, P ≤ 0.01; ***, P ≤ 0.001; ****, P ≤ 0.0001. See also Supplementary Fig. S3 and Supplementary Data S1.

Figure 4.

AML cells self-amplify Kyn production through HTR1B–SAA signaling in osteoblasts. A, Schematic of RNA-seq analysis strategy (left) and box plots (right) of the main secreted molecules significantly upregulated in primary human osteoblasts untreated (UT) or cocultured 24 hours with the THP-1 AML cell line (n = 2); Wald test, two-sided. B, Box plots for IDO1 and TPH1 from RNA-seq analysis of THP-1 cells exposed 24 hours to primary human osteoblasts (n = 2); Wald test, two-sided. C,IDO1 mRNA levels in OCI-AML3 cells exposed overnight to the indicated molecules (UT and SAA1, n = 15; IL1α, IL1β, IL6, CXCL1, and CXCL8, n = 6; IL33, IL34, CXCL3, CXCL5, CCL2, and CCL20, n = 3). D,Ido1 mRNA levels in WEHI-3B cells exposed overnight to recombinant mouse SAA3 or recombinant human SAA1 (n = 8). E,Saa3 mRNA relative level in primary differentiated mouse calvaria from Htr1b^−/− and Htr1b^+/+ littermates, exposed for 24 hours to 5-HT (25 nmol/L, n = 7–8), Kyn (25 nmol/L, n = 5), or the WEHI-3B cell line (n = 10–12); two-way ANOVA. F, Violin plots of SAA3 peripheral blood (PB) serum levels in control (n = 20) and MLL/AF9-injected mice (n = 20); unpaired t test. G, Violin plots of SAA1 BM plasma levels in healthy donors (n = 30), patients with MDS (n = 35), and patients with AML (n = 23). H, SAA1 BM plasma levels in paired samples from patients (MDS and corresponding AML-transformed stage; n = 6 paired samples); paired t test. I, Multiple variable data plot of BM plasma levels for SAA1 and Kyn/Trp ratio along healthy, MDS, or AML samples; Pearson correlation values are shown for Kyn/Trp ratio and SAA1 BM plasma levels. All data expressed as mean ± SEM. Statistical analysis was done with one-way ANOVA unless otherwise stated. **, P ≤ 0.01; ***, P ≤ 0.001; ****, P ≤ 0.0001. See also Supplementary Fig. S4.

Figure 5.

SAA1 selectively promotes leukemic cell proliferation by upregulating IDO1 expression through activation of the aryl hydrocarbon receptor (AHR) pathway. A, Proliferation of human THP-1 and OCI-AML3 (n = 22) and mouse WEHI-3B (n = 8) AML cell lines exposed to SAA1 or SAA3, respectively (1 μg/mL, 24–72 hours). Proliferation (B) and IDO1 mRNA levels (C) of human BM-MNCs isolated from MDS or AML (lineage-depleted) BM aspirates (n = 8) and exposed to SAA1 (5 μg/mL, 24 hours), paired t test. D, Left, schematic of PDX model used. Right, proliferation of total human BM cells isolated from the PDX mice injected with either healthy CD34⁺ (n = 3) or patient-derived AML cells (n = 8) exposed to vehicle (PBS) or SAA1 (1 μg/mL, 24 hours). E,IDO1 mRNA level from cells in D; two-way ANOVA. In vivo proliferation of leukemic blasts (hCD45⁺CD33⁺; F) and BM AML burden (G) in mice treated for 2 or 8 days with either vehicle (n = 10 and n = 7, respectively) or SAA1 (n = 14 and n = 9, respectively); two-way ANOVA. H, Proliferation of total human AML BM cells isolated from PDX mice and nucleofected with Cas9 (n = 5) or Cas9 and the combination of sgRNA#126 and sgRNA#170 (n = 8) exposed to vehicle or SAA1 (1 μg/mL, 24 hours); two-way ANOVA. I, mRNA level of CYP1A1 and CYP1A2 from cells in D; two-way ANOVA. J, Violin plots for mRNA levels of CYP1A1 and CYP1A2 in BM-MNCs from healthy donors (n = 15) and patients with AML (n = 17). K,CYP1A1 and CYP1A2 mRNA levels from cells in B. L, GSEA of AHR activation signature genes in THP-1 cells cocultured with human osteoblasts for 24 hours. All data expressed as mean ± SEM. Statistical analysis was done with an unpaired t test unless otherwise stated. *, P ≤ 0.05; **, P ≤ 0.01; ***, P ≤ 0.001; ****, P ≤ 0.0001. See also Supplementary Fig. S5.

Figure 6.

Pharmacologic targeting of the Kyn–HTR1B–SAA–IDO1 axis in PDXs. A, Survival curve comparing vehicle (n = 26) and epacadostat-treated mice (n = 18 for 0.8 g/kg and n = 13 for 1.6 g/kg). Kaplan–Meier curve with P value of log-rank (Mantel–Cox) test. SAA3 (B) and Kyn/tryptophan (Trp) ratio (C) serum levels in NSGS mice transplanted with CD34⁺ healthy cells (n = 11) or with patient-derived AML cells (n = 27). D, Schematic describing pharmacologic targeting of IDO1 (epacadostat) in patient-derived AML xenograft (PDX) in NSGS mice. E, Kyn/Trp ratio in serum of PDX mice 5 weeks after AML transplant and 2 weeks after epacadostat treatment (n = 8 vehicle; n = 10 epacadostat). F, Representative flow cytometry plots depicting the percentage of human or mouse CD45⁺ cells in the BM of PDX mice (left) and AML burden in the BM of PDX mice at harvest (right; n = 8 vehicle; n = 10 epacadostat). G, Representative flow cytometry plots (left) and cell-cycle analysis of leukemic blasts (CD45⁺ CD33⁺) of PDX mice treated with either vehicle (n = 8) or epacadostat (Epac; n = 8). H, Cell-cycle analysis of mice in G. I, Schematic diagram showing the in vivo PDX mouse model treated with the combination therapy (Ara-C 60 mg/kg 1–5 days + epacadostat 1.6 g/kg ad libitum 3 weeks). AML burden in BM (J) and SP (K) 11 weeks after transplant, three weeks after combination therapy; control chow (ctrl; n = 4), Ara-C (n = 3), epacadostat (n = 4), and combination therapy (Ara-C + epacadostat, n = 3); one-way ANOVA; unpaired t test P values are shown for BM ctrl versus Ara-C and epacadostat groups. L, Schematic model of the Kyn–HTR1B–SAA–IDO1 axis depicting the AML-mediated osteoblastic self-reinforcing niche remodeling. All data expressed as mean ± SEM. Statistical analysis done with unpaired t test unless otherwise stated. *, P ≤ 0.05; **, P ≤ 0.01; ****, P ≤ 0.0001. See also Supplementary Fig. S6. Hg, hemoglobin; PLT, platelets.