Acute myeloid leukemia-induced remodeling of the human bone marrow niche predicts clinical outcome

Abstract

Murine models of myeloid neoplasia show how leukemia infiltration alters the hematopoietic stem cell (HSC) niche to reinforce malignancy at the expense of healthy hematopoiesis. However, little is known about the bone marrow architecture in humans and its impact on clinical outcome. Here, we dissect the bone marrow niche in patients with acute myeloid leukemia (AML) at first diagnosis. We combined immunohistochemical stainings with global gene expression analyses from these AML patients and correlated them with clinical features. Mesenchymal stem and progenitor cells (MSPCs) lost quiescence and significantly expanded in the bone marrow of AML patients. Strikingly, their HSC- and niche-regulating capacities were impaired with significant inhibition of osteogenesis and bone formation in a cell contact-dependent manner through inhibition of cytoplasmic β-catenin. Assessment of bone metabolism by quantifying peripheral blood osteocalcin levels revealed 30% lower expression in AML patients at first diagnosis than in non-leukemic donors. Furthermore, patients with osteocalcin levels ≤11 ng/mL showed inferior overall survival with a 1-year survival rate of 38.7% whereas patients with higher osteocalcin levels reached a survival rate of 66.8%. These novel insights into the human AML bone marrow microenvironment help translate findings from preclinical models and detect new targets which might pave the way for niche-targeted therapies in AML patients.

Graphical abstract

Figure 1.

AML infiltration induces proliferation of bone marrow MSPCs. (A) Exemplary procedure for quantification of CD271. First, tile images were fused into larger images (1). Second, a color deconvolution was performed (ImageJ, IHC Profiler plugin) to obtain separate hematoxylin and DAB images. The hematoxylin image was thresholded to obtain a mask of the total bone region (2). The DAB image (3) was inverted and thresholded to obtain a mask of the positive DAB regions (4). Scale bars, 1 mm. (B) Representative images of bone marrow biopsy samples of non-leukemic donors (controls) and AML patients stained for CD271⁺ MSPCs (brown) and counterstained with hematoxylin (blue). Scale bars, 400 µm. (C) Quantification of CD271⁺ bone marrow MSPCs in controls (n = 58) and AML patients (n = 36); CD271⁺ area is calculated by CD271⁺ stained area divided by total tissue surface. (D) Left: representative image from bone marrow after silver staining representing reticular fibers. Right: image converted by trainable Weka segmentation showing 3 different classes: red, reticular fiber; green, tissue; purple, background. Scale bars, 50 µm. (E) Representative images of bone marrow biopsy samples from controls and AML patients stained for reticular fibers with silver staining (black fiber) and counterstained with the fast-red nuclear solution. Scale bars, 100 µm. (F) Quantification of reticular fibers in control (n = 19) and AML (n = 37) bone marrow by calculating positive silver stained area divided by total tissue surface (n = 56). Data are shown as mean ± standard error of the mean (SEM). *P < .05; **P < .01 (determined by Mann-Whitney U test).

Figure 2.

Bone marrow MSPCs from AML patients harbor a distinct gene expression profile. (A) Principal component (PC) analysis of MSPCs isolated from AML patients and non-leukemic donors. (B) Hierarchical cluster analysis of genes with significantly different expression between MSPCs from AML patients (n = 8) and controls (n = 3) by the unweighted pair-group method with arithmetic averages. Gene signal intensities were normalized to the mean signal of all samples with log₁₀ transformation. Horizontal rows represent individual samples; vertical columns represent individual genes. Black indicates average signal intensity, brightest red represents twofold or higher upregulated, and brightest blue represents twofold or lower downregulated gene expression relative to the mean. (C) Twenty top altered canonical pathways (P < .05) in bone marrow MSPCs from AML patients compared with controls were generated by Ingenuity Pathway Analysis (IPA) software (QIAGEN) and subcategorized into 4 functional groups. (D) Overlapping canonical pathway analysis generated by IPA. Black line indicates 4 genes and red line indicates more than 4 genes were shared between connected pathways. (E) Gene set enrichment analysis (GSEA) of proliferation-related gene sets with normalized enrichment score (NES) >1.5 and P < .05. (F) GSEA of osteogenesis-related gene sets with NES >1.5 and P < .05. (G) Left: heatmap of HSC-regulating genes in MSPCs from AML patients and controls. Gene signal intensities were central to the mean signal of individual gene symbols. Horizontal rows represent individual samples; vertical columns represent individual genes. Black indicates average signal intensity, brightest red represents 1.5-fold or higher upregulated, and brightest blue represents twofold or higher downregulated gene expression relative to the mean. Right: complete blood count of peripheral blood from corresponding patients at diagnosis including hemoglobin (Hb), absolute neutrophil count (ANC), and platelet (Plt). *P < .05 (determined by unpaired Student t test).

Figure 3.

AML infiltration inhibits osteogenesis. (A) Alizarin red staining to detect calcium deposition after 14 days of induction of mineralization in HuMSPCs in the presence of PBMNCs from healthy donors (control) or different AML cell lines. Left: low-magnification image of a 48-well plate; right: images (4× objective) showing calcium deposition. (B) Colorimetric detection of alizarin red staining using absorbance at 450 nm to quantify calcium deposition in HuMSPCs after 14 days of induction of mineralization in the presence of healthy PBMNCs as controls or different AML cell lines (representative of 9 independent experiments). (C) Colorimetric detection of alizarin red staining using absorbance at 450 nm to quantify calcium deposition in SaOS2 cells after 7 days of induction of mineralization in the presence of controls or different AML cell lines (representative of 6 independent experiments). (D) Images (4× objective) of 96-well plate and colorimetric detection of alizarin red staining using absorbance at 450 nm to quantify calcium deposition in HuMSPCs after 14 days of induction of mineralization in the presence of controls or primary AML cells (3 replicates per sample). (E) Image (4× objective) and colorimetric detection of alizarin red staining using absorbance at 450 nm to quantify calcium deposition in HuMSPCs after 21 days of induction of mineralization in the presence of different AML cell lines or corresponding conditioned medium (representative of 3 independent experiments). Data are shown as mean ± SEM. *P < .05; **P < .01; ***P < .001; ****P < .0001 (determined by unpaired Student t test).

Figure 4.

AML inhibits osteogenesis through interference in WNT/β-catenin pathway. (A) GSEA enrichment plot of GO_WNT_ACTIVATED_RECEPTOR_ACTIVITY of bone marrow MSPCs from AML patients and controls (NES, −1.55; P = .021; false discovery rate q = 0.114). (B) Representative overlay histogram of cytoplasmic β-catenin fluorescence intensity in HuMSPCs after mono- or coculture with healthy PBMNCs (controls) or indicated AML cell lines. (C) Mean fluorescence intensity (MFI) of cytoplasmic β-catenin levels in HuMSPCs after mono- or coculture with the indicated cells, normalized to HuMSPCs (representative of 5 independent experiments). (D) Colorimetric detection of alizarin red staining using absorbance at 450 nm to quantify calcium deposition in SaOS2 cells transduced with either empty vector or different CTNNB1 short hairpin RNAs after 7 days of induction of mineralization (representative of 4 independent experiments). (E) Colorimetric detection of alizarin red staining using absorbance at 450 nm to quantify calcium deposition in HuMSPCs after 14 days of induction of mineralization in the presence of healthy PBMNCs or different AML cell lines with either 10 µM SKL2001 or dimethyl sulfoxide (DMSO) (representative of 8 independent experiments; result is normalized to HuMSPCs with DMSO). (F) MFI of cytoplasmic β-catenin levels in HuMSPCs after mono- or coculture in the presence of healthy PBMNCs or different AML cell lines with either 10 µM SKL2001 or DMSO (representative of 5 independent experiments; result is normalized to HuMSPCs with DMSO). (G) Colorimetric detection of alizarin red staining using absorbance at 450 nm to quantify calcium deposition in SaOS2 cells transduced with either empty vector or CTNNB1 overexpressing plasmids after 7 days of induction of mineralization in the presence of different AML cell lines (representative of 3 independent experiments). Data are shown as mean ± SEM. *P < .05; **P < .01; ***P < .001 (determined by unpaired Student t test).

Figure 5.

Deficient bone metabolism predicts clinical outcome. (A) Osteocalcin (OCN) levels measured in the peripheral blood of controls (n = 31) and AML patients (n = 58). Data are shown as mean ± SEM. **P < .01 (determined by Mann-Whitney U test). (B) Bone marrow blast infiltration in patients with osteocalcin level ≤11 ng/mL (OCN-Low, n = 16) and osteocalcin levels >11 ng/mL (OCN-High, n = 26). Data are shown as mean ± SEM. *P < .05 (determined by Mann-Whitney U test). (C) Peripheral blood blast percentage in patients with OCN-Low (n = 24) and OCN-High (n = 32). Data are shown as mean ± SEM. ****P < .0001 (determined by Mann-Whitney U test). (D) Overall survival of OCN-Low (n = 24) and OCN-High (n = 34). *P = .0339 (determined by log-rank test). (E) Relapse-free survival of AML patients receiving curative-intent treatment with OCN-Low (n = 14) and OCN-High (n = 20). P = .1105 (determined by log-rank test).