Bone marrow stroma cells promote induction of a chemoresistant and prognostic unfavorable S100A8/A9high AML cell subset

Abstract

The bone marrow (BM) stroma represents a protective niche for acute myeloid leukemia (AML) cells. However, the complex underlying mechanisms remain to be fully elucidated. We found 2 small, intracellular, calcium-sensing molecules, S100A8 and S100A9, among the top genes being upregulated in primary AML blasts upon stromal contact. As members of the S100 protein family, they can modulate such cellular processes as proliferation, migration, and differentiation. Dysregulation of S100 proteins is described as a predictor of poor survival in different human cancers, including increased S100A8 expression in de novo AML. Thus, we wanted to decipher the underlying pathways of stroma-mediated S100A8/A9 induction, as well as its functional consequences. Upregulation of S100A8/A9 after stromal cross talk was validated in AML cell lines, was contact independent and reversible and resulted in accumulation of S100A8/A9high cells. Accordingly, frequency of S100A8/A9high AML blasts was higher in the patients' BM than in peripheral blood. The S100A8/A9high AML cell population displayed enhanced utilization of free fatty acids, features of a more mature myeloid phenotype, and increased resilience toward chemotherapeutics and BCL2 inhibition. We identified stromal cell-derived interleukin-6 (IL-6) as the trigger for a Jak/STAT3 signaling-mediated S100A8/A9 induction. Interfering with fatty acid uptake and the IL-6-Jak/STAT3 pathway antagonized formation of S100A8/A9high cells and therapeutic resistance, which could have therapeutic implications as a strategy to interfere with the AML-niche dynamics.

Graphical abstract

Figure 1.

BM stromal cells induce S100A8 and S100A9 in AML cells. (A) FACS-sorted primary CD33⁺CD34⁺ AML blasts (n = 6) were cultured for 24 hours in the absence (w/o) or presence (w) of a confluent layer of the human BM-derived stroma cell line HS-5 and subsequently analyzed by microarray analysis (left). The heat map (right) shows the top 30 highest differentially expressed genes in the group of AML blasts cultured in the presence of HS-5 cells. (B) Publicly available data from TCGA (LAML data set) was divided into S100A8^hi/S100A9^hi and S100A8^lo/S100A9^lo groups based on the mean expression, and depicted as survival curves using the Mantel-Cox test for calculation of significance. (C) AML cell lines OCI-AML3 and MOLM-13 were cultured for 48 hours in the presence or absence of HS-5 cells (contact; OCI-AML, n = 11; MOLM-13, n = 10) or HS-5 CM (CM; OCI-AML, n = 5; MOLM-13, n = 5). Gene expression of S100A8 and S100A9 was analyzed by qPCR and is depicted as the fold change of treated/untreated cells (untreated set as 1). (D) Frequency of S100A8/S100A9^hi cells as shown in the representative pseudocolored flow cytometry plot was determined in AML cell lines (OCI-AML, n = 27; MOLM-13, n = 27) cultured for 48 hours in the absence or presence of HS-5 CM. (E) AML cell lines (OCI-AML, n = 13; MOLM-13, n = 11) were cultured for 48 hours in the absence or presence of CM from MSCs of 7 patients with AML (AML-MSC CM, patient ID 11-17; supplemental Table 1) and analyzed for the frequency of the S100A8/A9^hi population by flow cytometry. (F) Log₂-transformed normalized counts of the S100A8 and S100A9 gene expression (from the TCGA LAML data set) were correlated by the Spearman test. (G) The S100A8/A9^hi population among matched-pair PB- and BM-derived AML blasts (n = 10; patient ID 1-10; supplemental Table 1) was analyzed by flow cytometry. Data are expressed as the standard error of mean. ∗P < .05; ∗∗P < .01; ∗∗∗P < .001. FACS, fluorescence-activated cell sorting; qPCR, quantitative real-time polymerase chain reaction.

Figure 2.

S100A8 and S100A9 is induced via IL-6/Jak/STAT3 signaling. (A) CM from HS-5 cells (n = 8) was analyzed with a bead-based multiplex assay for 25 human cytokines (IL-9, IL-10, IL-17A, and interferon α [IFN-α] not detected). The pie chart displays the individual proportion among the total analytes. (B) Induction of S100A8 and S100A9 on the mRNA (left; OCI, n = 4; MOLM-13, n = 5) and induction of an S100A8/A9^hi population (right; OCI-AML, n = 4; MOLM-13, n = 4) were analyzed after treatment with 10 ng/mL of rhIL-6 for 48 hours by qPCR and flow cytometry, respectively. mRNA levels are shown as the fold change with or without rhIL-6 (untreated set as 1). (C) Frequency of S100A8/A9^hi AML cells was determined by flow cytometry after culturing HS-5 CM-treated MOLM-13 cells (n = 4) in the absence (isotype) or presence of 10 μg/mL anti-IL-6 or anti-IL-6R blocking antibodies. (D) CM from HS-5 cells was prepared after inflammatory priming of the HS-5 cells (with 15 ng/mL tumor necrosis factor-α and 10 ng/mL IFN-γ). Gene expression of IL-6 in HS-5 cells was determined by qPCR and expressed as fold change between primed and nonprimed samples (nonprimed samples set as 1; left, n = 4). The frequency of S100A8/A9^hi AML cells (right; OCI-AML, n = 2; MOLM-13, n = 4) cultured with nonprimed and primed HS-5 CM for 48 hours was determined by intracellular flow cytometry. (E) Gene expression of IL-6 was determined in healthy donor– and AML patient–derived BM-MSCs (both n = 5). (F) Gene expression of PIM1 and SOCS3 (2 STAT3 target genes) was determined in AML cell lines (OCI-AML n = 3, MOLM-13 n = 3) cultured in the absence or presence of HS-5 CM for 48 hours by qPCR and is shown as the fold change (untreated set as 1). (G) Phosphorylation of the STAT proteins (p-STAT) 1, 3, 5, and 6 was determined by PhosFlow in AML cell lines (OCI-AML, n = 3; MOLM-13, n = 3) cultured in the absence or presence of HS-5 CM for 20 minutes and is depicted as a heat map showing the fold change with or without HS-5 CM (left; untreated set as 1). Most prominently, p-STAT3 was induced by HS-5 CM treatment as representatively shown in the histogram (right; scale indicates median fluorescence intensity). (H-J) STAT3 phosphorylation was analyzed by PhosFlow in AML cell lines cultured in the presence of HS-5 CM for 20 minutes after pretreatment for 2 hours with anti-IL-6–blocking antibodies (H; OCI-AML, n = 3; MOLM-13, n = 3), ruxolitinib (I; OCI-AML, n = 2; MOLM-13, n = 2), and Jak1 or Jak2 inhibitor (J; OCI-AML, n = 2; MOLM-13, n = 2). (K-L) Induction of S100A8 and S100A9 gene expression after 48 hours of culture with HS-5 CM was analyzed in the absence (0 μM) or presence (1 μM) of the pan-Jak inhibitor ruxolitinib (K; OCI-AML, n = 3; MOLM-13, n = 3) or the p-STAT3 inhibitor C188-9 (L; OCI-AML, n = 5; MOLM-13, n = 5). Values are depicted as fold change with or without HS-5 CM (untreated set as 1). Data are expressed as the standard error of mean. ∗P < .05; ∗∗P < .01; ∗∗∗P < .001; ns, not significant. qPCR, quantitative real-time polymerase chain reaction.

Figure 3.

S100A8/A9^hiAML cells display enhanced fatty acid metabolism. (A) Gene expression of the fatty acid translocase CD36 was determined in FACS-sorted S100A8/A9^hi and S100A8/A9^lo OCI-AML (n = 3) and MOLM-13 (n = 3) cells cultured for 48 hours in the presence of HS-5 CM by qPCR and is shown as the fold change (S100A8/A9^lo cells were set as 1). (B) Surface levels of CD36 were determined by flow cytometry on S100A8/A9^hi and S100A8/A9^lo cells among the AML cell lines cultured for 48 hours in the presence of CM from either HS-5 cells (left; OCI-AML, n = 4; MOLM-13, n = 4) or AML-MSCs isolated from 6 patients (right; OCI-AML, n = 12; MOLM-13, n = 12, patient ID 11-16; supplemental Table 1). (C) Surface levels of CD36 were analyzed by flow cytometry in matched-pair (n = 10; patient ID 1-10; supplemental Table 1) PB- and BM-derived S100A8/A9^hi and S100A8/A9^lo AML blasts, as representatively shown in histograms (left) and summarized in a bar graph (right). (D) Localization of S100A8 and S100A9 was visualized by fluorescence microscopy (left, scale bars = 20 μm). MOLM-13 cells (n = 5) were cultured in the absence (top) or presence (bottom) of HS-5 CM and stained for the cell membrane (WGA, green), S100A8 (left, red) or S100A9 (right, red), and the nucleus (blue, DAPI). Co-localization of S100A8/S100A9 and WGA were calculated with ZEN software (bar graphs; Zeiss). (E) A proximity ligation was performed with a Duolink Flow kit (Sigma-Aldrich) with antibodies against S100A8, S100A9, and CD36. The representative images show the bright-field (left, BF) and the fluorescence signal (right) in MOLM-13 cells. Bars represent 20 μm for 20× magnification. (F) Long-chain fatty acid uptake by AML cell lines (OCI-AML, n = 5; MOLM-13, n = 5) cultured for 48 hours in absence or presence of HS-5 CM was measured by flow cytometry using the fluorescent probe Bodipy FL_C16 based on the median fluorescence intensity (MdFI). (G) Long-chain fatty acid uptake was analyzed in AML cell lines cultured for 48 hours in the absence or presence of HS-5 CM, with or without ruxolitinib (OCI-AML, n = 2; MOLM-13, n = 3). (H) Oxygen consumption rate (OCR; as a surrogate for oxidative phosphorylation) was determined for MOLM-13 cells (n = 3) cultured for 48 hours in the absence (absence) or presence of HS-5 CM (CM) at baseline and after consecutive injection of SSO and etomoxir, which enabled calculation of the dependence on exogenous (exo) and endogenous (endo) fatty acids for fueling mitochondrial respiration. (I) S100A8/A9^hi frequency was analyzed in AML cell lines (OCI-AML, n = 5; MOLM-13, n = 4) cultured for 48 hours in the presence of HS-5 CM, with or without the CD36 inhibitor SSO. Data are expressed as the standard error of the mean. ∗P < .05; ∗∗P < .01; ∗∗∗P < .001; ns, not significant. DAPI, 4′,6-diamidino-2-phenylindole; FACS, fluorescence-activated cell sorting; qPCR, quantitative real-time polymerase chain reaction.

Figure 4.

S100A8/A9^highAML cells display signs of myeloid differentiation. (A-C) AML cell lines were analyzed by flow cytometry for the myeloid maturation markers CD11b and CD14 in S100A8/A9^hi and S100A8/A9^lo cells after 48 hours of culture in the presence of HS-5 CM (A; OCI-AML, n = 5; MOLM-13, n = 5), AML-MSC CM from 6 patients (B; OCI-AML, n = 12; MOLM-13, n = 12, patient ID 11-16; supplemental Table 1), and recombinant human IL-6 (C; OCI-AML, n = 3; MOLM-13, n = 3), based on median fluorescence intensity (MdFI). (D) The myeloid maturation markers CD11b and CD14 were analyzed by flow cytometry in matched-pair (n = 10; patient ID 1-10; supplemental Table 1) PB- and BM-derived S100A8/A9^hi and S100A8/A9^lo AML blasts, as shown in representative histograms (left) and as summarized by fold change (right; S100A8/A9^lo set as 1). (E-F) Surface levels of CD11b and CD14 were analyzed in S100A8/A9^hi and S100A8/A9^lo (OCI-AML, n = 4; MOLM1,3 n = 4) populations after culture for 48 hours with HS-5 CM in the absence or presence of the pan-Jak inhibitor ruxolitinib, as indicated. (G) Surface levels of CD11b and CD14 were analyzed in S100A8/A9^hi cells among the AML cell lines (OCI-AML, n = 5; MOLM-13, n = 5) after 48 hours of culture with HS-5 CM in the absence or presence of the CD36 inhibitor SSO. Data are expressed as the standard error of mean. ∗P < .05; ∗∗P < .01; ∗∗∗P < .001.

Figure 5.

S100A8/A9 expression is associated with enhanced therapeutic resistance. (A) AML cell lines (OCI-AML, n = 5; MOLM-13, n = 5) were incubated for 48 hours in the absence (−, circles) or presence (+, squares) of HS-5 CM and treated with 50 nM doxorubicin for the last 24 hours of culture. Viability was analyzed by annexin-V/7-AAD staining via flow cytometry, and specific cell death by doxorubicin was calculated. (B-C) Similarly, specific cell death by 50 nM doxorubicin of AML cell lines (OCI-AML, n = 3; MOLM-13, n = 2-5) cultured in the presence of HS-5 CM was analyzed in untreated (circles) and treated (squares) samples treated with ruxolitinib (B) or SSO (C). (D) The frequency of S100A8/A9^hi OCI-AML (n = 8) and MOLM-13 (n = 11) cells was determined by flow cytometry after 48 hours of culture in the absence and presence of HS-5 CM and upon treatment with 50 nM doxorubicin for the last 24 hours, as indicated. (E-F) Frequency of S100A8/A9^hi OCI-AML (n = 3-4) and MOLM-13 (n = 4) cells after culture for 48 hours in the presence of HS-5 CM and for the last 24 hours in the presence of 50 nM doxorubicin was determined by flow cytometry upon treatment with ruxolitinib (E) or SSO (F). (G) Free cytosolic calcium was semiquantified in AML cells (OCI-AML, n = 3; MOLM-13, n = 3) cultured for 48 hours in the absence (medium) or presence (CM) of HS-5 CM with the flow cytometric probe Fluo-8. (H) Specific cell death of AML cells (OCI-AML n = 3; MOLM-13 n = 2) by 50 nM doxorubicin was estimated via annexin-V/7-AAD staining in untreated samples and upon scavenging extracellular (EGTA) and free cytosolic (BAPTA) calcium. (I) Log₂-transformed normalized counts of the S100A8 and S100A9 gene expression (from TCGA LAML data set) correlated with the log₂-transformed normalized count of MCL1 by the Spearman test. (J) Cell death of S100A8/A9^lo and of S100A8/A9^hi AML cells (OCI-AML, n = 3; MOLM-13, n = 3) cultured for 24 hours in the presence of HS-5 CM was analyzed upon treatment with increasing concentrations of venetoclax by annexin-V/7-AAD staining in flow cytometry. (K) Frequency of S100A8/A9^hi AML cells (OCI-AML, n = 2; MOLM-13, n = 2) cultured for 24 hours in the absence or presence of HS-5 CM was determined by flow cytometry after treatment with increasing concentrations of venetoclax . (L-M) Intracellular protein levels of Mcl-1 (L, OCI-AML, n = 4; MOLM-13, n = 4) and Bcl-2 (M, OCI-AML, n = 3; MOLM-13, n = 3) were measured by flow cytometry in S100A8/A9^lo and S100A8/A9^hi subsets of AML cells cultured for 48 hours in the presence of HS-5 CM. Data are expressed as the standard error of mean. ∗P < .05; ∗∗P < .01; ∗∗∗P < .001; ns, not significant.