Downregulation of stromal syntenin sustains AML development

Abstract

The crosstalk between cancer and stromal cells plays a critical role in tumor progression. Syntenin is a small scaffold protein involved in the regulation of intercellular communication that is emerging as a target for cancer therapy. Here, we show that certain aggressive forms of acute myeloid leukemia (AML) reduce the expression of syntenin in bone marrow stromal cells (BMSC). Stromal syntenin deficiency, in turn, generates a pro-tumoral microenvironment. From serial transplantations in mice and co-culture experiments, we conclude that syntenin-deficient BMSC stimulate AML aggressiveness by promoting AML cell survival and protein synthesis. This pro-tumoral activity is supported by increased expression of endoglin, a classical marker of BMSC, which in trans stimulates AML translational activity. In short, our study reveals a vicious signaling loop potentially at the heart of AML-stroma crosstalk and unsuspected tumor-suppressive effects of syntenin that need to be considered during systemic targeting of syntenin in cancer therapy.

Keywords: cell-to-cell communication; syntenin; tumor aggressiveness; tumor-stroma.

Figure 1. Aggressive human FLT3‐ITD⁺ AML downregulates stromal syntenin expression through miR‐155

1. The volcano plot was generated from the publicly available database (GSE97194). It illustrates the changes in gene expression, measured by microarray, with log₂‐fold change (x‐axis) and −log₁₀ (P‐value) (y‐axis), in BMSC isolated from C57BL/6 mice inoculated with AML (AML‐BMSC) compared to BMSC isolated from control animals (normal‐BMSC). Genes, down‐ or upregulated in AML‐BMSC compared to normal BMSC, were selected based on a P‐value < 0.01 and a difference > 2 (see Dataset EV1 for raw data).

2. Dot plot representing Sdcbp expression in AML‐BMSC, compared to normal BMSC, according to the AML subtypes inoculated in the animals (MLL‐ENL + FLT3‐ITD, MLL‐ENL, and AML1‐ETO9a). The selected threshold is based on a P‐value < 0.01 and a difference > 2 (see Dataset EV2 for raw data).

3. Real‐time PCR analysis of syntenin mRNA expression in HS5 transfected with hsa‐miR‐155‐5p (miR‐155; 30 nM) or control (miR‐Ctrl). Values were normalized to housekeeping L32/GAPDH genes. Data represent the mean ± SEM of three independent experiments performed in duplicate. Statistical analysis was performed using Student's t‐test.

4. Left, western blots on cell lysates from HS5 cells transfected with hsa‐miR‐155‐5p (miR‐155) or control (miR‐Ctrl), in the presence or absence of miR‐155‐inhibitor (anti‐155), illustrating the miR‐155‐dependent loss of syntenin expression. GAPDH is used as loading control. Right, histograms representing mean syntenin signal intensities ± SEM, relative to GADPH, calculated from the analysis of three independent experiments. Statistical analysis was performed using the one‐way analysis of variance (ANOVA).

5. Left, western blots of total cell lysates from HS5 transfected with miR‐155 inhibitor (anti‐155) or control, cultured in the absence or presence of U937 (AML‐ETO, miR‐155^low) or MOLM14 (AML‐FLT3‐ITD, miR‐155^high) cells for 24 h, illustrating the impact on syntenin signals. GAPDH was used as loading control. Right, histograms representing mean signal intensities ± SEM, relative to loading control, calculated from the analysis of three independent experiments. Statistical analysis was performed using the one‐way analysis of variance (ANOVA).

Source data are available online for this figure.

Figure 2. FLB1 blasts gain in aggressiveness upon serial transplantation in a syntenin‐deficient host

1. Scheme illustrating the serial transplantation assay. Briefly, WT and syntenin knock‐out (synt‐KO) mice were injected, in parallel, with FLB1 cells. Upon complete invasion of the bone marrow, FLB1 cells were harvested and re‐injected for a new round of transplantation. FLB1 were serially transplanted into WT and synt‐KO animals in parallel for four rounds (graft 1 to graft 4)

2. FACS analysis of blood samples collected from mice on day 14 post‐injection. Leukemia burden is expressed as percentage of FLB1 (CD45.1⁺) cells relative to total CD45⁺ (CD45.1⁺+CD45.2⁺) in the peripheral blood ± SEM. Statistical analysis was performed using the two‐way analysis of variance (ANOVA). Note the stability of the leukemia burden upon serial transplantation in WT animals. Note leukemia “outbreak,” from graft 3 on, in synt‐KO animals.

Source data are available online for this figure.

Figure 3. Education by a syntenin‐deficient host sustains cell survival and protein synthesis in leukemia blasts

1. Volcano plot illustrating the proteins that are differentially expressed, with log₁₀ levels (x‐axis) and –log (P‐value) (y‐axis), in FLB1‐G4KO (aggressive) blasts versus FLB1‐G4WT blasts (see Dataset EV3 for raw data).

2. Molecular and cellular functions identified using ingenuity pathway analysis (IPA) based on the selected threshold (P < 0.005; difference > 1). The bar diagram indicates the most significant molecular and cellular functions upregulated in FLB1 cells transplanted for four rounds in Synt‐KO mice compared to FLB1 cells transplanted in WT animals. The x‐axis indicates the number of molecules involved in the corresponding y‐axis function.

3. FACS analysis of FLB1 (CD45.1⁺) cells collected after graft 4 from WT (FLB1‐G4WT) or synt‐KO animals (FLB1‐G4KO). Apoptosis was evaluated after 72 h of maintenance in complete RPMI media, using double staining with 7AAD/AnnexinV, as illustrated on the representative dot blots (Left panel). Results from three independent experiments are expressed as mean value of living (AnnexinV⁻, 7AAD⁻) FLB1cells ± SEM. Statistical analyses were performed using the nonparametric Mann–Whitney U‐test (Right panel).

4. Left, scheme of the experiment. FLB1‐G4WT cells were injected at day 0 into WT mice, while the inoculation of FLB1‐G4KO cells into synt‐KO mice was delayed for 1 week to obtain similar stages of disease progression in the two different hosts at day 21. To address in vivo protein synthesis, O‐propargyl‐puromycin (OPP) was injected intraperitoneally. One hour later, BM was collected and OPP incorporated into nascent polypeptide chains was fluorescently labeled via “Click‐it Chemistry.” Right, graph representing the relative levels of protein synthesis in host cells (CD45.1‐) and in FLB1 cells (CD45.1⁺) relative to the BM cells from OPP‐untreated mice ± SEM as measured by FACS and calculated from the analysis of three independent mice. Statistical analysis was performed using the nonparametric Mann–Whitney U‐test.

5. Left, western blots of FLB1‐G4WT and ‐G4KO total cell lysates analyzed for different markers, as indicated. Right, histograms representing FLB1‐G4KO mean signal intensities ± SEM, relative to signals obtained with FLB1‐G4WT lysates, calculated from the analysis of five independent mice. Statistical analysis was performed using the nonparametric Mann–Whitney U‐test.

6. Left, western blot from FLB1‐G4WT and ‐G4KO total cell lysates illustrating EEF1A2 expression. The ubiquitous GRB2 signal was used as loading control. Right, histogram representing FLB1‐G4KO EEF1A2 mean signal intensities ± SEM, relative to signals obtained with FLB1‐G4WT lysates, calculated from the analysis of five independent mice. Statistical analysis was performed using the nonparametric Mann–Whitney U‐test.

Source data are available online for this figure.

Figure 4. Syntenin‐deficient BMSC suffice to educate AML cells for aggressiveness

1. Scheme for ex vivo long‐term FLB1‐BMSC co‐culture experiments coupled to in vivo leukemia burden assays. FLB1 cells were co‐cultured with murine BMSC isolated from WT or synt‐KO mice. After 1 month of co‐culture, surviving FLB1 cells, “educated” by murine BMSC (WT or synt‐KO), were injected into the retro‐orbital vein of WT mice. Leukemia progression was assessed weekly by FACS analysis as described before.

2. Results of FACS analyses showing FLB1 (CD45.1⁺) cell frequencies measured in PB, BM, and the spleen of the different mice in the different groups 15 days after cell inoculation. Results are expressed as percentage of FLB1 (CD45.1⁺) cells relative to total CD45⁺ cells, and calculated as means ± SEM. Statistical analysis was performed using one‐way ANOVA test.

Source data are available online for this figure.

Figure 5. Syntenin deficiency in BMSC affects the subcellular expression of endoglin

1. Left, FACS analysis of cell surface BMSC markers (as indicated) in WT (black lane) or Synt‐KO (gray lane) murine expanded BMSC (passage 4). Right, histogram representing the fold change in the median of fluorescence relative to that measured in BMSC WT ± SEM, calculated from the analysis of five independent experiments. Statistical analysis was performed using parametric Student's t‐test.

2. Upper, FACS detection of the cell surface expression of endoglin in HS5 and HS27a stromal cells transfected with siRNA‐targeting syntenin (siSynt; gray histogram) or control siRNA (siCtrl; black histogram). Lower, histogram representing the median of fluorescence of endoglin ± SEM. Values were calculated from three independent experiments. Statistical analysis was performed using the two‐way analysis of variance (ANOVA).

3. HS5 and HS27a cells were transfected with siRNA‐targeting syntenin (siSynt) or control siRNA (siCtrl). After 48 h, the media were changed and the transfected cells were cultured in medium containing EV‐depleted FCS (10%) for 16 h. Conditioned medium was submitted to differential centrifugation and small EVs were pelleted at 100,000 g. Upper, total cell lysates and corresponding pelleted extracellular particles (small EVs) were analyzed for indicated markers. Lower, histograms represent the endoglin protein levels in HS5 and HS27a cell lysates (normalized to GAPDH) and in corresponding small EVs. Values were calculated from three independent experiments. Statistical analysis was performed using one‐way ANOVA test.

4. Surface plasmon resonance experiment illustrating the direct syntenin–endoglin interaction, tested with recombinant polypeptides. Increasing concentrations of syntenin (comprising only the tandem‐PDZ domains + C‐terminal domain; 0.5 μM to 120 μM) were perfused on an immobilized peptide corresponding to the last 25 C‐terminal amino acids of wild‐type endoglin.

5. Left, representative confocal micrographs showing the steady‐state distributions of endogenous syntenin and endogenous endoglin in HS5 and HS27a stromal cells. In merge, nuclei are stained with DAPI (blue), syntenin is in red, and endoglin is in green. Right, Pearson correlation coefficient in 16 HS5 and 14 HS27a cells using the JACoP plugin on ImageJ. Dot plots with bars representing the mean Pearson coefficient ± SEM from three independent experiments.

Source data are available online for this figure.

Figure EV1. (Related to Fig 5). Syntenin–endoglin relationship

Effect of loss of syntenin on endoglin expression (A–D) and endoglin–syntenin interaction assays (E–G).

1. Representative confocal micrographs showing the subcellular distribution of endogenous endoglin in HS5 and HS27a cells, depleted for syntenin expression (siSynt) or treated with control siRNA (siCtrl). Nuclei are stained with DAPI (blue), and endoglin is in green.

2. Real‐time PCR analysis of endoglin mRNA expression in HS5 and HS27a cells depleted for syntenin (siSynt) and endoglin (siEng) or treated with control siRNA (siCtrl). Values were normalized to housekeeping L32/GAPDH genes. Data represent the mean ± SEM of four independent experiments performed in duplicate.

3. Left, FACS analysis of endoglin levels in HS5, transfected with miR‐155‐5p (miR‐155) or control (miR‐Ctrl). Cells were immunostained with the anti‐human endoglin monoclonal antibody or isotype control antibody (untreated). Right, data represent the mean of the means of fluorescence intensity ± SEM, collected from three independent experiments. Statistical analysis was performed using the nonparametric Mann–Whitney U‐test.

4. HS5 cells were transfected with miR‐155‐5p (miR‐155) or control (miR‐Ctrl). After 48 h, the media were changed and the transfected HS5 cells were cultured in medium containing EV‐depleted FCS (10%) for 16 h. Conditioned medium was submitted to differential centrifugation and small EVs were pelleted at 100,000 g. Total cell lysates and corresponding pelleted extracellular particles (small EVs) were analyzed for indicated markers.

5. Surface plasmon resonance experiment illustrating the direct interaction of endoglin with GIPC1, used as positive control. Sensorgrams illustrating the binding of GIPC1, at increasing concentrations of recombinant GIPC1 (0.5 μM to 120 μM), to an immobilized peptide corresponding to the last 25 C‐terminal amino acids of wild‐type endoglin.

6. Langmuir graph showing the binding (in resonance units, RU) of recombinant syntenin (blue curve) and GIPC1 (red curve) as observed at equilibrium, at increasing concentrations of analyte. KD (apparent) was calculated from the protein concentration required to observe half of the maximal response.

7. Proximity ligation assay (PLA) detecting the proximity of endogenous syntenin and endoglin in HS5 cells. Left, representative images of confocal microscopy showing the PLA signals. Nuclei were stained with DAPI. Syntenin and endoglin antibodies were incubated alone or together. Each picture is representative of a typical cell staining pattern observed in five fields chosen at random. Right, the quantification of the number of PLA dots per nucleus is represented as the mean values ± SEM from a single experiment performed in duplicate. Statistical analysis was performed using the one‐way analysis of variance (ANOVA).

Figure 6. High endoglin cell surface expression in BMSC promotes AML aggressiveness

1. FACS analysis of cell surface expression of endoglin in stromal cells from AML patients with wild‐type FLT3 (WT) or bearing FLT3‐ITD mutation. As indicated in Dataset EV4, endoglin expression was assessed in samples where at least 300 stromal cells were reliably detected. The violin plots represent the median fluorescence of endoglin expression at the surface of the stromal cells minus the value of the isotypic control. Each dot represents one patient. Statistical analysis was performed using the nonparametric Mann–Whitney U‐test.

2. Scheme illustrating long‐term co‐culture experiments with syntenin‐ and endoglin‐deficient stromal cells addressing stromal effects on AML protein synthesis. HS5 cells were transfected with siRNA‐targeting syntenin (siSynt), endoglin (siEng), or control siRNA (siCtrl). Twenty‐four hours later, human AML HL60, U937, or OCI‐AML3 cells were co‐cultured with these HS5 cells. After 1 week of co‐culture, AML cells were collected and re‐seeded on freshly transfected HS5 cells. After 3 weeks of co‐culture, AML cells were collected and treated with O‐propargyl‐puromycin (OPP) to measure protein synthesis by FACS.

3. Upper, FACS detection of OPP incorporation in AML cells educated on HS5 treated as indicated. Untreated refers to AML cells not treated with OPP. Lower, graph representing the relative levels of protein synthesis in AML cells (HL60, U937, or OCI‐AML3) normalized to untreated AML cells ± SEM. Values were calculated from four independent experiments. Statistical analysis was performed using the one‐way analysis of variance (ANOVA).

Source data are available online for this figure.

Figure EV2. (Related to Fig 6). Endoglin/syntenin suppression in HS5 cells (control transfection experiments)

1. A

   Western blots of the lysates of HS5 cells transfected with control siRNA (siCtrl), with control siRNA and siRNA‐targeting syntenin (siSynt + siCtrl) or endoglin (siEng + siCtrl), or siRNA‐targeting syntenin and endoglin (siSynt + siEng; 30 nM), showing the effects on syntenin levels. GAPDH was used as control.

2. B

   FACS analysis of the endoglin expression at the surface of HS5 cells transfected with control siRNA (siCtrl, dark gray), with control siRNA and siRNA‐targeting syntenin (siSynt + siCtrl, light gray) or endoglin (siEng + siCtrl, blue), or siRNA‐targeting syntenin and endoglin (siSynt + siEng; red). Isotype control antibody was used as control (unstained).

3. C

   HS27a (stroma) cells were transfected with siRNA‐targeting syntenin (siSynt), endoglin (siEng), both (siSynt + siEng), or control siRNA (siCtrl). After 24 h, HL60 (leukemia) cells were put in contact with the transfected HS27a cells. After 1 week of co‐culture, HL60 cells were collected and re‐seeded on freshly transfected HS27a cells. After 3 weeks of co‐culture, AML cells were collected and treated with O‐propargyl‐puromycin (OPP) to measure protein synthesis by FACS. Upper panel, FACS detection of OPP incorporation in HL60 cells educated on HS27a treated as indicated. Untreated refers to HL60 cells not treated with OPP. Lower panel, graph representing the relative levels of protein synthesis in HL60 cells, normalized to untreated HL60 cells ± SEM. Values were calculated from four independent experiments. Statistical analysis was performed using the one‐way analysis of variance (ANOVA).

4. D–F

   HS5 cells were transfected with siRNA‐targeting syntenin (siSynt), endoglin (siEng), both (siSynt + siEng), or control siRNA (siCtrl). Twenty‐four hours later, human AML HL60, U937, or OCI‐AML3 cells were co‐cultured with these HS5 cells. After 1 week of co‐culture, AML cells were collected and re‐seeded on freshly transfected HS5 cells. After 3 weeks of co‐culture, AML cells were collected for FACS analysis of (D) EEF1A2, (E) pAKT (Ser473), and (F) pRPS6 (Ser235), respectively. The histograms represent the fold change in mean of fluorescence intensity relative to the signal in AML cells maintained on HS5 cells transfected with siCtrl ± SEM calculated from the analysis of three independent experiments. Statistical analysis was performed using the nonparametric two‐way analysis of variance (ANOVA).

Figure 7. Model recapitulating the findings presented in this study

We here summarize cues that might provide novel insight into the mechanisms at work in tumor–stroma crosstalk.

1. A

   Aggressive AML (FLT3‐ITD) cells overexpressing miR‐155 can suppress syntenin expression in stromal cells.

2. B

   AML blasts confronted with a syntenin‐deficient stroma acquire a cell survival advantage associated with increased levels of protein synthesis. This AML response relies on the upregulation of EEF1A2, an elongation factor and notable activator of the AKT/RPS6 signaling pathway.

3. C

   Syntenin loss/dysregulation increases the level of endoglin at stromal cell surfaces, while decreasing the loading of endoglin into exosomes. Stromal endoglin, and likely high endoglin at stromal cell surfaces, is needed for the in trans support of AML translational activity by a syntenin‐deficient stroma.