Human bone marrow mesenchymal stem cells secrete endocannabinoids that stimulate in vitro hematopoietic stem cell migration effectively comparable to beta-adrenergic stimulation

Department of Stem Cell Sciences, Institute of Health Sciences, Center for Stem Cell Research and Development (PEDI-STEM), Hacettepe University, Ankara, Turkey; Department of Histology and Embryology, Faculty of Medicine, Yuksek Ihtisas University, Ankara, Turkey; Department of Analytical Chemistry, Faculty of Pharmacy, Hacettepe University, Ankara, Turkey; Pediatric Hematology, BMT Unit, Hacettepe University Children’s Hospital, Ankara, Turkey; Department of Medical Genetics, Faculty of Medicine, Hacettepe University, Ankara, Turkey; Department of Microbiology, Faculty of Medicine, Hacettepe University, Ankara, Turkey; Department of Histology and Embryology, Faculty of Medicine, Hacettepe University, Ankara, Turkey

Hematological malignancies are currently treated with hematopoietic stem cell transplantation (HSCT) [1]. Rapid and sustained recovery of hematopoietic functions after HSCT correlates with the number of CD34 + hematopoietic stem cells (CD34 + HSCs) infused [2]. CD34 + HSCs reside mainly in the bone marrow (BM) microenvironment(s) and mobilize to the peripheral blood (PB) after administration of growth factors or antagonists, such as granulocyte colony-stimulating factor (G-CSF). The release mechanism of these cells from the hematopoietic microenvironment is still not completely understood [3,4]. Infection and stress increase HSC migration through lipopolysaccharide (LPS) release [5] and activation of the sympathetic nervous system via β-adrenergic receptors (Adrβ1, Adrβ2, and Adrβ3) [6,7] in a circadian rhythm [8]. However, they may not be the only regulators of HSC migration [9]. Chemokine stromal derived factor-1 (SDF-1, also termed CXCL12) and its major receptor, CXCR4, are crucial in mediating both retention and mobilization of HSCs. G-CSF and the CXCR4 antagonist AMD3100 are the only U.S. Food and Drug Administration-approved agents for patients when HSCs fail to mobilize [10]. G-CSF-based migration requires a multiday dosing regimen and is associated with some morbidity and rare but serious complications [11].
The sympathetic activation and the subsequent betaadrenergic system involvement is well described in BM under physiologic and stress conditions, but it is not the only/ efficient stimulator for HSC mobilization. eCBs functioning as neurotransmitters and paracrine factors may play role in HSC mobilization similar to mobilization induced as a result of stress-induced sympathetic hyperactivity in the human BM microenvironment. Therefore, the utility of eicosanoidbased therapeutic strategies including cannabinoids is being investigated but still requires further investigation for improving HSC mobilization [6][7][8]22,23]. Here, we assessed the following: (1) if healthy donor HSCs and MSCs release eCBs (AEA and 2-AG) and express CB1 and CB2 receptors, (2) if HSCs migrate toward eCBs and MSCs, and (3) if there is a difference between cells obtained from donors after G-CSF treatment.
The aim of the present study was to elucidate the role of the eCB system in the human BM microenvironment on migration of HSCs and the effects of G-CSF thereon. We found expression of eCBs in stromal BM microenvironments and cannabinoid receptors in HSCs. HSC migration in the presence of the eCB ligands (AEA and 2-AG) increased compared with controls and decreased in presence of the eCB receptor antagonists AM281 and AM630. Therefore, eCBs may be a novel candidate to enhance or facilitate/accelerate G-CSFmediated HSC mobilization in a clinical setting.

LC-ESI-MS/MS
2-AG and AEA levels were measured in BM, PB plasma, and MSC supernatant by liquid chromatography-tandem electrospray ionization-mass spectrometry (LC-ESI-MS/MS) (#LCMS-8030, Shimadzu, Japan) following the optimization protocol with standards (protocol adopted from Bradshaw et al [28].). The daily calibration curves were used for quantification of AEA and 2-AG; all samples were just prepared before the run and analyzed twice. The solvent controlled validation results are shown as Supplementary  Fig. E1 (online only; available at www.exphem.org). The experiment was performed with 10 independent samples for each group. Polymeric sorbent-based solid-phase extraction (SPE) cartridges (Strata-X0.5 mL) were used for high-throughput sample preparation. Briefly, 250 µL of BM and PB plasma or 7 mL of the supernatant of 7 × 10 6 MSCs were passed through the cartridges activated with methanol. Cartridges were washed with a water:methanol mixture (45:55, v/v) and elution of 2-AG (#8923, Sigma-Aldrich, Germany) and AEA (#A0580, Sigma-Aldrich, Germany) was performed with 2 mL of MeOH. The calibration curves of 2-AG and AEA were constructed with the peak area of the analyte versus the concentration. The chromatographic separation was achieved on a C18 column (Hypersill ODS4, 50 × 3.0 mm, 2.1 µm) using a mobile phase consisting of acetonitrile containing 0.1% formic acid and water containing 0.1% formic acid at 0.3 mL/min flow rate.
qRT-PCR CNR1, CNR2 (genes for CB1 and CB2, respectively), ADRB1, ADRB2, and ADRB3 (genes for Adrβ1, Adrβ2, and Adrβ3, respectively) expression levels were measured in MNCs and MSCs of both groups by quantitative reverse transcription polymerase chain reaction (qRT-PCR). The experiment was performed with six independent samples for each group and three repeats for each sample. Total RNA was isolated from MNCs and MSCs using the RNAeasy mini RNA isolation kit (#74104, Qiagen, USA) following the manufacturer's instructions. Total RNA concentrations and ratios were determined by spectrophotometry (Nanodrop 2000, Thermo Fisher Scientific, USA) and stored at −80°C. cDNA was generated from 200 ng of total RNA using the ProtoScript® First Strand cDNA synthesis kit (#E6300, New England BioLabs, USA) following the manufacturer's instructions. qRT-PCR was performed with the PowerUp™ SYBR® Green Master Mix (#A25741, Thermo, USA) using the ViiA™ 7 Real-Time PCR System (Thermo, USA). Relative mRNA expression analysis was calculated by delta delta Ct method and ViiA™ 7 Software (version 1.2.4).
For the migration assay itself, CD34 + PBSCs were placed onto prewetted filters of Transwells with 100 µL of high-glucose Dulbecco's modified Eagle medium (DMEM-HG). DMEM-HG containing SDF-1, the cannabinoid agonists AEA and 2-AG (both at 30 nmol/L, 1 µmol/L, 50 µmol/L; Sigma-Aldrich, Germany) with or without AM281 and AM630 (both at 10 µmol/L; Tocris Bioscience, UK) were added to the lower wells alone or combined with LPS-stimulated or LPS-unstimulated MSCs in 600 µL of DMEM-HG. Cells were allowed to migrate for 4 h at 37°C in a humidified atmosphere with 5% CO2. Filters were then removed from the chambers and counting was performed with Turk's solution. All Transwell assay experiments were performed with six independent PBSC donors and three repeats for each donor.

Statistical analysis
Descriptive results were presented as mean ± SEM and median (minimum-maximum). Normality of the distribution of variables in every study group was evaluated by the Shapiro-Wilk test and differences between study groups (with nonparametric distribution) were assessed by Wilcoxon's test. p values < 0.05 were considered statistically significant.

MSCs from G-CSF-treated and G-CSF-untreated donors are phenotypically and morphologically similar
MSCs showed plastic adherence and fibroblastic morphology; positive expression for CD73, CD44, CD90, and CD29; and were negative for hematopoietic markers, including CD34 and CD45, in the G-CSF-untreated and G-CSF-treated group, respectively (Figs. 1A and 1B). Expression of MSC markers was similar in the G-CSF-untreated and G-CSF-treated groups (p ≥ 0.05; Fig. 1B). Adipogenic differentiation was confirmed by morphology and the amount of Oil Red O (Figs. 1C and 1E). Osteogenic differentiation was confirmed by morphology using Alizarin Red S and the production of calcium phosphates (Figs. 1D and 1E). G-CSF-treated and G-CSF-untreated MSCs exhibited similar adipogenic and osteogenic differentiation capacity (p ≥ 0.05). The morphological, differentiation, and immunophenotypic characteristics confirmed stromal and multipotential nature of the MSCs used in this study [31].

Endogenous AEA and 2-AG was found in BM and PB plasma and MSC supernatant
MSCs secrete 2-AG and AEA in culture supernatants. AEA was detected in PB and BM plasma of G-CSF-treated and G-CSF-untreated groups at similar levels (Table 1). However, the 2-AG level in PB of the G-CSF-treated group was significantly higher compared with the BM plasma samples (p = 0.01; Table 1). The 2-AG level was higher in PB plasma compared with AEA in both groups, but the difference was not statistically significant ( Table 1). MSCs of G-CSFtreated and G-CSF-untreated groups exhibited a similar profile of AEA and 2-AG secretion with PB and BM cells (Table 1).

BM-MNCs and CD34+ HSCs express CB1 and CB2 receptor and Adrβ subtypes
MNCs and CD34 + cells expressed significantly higher levels of CB1 and CB2 receptors compared with MSCs of those groups (p = 0.001 for all groups) by FC (Fig. 2). CB1  -maximum). In the G-CSF-treated group, the level of 2-AG was found to be higher in PB plasma than in BM plasma (*p < 0.05, n = 10 for each group; *comparison of PB vs BM for G-CSF-treated group). receptor expression was highest on CD34 + HSCs. This difference was significant for the G-CSF-untreated group only (p = 0.02; Fig. 2A). Data were confirmed by qRT-PCR, with which CB1 and CB2 gene expressions were detected in MNCs, but not in MSCs (Figs. 2B and 2C). CB2 expression of MNCs was higher in the G-CSF-untreated group compared with the treated group (p = 0.04; Fig. 2B). MNCs, CD34 + HSCs, and MSCs expressed various subtypes of Adrβ subtypes by FC; however, MNCs and MSCs mainly expressed Adrβ2, as confirmed by qRT-PCR (Fig. 2). MNCs and CD34 + HSCs exhibited significantly higher expression of Adrβ2 (p = 0.02 and p = 0.004 for MNC; p = 0.001 and p = 0.001 for CD34 + HSCs in the G-CSF-untreated and G-CSF-treated groups, respectively) compared with MSCs by FC ( Fig. 2A). CD34 + HSCs exhibited higher expression for Adrβ2 and Adrβ3 in all groups compared with MNCs, but the difference was only significant for Adrβ3 (p = 0.001 and p = 0.013 for MNC; p = 0.001 and p = 0.002 for MSC in the G-CSF-untreated and G-CSF-treated groups, respectively) ( Fig. 2A). Increased Adrβ2 expression was found in G-CSF-treated MNCs compared with untreated (p = 0.02) MNCs by qRT-PCR (Fig. 2B). Differences in expression of Adrβ1 as measured by FC were not found.

PBSCs migration toward MSCs is blocked by CB receptor and β-AR antagonists
CD34 + PBSCs effectively migrated toward LPS-stimulated and LPS-unstimulated MSCs in the coculture system (Figs. 3B and 3E). However, the LPS-stimulated MSCs exhibited higher eCB-receptor-mediated migration stimulation compared with unstimulated cells. The CB1 antagonist AM281 (p = 0.03), the CXCR4 antagonist AMD3100 (p = 0.05), and the Adrβ inhibitor SR59230A (p = 0.03) all significantly blocked migration toward LPS-stimulated MSCs (Fig. 3E). The difference was not significant for the CB2 antagonist AM630 (Fig. 3E). Inhibition of migration toward LPS-unstimulated MSCs was not significant. The LPS-only control group revealed a response at the baseline with almost no migrating CD34 + PBSCs.

G-CSF untreated group G-CSF treated group
Ca 2+ (mg/dL) 15,79 ± 5,1 16,97 ± 3,7 ORO dye (mg/mL) 1,89 ± 1,3 1,97 ± 1,5 C D survival [32] under stress conditions, which results in sympathetic Adrβ stimulation [6,7]. In this study, AEA and 2-AG caused CD34 + PBSC migration, reflecting migration via the β-adrenergic system. BM-HSCs and MSCs express Adrβ2 and Adrβ3 [33]. Adrβ signaling and LPS release favor the positive effect of G-CSF on HSC mobilization [6,34]. PB and BM plasma and MSC supernatants contained AEA and 2-AG. AEA and 2-AG levels in the PB plasma of healthy controls were found previously to be 0.8 ± 0.12 and 19.0 ± 2.61 nmol/L, respectively, using LC-ESI-MS/MS [35]. Levels of AEA and 2-AG in PB were previously found to be 0.56 ng/mL and 2,0 ng/mL in healthy donors [36]. We report levels of AEA and 2-AG of approximately 50 times higher compared with findings by Jean-Gilles et al. [35] and Quercioli et al. [36]. However, others reported a range from pmol/L to µmol/L levels for AEA and 2-AG at a different matrix [37] (Human Metabolite Data Base, www.hmdb.ca). AEA levels (0.36 ± 0.14 ng/mL) and 2-AG (6.26 ± 2.10 ng/mL) were also measured in PB plasma of patients during traumatic stress exposure and posttraumatic stress disorder [38]. Our study on AEA and 2-AG levels in PB plasma of G-CSF-treated and G-CSF-untreated donors correlated well with that study. eCB levels were not examined in BM plasma of humans or other species previously. AEA levels in BM and PB plasma were generally equivalent between G-CSF-treated and G-CSFuntreated donors in our study. The 2-AG level was higher in PB plasma compared with BM plasma in the G-CSF-treated group. The 2-AG level in PB increased approximately 10fold after G-CSF treatment compared with the untreated group. This may suggest a potential function for 2-AG during G-CSF treatment.
Rossi and colleagues [39] reported a gradual decrease in AEA and 2-AG levels secreted by human BM-MSCs from passage 1 (AEA: 5 pmol/mg protein and 2-AG: 11 pmol/ mg protein 2-AG, at passage 1). In our study, we used passage 1 MSCs to assess CD34 + PBSC migration. Human adipose tissue-derived MSCs have been reported to secrete AEA and 2-AG (AEA: 3.5 pmol/mg protein and 2-AG: 7.3 pmol/mg protein) [40]. These findings were consistent with our results showing AEA and 2-AG secretion by BM-MSCs. Therefore, endogenous eCBs found in BM plasma samples are likely to be secreted by MSCs.
Here, we report for the first time the distribution of eCB receptor and Adrβ subtypes on human BM and PB cells simultaneously with or without stimulation by G-CSF. We found significantly higher CB1 and CB2 receptor expression on MNCs and CD34 + HSCs compared with MSCs. There are limited studies reporting distribution of CB1 and/or CB2 receptors on HSCs [14,15]. Although CB1 receptor expression is reported in a single murine T-lymphoid cell line, CB2 receptor expression was found in a multitude of myeloid, macrophage, erythroid, B/T-lymphoid, mast cell lines [14]. In rodents, CB1 and CB2 receptors were detected in BM-MSCs and HSCs [15].
We report that the migratory effect of AEA (a full agonist for CB1 and CB2) and 2-AG (a full agonist for CB2 and a weak agonist for CB1) [12,13] was generally inhibited by CB1 and CB1 receptors separately. Conversely, AEA activated CB2 receptors mainly when applied at 50 µmol/L and its effect wais blocked with AM630, possibly due to the high dose activation ability for and a shift to CB2 receptors.
In our first set of experiments, we showed higher Adrβ2 and CB2 receptor expression in G-CSF-treated MNCs compared with untreated cells. We also demonstrated that AEA and 2-AG mediate CD34 + PBSC migration toward MSCs in a coculture system. LPS-stimulated MSCs exhibited higher eCB-receptor-mediated migration stimulation effect to CD34 + PBSCs compared with unstimulated. Those findings were not shown previously. Human BM-MSCs endogenously secrete AEA and 2-AG, which in turn induces CD34 + PBSC migration in vitro.
Our study results are limited due to the small number of human samples used due to ethical considerations. Second, because our assay does not simulate a 3D BM niche accurately, the true mobilizing effects of eCBs on HSCs should be tested further in vivo. These limitations, however, do not constrain future in vivo and clinical studies because statistical accuracy was validated at the beginning of the study. We have demonstrated that the effects of the components of the eCB system and other mobilizing agents on the interaction of HSCs with MSCs should be further assessed in animal models and in clinical trials. In particular, the lower Adrβ2 and CB2 receptor expression pattern in G-CSF-treated MNCs should be confirmed with high numbers of fresh donor samples immediately after treatment.
Rapid and sustained recovery of hematopoietic functions after HSCT correlates with the number of CD34 + HSCs infused. New solutions should be explored to increase the number of CD34 + HSCs. The sympathetic activation and the subsequent beta-adrenergic system involvement is well described, but it is not the only/efficient stimulator for HSC mobilization. Therefore, the utility of eicosanoid-based therapeutic strategies including cannabinoids requires investigation [8,18]. If the mechanism of mobilization of HSCs by eCB agonists or antagonists can be controlled or regulated, then this may result in the development of clinically applicable new mobilization strategies and a better understanding of BM niche dynamics and HSCT strategies.
In conclusion, we found an important role for both the eCB systems and β-adrenergic systems in the migration of HSCs and demonstrate interactions between the HSCs and MSCs of G-CSF-treated and G-CSF-untreated healthy age-matched donors. The eCB system works well in both G-CSF-treated and G-CSF-untreated donors. Therefore, cannabinoid agonists may be strong candidates for new potential therapies of various hematological diseases.