Loss of bisecting GlcNAcylation on MCAM of bone marrow stoma determined pro-tumoral niche in MDS/AML

Bone marrow (BM) stroma plays key roles in supporting hematopoietic stem cell (HSC) growth. Glycosylation contributes to the interactions between HSC and surrounding microenvironment. We observed that bisecting N-acetylglucosamine (GlcNAc) structures, in BM stromal cells were significantly lower for MDS/AML patients than for healthy subjects. Malignant clonal cells delivered exosomal miR-188-5p to recipient stromal cells, where it suppressed bisecting GlcNAc by targeting MGAT3 gene. Proteomic analysis revealed reduced GlcNAc structures and enhanced expression of MCAM, a marker of BM niche. We characterized MCAM as a bisecting GlcNAc-bearing target protein, and identified Asn 56 as bisecting GlcNAc modification site on MCAM. MCAM on stromal cell surface with reduced bisecting GlcNAc bound strongly to CD13 on myeloid cells, activated responding ERK signaling, and thereby promoted myeloid cell growth. Our findings, taken together, suggest a novel mechanism whereby MDS/AML clonal cells generate a self-permissive niche by modifying glycosylation level of stromal cells.


INTRODUCTION
The bone marrow (BM) niche is part of a physiological microenvironment in which hematopoietic stem cells (HSCs) are maintained and respond to regulatory signals under various physiological conditions [1]. BM niche dysfunction strongly affects development of hematological malignancies, and vice versa. BM stroma plays major roles in regulation of a variety of cellular processes, particularly quiescence, differentiation, proliferation, maturation, and apoptosis of HSCs [2,3]. Conversely, the BM niche can be remodeled by hematopoietic malignant cells to generate a pro-tumoral niche surrounding [4,5]. Crosstalk between BM stromal cells and malignant cells is thus strongly involved in disease initiation and progression.
Glycosylation is the most common post-translational modification of proteins, and specific glycan patterns frequently serve as stem cell markers [6,7]. Numerous types of glycoconjugates have been shown to interfere with neoplastic cell processes or microenvironments of these cells, leading to malignant progression [8,9]. We demonstrated in 2013 that human BM stromal cells HS27a (but not HS5) facilitated engraftment of clonal cells from myelodysplastic syndrome (MDS) patients, a process mediated by highly expressed melanoma cell adhesion molecule MCAM/CD146 [10]. Our follow-up 2015 study showed that HS5, in comparison with HS27a, have higher expression of bisecting N-acetylglucosamine (GlcNAc) (β1,4-linked GlcNAc attached to core β-mannose residue, catalyzed by MGAT3) [11]. Bisecting GlcNAc modification regulates physicochemical properties of numerous cell surface glycoproteins, notably integrins, growth factor receptors, and adhesion molecules [12][13][14]. Our analyses demonstrated that levels of MGAT3 and responding bisecting GlcNAc on BM stroma were significantly lower for MDS and acute myeloid leukemia (AML) patients than for healthy donors (HD). We evaluated the underlying mechanism whereby bisecting GlcNAc remodels the BM niche by modulating MCAM on stromal cells, and thereby affects proliferation of MDS/AML clonal cells.

MATERIALS AND METHODS Cell culture
Myeloid leukemia cell line KG1a, and BM-derived stromal cell lines HS5 and HS27a, were kindly donated by Prof. H. Joachim Deeg (Fred Hutchinson Cancer Center; Seattle, WA, USA). SKM-1, a cell line established from MDS progressing to AML, was donated by Prof. Xiao Li (Shanghai Jiao Tong University). These cells were all cultured as described previously [15].
Generation of conditional MGAT3 loxP/loxP mice MGAT3 fl/fl transgenic mice were generated using a homologous combination knockout strategy, in which exon 2 of MAGT3 gene was flanked with two loxP sites. Conditional deletion of MGAT3 gene was accomplished by crossbreeding leptin receptor (LepR)-Cre transgenic mice (donated by Prof. Caiwen Duan, Shanghai Jiao Tong University) with MGAT3 fl/fl mice. Animal experiments were approved by the Animal Care and Use Committee of Northwest University.

Levels of bisecting GlcNAc and MGAT3 in BM stroma
Expression of bisecting GlcNAc and its glycosyltransferase MGAT3 were examined in primary MDS/AML BM stroma (Fig. 1A). MGAT3 mRNA expression and bisecting GlcNAc levels were significantly lower for MDS/AML patients than HD (Fig. 1B). And bisecting GlcNAc in stromal cells was significantly downregulated by coculture with KG1a or SKM-1 cells (Figs. 1C, S1A-C). These findings demonstrate the aberrant expression of bisecting GlcNAc in MDS/ AML stromal cells, and the ability of myeloid cells to alter N-glycosylation levels of niche cells.
We generated LepR-Cre; MGAT3 fl/fl mice in order to investigate the functional role of bisecting GlcNAc in the BM niche in vivo (Fig.  S3A). These mice showed strongly reduced expression of MGAT3 and stromal bisecting GlcNAc (Fig. S3B), retarded growth relative to MGAT3 fl/fl (Fig. S3C), and development of severe anemia and thrombocytopenia (Fig. S3D). KG1a/SKM-1 cells grew well when transplanted into LepR-Cre; MGAT3 fl/fl (Figs. 2C, S3E). These findings illustrate substantial effects on the hematopoietic microenvironment of MGAT3 expression and bisecting GlcNAc production in BM stroma.
We hypothesized, in view of these findings, that MCAM expression is affected by stromal bisecting GlcNAc level. MCAM expression was upregulated in HS5-shM3 but downregulated in HS27a-M3, consistently with this hypothesis (Fig. S5E). We cloned MGAT3 gene into a tetracycline-inducible gene expression system and transfected it into HS27a. Treatment of these cells with doxycycline (dox) resulted in time-dependent increased MGAT3 and bisecting GlcNAc expression, and decreased MCAM expression (Fig. S5F). Introduction or silencing of MGAT3 in primary stroma led respectively to down-and upregulation of MCAM expression (Fig. 4B). These findings demonstrate that MCAM expression in stromal cells was regulated by bisecting GlcNAc modification.
Increase of bisecting GlcNAc levels did not alter MCAM expression at the mRNA level (Fig. S6A), suggesting that modulation of MCAM expression by bisecting GlcNAc modification is a post-translational event. MCAM is a typical transmembrane glycoprotein, and we accordingly used Sulfo-NHS-LC-Biotin to label MCAM on cell membrane (Fig. S6B). Total MCAM content was much lower in HS27a-M3 than in HS27a cells. In HS27a-M3, relative to HS27a, biotin-labeled MCAM level on cell membrane was lower, whereas MCAM level in cytoplasm was higher (Fig.  S6C). Blocking of cytosolic protein synthesis by cycloheximide accelerated MCAM degradation (Fig. S6D). In most cases, intracellular proteins are degraded via ubiquitin-proteasome pathways, whereas extracellular proteins and cell surface proteins enter cells by endocytosis and are degraded via lysosomal pathways [20][21][22]. MCAM expression in HS27a-M3 was enhanced by treatment with lysosomal inhibitor chloroquine, but unaffected by treatment with proteasome inhibitor MG132 (Fig. S6E). MCAM in HS27a-M3 was shown to be localized mainly in lysosomes (Fig.  S6F). These findings indicate that bisecting GlcNAc modification affects MCAM stability, and causes MCAM degradation via a lysosomal pathway.
Mutation at Asn 56 had no effect on MCAM expression at protein or mRNA levels (Figs. 4C, S7B). However, MCAM level on cell membrane was significantly lower in HS5-MCAM-Mu than in HS5-MCAM (Fig. S7C). Similarly, when MCAM was pulled down by Sulfo-NHS-LC-Biotin strategy as described above, MCAM level on cell membrane was significantly lower in HS5-MCAM-Mu than in HS5-MCAM (Fig. 4C). For the two cell types, MCAM expression was assayed in total cell lysates, on cell membrane, and in cytoplasm. In HS5-MCAM-Mu, relative to HS5-MCAM, MCAM level was lower on cell membrane, and higher in cytoplasm (Fig. S7D). For HS5-MCAM-Mu, cycloheximide treatment to block novel cytosolic protein synthesis accelerated MCAM degradation (Fig. S7E), while MCAM expression was enhanced by chloroquine treatment and unaffected by MG132 (Fig. S7F). MCAM in HS5-MCAM-Mu was shown to be localized in lysosomes (Fig. S7G). Thus, localization of MCAM on cell membrane could be reduced by either removal of N-glycan structures at Asn 56 by mutation, or enhanced bisecting GlcNAc modification at this site.

Effect of MCAM on myeloid cell proliferation
In vivo and in vitro experiments showed that proliferation of myeloid cells co-cultured with HS5-MCAM increased, whereas that of cells co-cultured with HS5-MCAM-Mu declined to the level of cells co-cultured with HS5 (Figs. 5A, S8A, B). MCAM on cell membrane evidently supports myeloid cell proliferation.
Based on the above finding, transplantation experiments were performed to clarify the in vivo role of bisecting GlcNAc modification of MCAM in myeloid cell proliferation (Fig. S8C). Proportion of KG1a/SKM-1 cells in peripheral blood of NSG mice was significantly higher when KG1a/SKM-1 were co-injected with HS5-MCAM cells, relative to co-injection with HS5 or HS5-MCAM-Mu (Figs. 5B, S8D, E). Expression in BM and spleen of CD45 + signal, a myeloid cell marker, was significantly lower for C Lectin blotting analysis of MGAT3 and bisecting GlcNAc levels of HS5, HS27a, and NBM stroma treated with miR188-5p mimic. HEK293T cells were co-transfected with miR188-5p mimic or miR188-5p mimics negative control ("miNC") and two reporter plasmids psiCHECK2 (wild-type or mutant MGAT3 3′-UTR sequence), and luciferase activities of transfectant cells were assayed.
HS5-MCAM-Mu co-injected relative to HS5-MCAM co-injected group (Figs. 5C, S8F-H). These findings indicate that the supportive effect of stromal cells on myeloid cell proliferation was disrupted in glycosylation-deficient N56D mutant of stroma because of reduced MCAM expression on stromal cell membrane.
These findings, taken together, demonstrate that MCAM on stromal cells binds to CD13 on myeloid cells, activates ERK signaling, and promotes myeloid cell growth.

DISCUSSION
Abnormal hematopoiesis may result from genetic dysregulation, or from a dysfunctional structure acting by itself or affecting crosstalk with hematopoietic stem cells [25]. Many factors are involved in crosstalk between stroma and hematopoietic cells; among these, glycoconjugates play key roles in modulating functions of HSPCs in BM [26][27][28]. Glycoconjugates on cell membranes affect nearly all interactions between cells and their surrounding environment. Several glycoengineering strategies have been developed to improve homing and engraftment of cells following hematopoietic stem cell transplantation [29][30][31]. In this study, we observed downregulated levels of bisecting GlcNAc and MGAT3 in BM stroma of MDS/AML patients. Bisecting GlcNAcylation is a specific type of N-glycosylation that affects adhesion, migration, and other cellular functions by modifying adhesion molecules and receptors (notably E-cadherin, integrins, tetraspanins, and EGFR) [14,32,33]. Bisecting GlcNAc is also involved in organ growth and development [34]. Low bisecting GlcNAc levels in stromal cells promote proliferation of hematopoietic clonal cells. Malignant clonal cells packaged miR-188-5p into exosomes for delivery to recipient stromal cells, where it reduced MGAT3 expression and bisecting GlcNAc level. These and similar findings demonstrate the ability of malignant cells to reprogram their microenvironment [35].
We identified MCAM as a bisecting GlcNAc-bearing target protein in stromal cells. MCAM was originally identified in human melanoma as an adhesion molecule glycoprotein [36]. It was subsequently shown to be highly expressed in other tumors and in endothelial cells. MCAM plays functional roles in a variety of cellular processes, including transendothelial migration, proliferation, and cancer metastasis [37][38][39]. The previous findings show that MCAM is a useful marker associated with BM niche, together with LepR [40][41][42]. MCAM overexpression in mesenchymal stromal cells (MSCs) enhanced adhesion of HSPCs to MSCs, and supported HSPC growth. In contrast, silencing of MCAM in MSCs suppressed HSPC proliferation, and strongly reduced formation of long-term culture-initiating cells [43]. Our 2013 study indicated that high MCAM expression in stromal cells facilitated engraftment of cloned MDS patient cells in a mouse xenotransplantation model [10]. Consistently with the finding by P. Bianco's group of high MCAM expression in MDS/AML patients [44], we observed that high MCAM expression in stromal cells supported myeloid cell proliferation in vitro and in vivo.  N-glycosylation plays essential roles in protein folding and trafficking, and whole N-glycosylation knockdown or mutation at certain N-glycosylation sites resulted in misfolding, mis-distribution, instability, and/or degradation of such glycoproteins as glucose transporter GLUT4, human tripeptidyl-peptidase I, and dopamine transporter [45][46][47]. When bisecting GlcNAc level was high, MCAM content in membranes was significantly decreased and localization of MCAM in lysosomes was increased. Bisecting GlcNAc appeared to modulate MCAM translocation to cell membrane, and to induce degradation of MCAM via lysosomal pathways. IP-MS analysis identified Asn 56 on MCAM as the key bisecting GlcNAc-bearing site. Promoting effect on malignant clonal cell proliferation in vitro and in vivo was weaker for glycosylation-deficient N56D mutant of HS5 (HS5-MCAM-Mu) than for HS5-MCAM. Knockdown of N-glycosylation at Asn 56 reduced MCAM localization on cell membrane, indicating that such localization depends on other N-glycan structures not yet identified.
In regard to functional mechanism, we found that MCAM on stromal cell membrane supported malignant clonal cell growth through interaction with the transmembrane aminopeptidase CD13. CD13 is widely expressed in myeloid cells, stromal cells and other cells [48,49]. It is overexpressed in many tumor cells and plays a key role in tumor angiogenesis, invasion, and metastasis, and also has been associated with multidrug resistance [50]. CD13 + myeloid BM-derived cells (BMDCs) promote angiogenesis and vascular maturation by regulating pericyte recruitment through increased production of MCP-1 and MMP-9 [51]. It was documented that specific CAR-T cells have been developed to target CD13 to eradicate AML cells [49]. We observed that elimination of bisecting GlcNAc on MCAM by Asn 56 mutation, or by bestatin treatment, notably reduced CD13 binding, inactivated ERK signaling, and suppressed myeloid cell growth. The supportive function of stroma on clonal cell growth therefore appears to depend on interaction between MCAM on stromal cells and CD13 on clonal cells.
In conclusion, MDS/AML clonal cells can modify bisecting GlcNAc levels of BM stroma through exosome secretion, thereby promoting malignant clonal cell proliferation and survival, and suppressing normal hematopoiesis (Fig. 6D). Novel therapeutic strategies that restore a healthy BM niche, based on glycosylation modification, may show improved effectiveness against MDS/AML and similar disorders.

DATA AVAILABILITY
The data supporting the conclusions of this article have been given in this article and its additional files.