Dysregulation of developmental and cell type-specific expression of glycoconjugates on hematopoietic cells: a new characteristic of myelodysplastic neoplasms (MDS)

Myelodysplastic neoplasms (MDS) are age-associated hematopoietic neoplasms characterized by myeloid dysplasia and cytopenias. Patients with MDS have a diverse clinical course, ranging from indolent conditions to acute myeloid leukemia (AML) [1]. Sequencing of leukocytes from MDS patients revealed somatic mutations that correlated with their clinical outcome. Studies that traced driver mutations back to hematopoietic stem and progenitor cells (HSPCs) supported the concept that myelodysplastic phenotypes arise from cancer stem cells [2]. Importantly, HSPC function and interaction with the bone marrow (BM) microenvironment depends partly on glycan- protein interactions [3]. Glycosylation is the post-translational modi ﬁ cation by which oligosaccharide chains are covalently attached to amino acids or lipids. Glycoproteins are decorated with glycans at nitrogen- and oxygen atoms in the endoplasmic reticulum or Golgi apparatus, yielding N - and O -linked glycosylation, respectively. Aberrant glycosylation is a hallmark of oncogenesis and results in modulated in ﬂ ammatory responses, apoptosis and cancer cell metastasis [4]. Insights into aberrant glycome

TO THE EDITOR Myelodysplastic neoplasms (MDS) are age-associated hematopoietic neoplasms characterized by myeloid dysplasia and cytopenias. Patients with MDS have a diverse clinical course, ranging from indolent conditions to acute myeloid leukemia (AML) [1]. Sequencing of leukocytes from MDS patients revealed somatic mutations that correlated with their clinical outcome. Studies that traced driver mutations back to hematopoietic stem and progenitor cells (HSPCs) supported the concept that myelodysplastic phenotypes arise from cancer stem cells [2]. Importantly, HSPC function and interaction with the bone marrow (BM) microenvironment depends partly on glycanprotein interactions [3]. Glycosylation is the post-translational modification by which oligosaccharide chains are covalently attached to amino acids or lipids. Glycoproteins are decorated with glycans at nitrogen-and oxygen atoms in the endoplasmic reticulum or Golgi apparatus, yielding N-and O-linked glycosylation, respectively. Aberrant glycosylation is a hallmark of oncogenesis and results in modulated inflammatory responses, apoptosis and cancer cell metastasis [4]. Insights into aberrant glycome structures have been applied for the development of biomarkers and therapeutic antibodies. Although described in other hematological malignancies, glycosylation is understudied in MDS [5,6]. This study explored glycosignatures in MDS and AML to elucidate pathological mechanisms that could serve as biomarker.
This study was conducted following the Helsinki Declaration and approved by the Medical Ethics Committee of the Amsterdam UMC location Vrije Universiteit Amsterdam (VUmc 2014-100, VUmc 2019-3448). Samples were obtained from patients with MDS (n = 14, Table S1), AML (n = 9) and iron deficiency and dysregulated iron metabolism (IDef, n = 17) (Supplementary Information). Normal bone marrow (NBM, n = 10) was acquired from cardiothoracic surgery patients after written informed consent. We used plant lectins as probes to recognize glycoconjugates based on their specific glycanbinding affinities. Cells were stained with an antibody backbone and one of the following lectins: Phytohemagglutinin-L (PHA-L), Concanavalin A (ConA), Maackia amurensis agglutinin II (MAA-II), Maackia amurensis leucoagglutinin I (MAL-I) and Sambucus nigra agglutinin (SNA, Table S2). The lectins PHA-L, ConA and SNA recognize tetra-antennary N-glycans, high-mannose glycans and di-antennary N-glycans, and α2-6 sialoglycans, respectively. The MAA-II and MAL-I lectins bind to α2-3 sialoglycans with distinct carbohydrate binding specificities: MAA-II has a preference for O-linked α2-3 sialic acids and MAL-I for N-linked α2-3 sialic acids. Lectins were selected based on their binding to hematopoietic cells as demonstrated in a pilot study (data not shown). Flow cytometry data were manually pre-gated on CD45 + leukocytes, aggregated into a dataset of 40 • 10 6 cells and subjected to unsupervised clustering ( Supplementary Information,  Fig. S1). Statistics are described in the (Supplementary Information  Tables S3-S5).

AML AND MDS
To explore glycosylation in myeloid disorders, we applied a principal component analysis on the lectin-binding intensities of the populations for each sample ( Fig. 2A). This analysis discriminated AML from other samples, indicating that AML-BM is characterized by a unique glycosignature. Whereas IDef showed overlap with NBM and MDS, MDS partly grouped together in-between NBM and AML. In search of aberrant glycoprofiles that separate MDS from other samples, we compared lectin-binding intensities on populations between diagnoses (Table S5A-C). This revealed aberrant glycan expression on hematopoietic cells across distinct maturational stages and cell lineages in MDS and AML (Fig. 2B). A common feature of MDS and AML was decreased α2-3 N-linked and α2-6 sialylation of GMPs [9] and aberrant expression of high-mannose glycans and/or di-antennary N-glycans, with increased expression on lymphoid populations [16,30,31] and reduced expression on myeloid subsets, particularly immature granulocytes [5][6][7] (Fig. 2C).
This study used unsupervised clustering to explore glycoprofiles in MDS and AML (Fig. S4). Conform a previous study on cord blood, we identified cell type-and maturational stage-specific glycosignatures in NBM [9]. Furthermore, we revealed that altered glycosylation is already detectable at HSCs in AML and MDS. Whereas α2-3 N-linked sialylation was increased on MDS-derived HSCs, it was reduced on HSCs in AML. The α2-3 N-linked sialic acids have been described to impair CD44-mediated binding to the extracellular matrix, thereby supporting HSPC migration and potentially hampering hematopoietic differentiation [10]. Upregulated α2-3 sialyation in breast cancer affected cell migration by supporting metastatic spread [11]. Contrarily, downregulated α2-3 sialyation was demonstrated in colorectal cancer, indicating that altered glycan expression depends on the tumor type [12]. In line with this hypothesis, a study on leukemic cell lines showed enhanced α2-3 sialylation within erythroid leukemia (M6) and reduced expression in myeloid (PLB985) and promyelocytic leukemia (HL60) [5]. We observed increased tetra-antennary Nglycosylation on most hematopoietic populations in MDS and on CLPs in AML. Tetra-antennary N-glycosylation has been linked to the suppressive potency of regulatory T-cells [13]. This suggests that increased tetra-antennary N-glycosylation on lymphoid subsets may play a role in tumor surveillance in MDS and AML, whereas general overexpression throughout hematopoiesis may  (Table S2). The background color of the cell clusters (n = 90) indicates their population (n = 32). The height of the plot pie visualizes the expression of the surface markers and the scatter properties. For visualization purpose, subsets of the minimal spanning tree are encircled (dotted line) with manual labels. B FlowSOM minimal spanning tree colored by the scatter intensities, antigen expressions and lectin binding intensities as a marker for glycan expressions for healthy controls. I. The scatter intensities and antigen expressions were used as input for cell clustering. II. The lectin binding intensities were projected on the FlowSOM minimal spanning tree. C Heatmap summary of protein expressions, scatter properties and lectin binding intensities for each of the 32 populations derived from all samples. D Table summary of the 32 populations as identified by FlowSOM. The populations were manually assigned to cell subsets based on biaxial dotplots, immunophenotypic criteria and the metric MEM. MO monocytes, pDC plasmacytoid dendritic cell, MEP megakaryocyte erythroid progenitor, HSC/CMP hematopoietic stem cell/common myeloid progenitor, LSC leukemic stem cell, GMP granulocyte macrophage progenitor, ERY erythrocyte, MGK megakaryocyte, PC plasma cell, LYM lymphocyte, CLP common lymphoid progenitor, GRN granulocyte, IM immature, EO eosinophil, PRO progenitor, MEM marker enrichment modeling.
be related to myelodysplastic phenotypes. Compared to low-risk, high-risk MDS showed upregulated α2-6 sialyation at HSC level. Differently, we observed increased α2-6 sialyation on AML-derived lymphoid cells. Previous literature showed that overexpression of STGAL1, the gene encoding for α2,6-sialyltransferase, facilitates progression of prostate cancer [14]. Another study showed higher α2-6 sialyation on tolerogenic DCs that is downregulated after DC maturation with proinflammatory cytokines [15]. We hypothesize that enhanced α2-6 sialylation on lymphoid cells may induce tumor surveillance.
To conclude, this study suggests that increased tetra-antennary N-glycosylation contributes to myelodysplastic phenotypes, whereas decreased α2-3 N-linked sialyation and increased α2-6 sialyation characterizes AML. Upregulation of α2-6 sialic acids and high-mannose glycans and/or di-antennary N-glycans on HSCs differed high-risk from low-risk MDS, indicating that glycoprofiles could be of value for MDS risk stratification. However, the main disadvantage of this exploratory study is the low sample size and power. Larger studies that combine profiling of antigen and lectin intensities with mass spectrometry and sequencing of glycosyltransferases are warranted.

DATA AVAILABILITY
The data that support the findings of this study are available from the corresponding author, A. A. van de Loosdrecht, on reasonable request.  [1,2,10] and aberrant CD34progenitors [12]. The first heatmap (gray background) summarizes the lectin binding intensities on NBM-derived populations. The other heatmaps present the difference in lectin-binding intensities in patients samples expressed by the fold change defined as (Y-X)/X using the lectin binding in NBM and patient diagnoses as X and Y, respectively. C Boxplots illustrating the median and range of glycan expression on distinct hematopoietic populations across diagnoses. Note that only a selection of the differentially expressed glycans is shown as an example. D Heatmap summary presenting lectin binding intensities from NBM-derived HSC/CMPs [3] as compared to aberrant stem cell populations, including AML-derived LSCs [1,2,10] and IDef-, MDS-and AML-derived HSC/CMPs [3]. E Boxplot summary of differentially expressed SNA-bound α2-6 sialoglycans, ConA-bound high-mannose glycans and/or di-antennary N-glycans and MAL-I-bound α2-3 N-linked sialoglycans between low risk (LR, good risk cytogenetics) and high risk (HR, intermediate or poor risk cytogenetics) MDS. The P values are based on the Mann-Whitney U test. MO monocytes, pDC plasmacytoid dendritic cell, MEP megakaryocyte erythroid progenitor, HSC/CMP hematopoietic stem cell/common myeloid progenitor, LSC leukemic stem cell, GMP granulocyte macrophage progenitor, ERY erythrocyte, MGK megakaryocyte, PLASMA plasma cell, LYM lymphocyte, CLP common lymphoid progenitor, GRN granulocyte, IM immature, EO eosinophil, PRO progenitor.