E-Cadherin Expression Distinguishes Mouse from Human Hematopoiesis in the Basophil and Erythroid Lineages

E-cadherin is a key regulator of epithelial cell–cell adhesion, the loss of which accelerates tumor growth and invasion. E-cadherin is also expressed in hematopoietic cells as well as epithelia. The function of hematopoietic E-cadherin is, however, mostly elusive. In this study, we explored the validity of mouse models to functionally investigate the role of hematopoietic E-cadherin in human hematopoiesis. We generated a hematopoietic-specific E-cadherin knockout mouse model. In mice, hematopoietic E-cadherin is predominantly expressed within the basophil lineage, the expression of which is dispensable for the generation of basophils. However, neither E-cadherin mRNA nor protein were detected in human basophils. In contrast, human hematopoietic E-cadherin marks the erythroid lineage. E-cadherin expression in hematopoiesis thereby revealed striking evolutionary differences between the basophil and erythroid cell lineage in humans and mice. This is remarkable as E-cadherin expression in epithelia is highly conserved among vertebrates including humans and mice. Our study therefore revealed that the mouse does not represent a suitable model to study the function of E-cadherin in human hematopoiesis and an alternative means to study the role of E-cadherin in human erythropoiesis needs to be developed.


Introduction
E-cadherin is a transmembrane protein and a master regulator of cell adherens junctions, through which it controls cell-cell adhesion. Epithelial cell-cell adhesion controls organ function, prevents tissue infection and suppresses malignancy formation. E-cadherin expression, however, can be detected in a wide variety of cells including hematopoietic progenitor cells (HPCs). Adherence junctions are formed by multimeric homophilic Ecadherin protein interactions at the interface between neighboring cells [1,2]. Intracellularly, E-cadherin binds signaling molecules αand β-catenin which connect the actomyosin cytoskeleton with the site of adhesion [3]. Cell adherens junctions thereby act as molecular bridges that are able to transmit mechanical forces generated by neighboring cells (e.g., by cell migration, proliferation or apoptosis [4][5][6][7]. Mechanical activation of E-cadherin/βcatenin signaling accordingly promotes cell proliferation, which underlies the regulation of E-cadherin in tissue homeostasis [6,7]. The importance of E-cadherin signaling during epithelial development is signified in E-cadherin-null mutant mice, which are embryonic lethal due to failure of trophectoderm epithelium formation [8]. E-cadherin is also a well-established tumor suppressor as demonstrated in mouse models with tissue-specific conditional inactivation of E-cadherin, which revealed impaired intestinal, skin and mammary gland formation and promoted tumor development and metastasis [9][10][11][12][13]. Moreover, CDH1 (encoding E-cadherin) inactivating mutations or promoter hypermethylation in humans are associated with the development of gastric, prostate, hepatocellular, breast

Human PBMC Isolation
Human material was obtained after informed consent was given in accordance with the Declaration of Helsinki, the Dutch National and Sanquin Internal Ethic Boards. BM was aspirated from patients undergoing cardiac surgery (permit MEC 04/042, No. 04.17.370; AMC, Amsterdam, The Netherlands). Human peripheral blood mononuclear cells (PBMCs) were purified by density separation using Ficoll (GEHealthcare, 17-1440-03), following the manufacturer's protocol. For hematopoietic stem and progenitor cell (HSPC) isolation, mobilized peripheral blood (MPB) was obtained by leukapheresis. HSPCs were isolated using CD34+ MACS beads (Miltenyi, 130-046-703) according to the manufacturer's protocol and cryopreserved in IMDM, 20% FCS; 5% DMSO as previously described [23].
Mouse tissue preparation mouse BM cells were obtained by crushing the femur and tibia, and a 70 µM cell strainer was used to filter the suspension. Spleen cells were directly mashed through a 70 µM cell strainer. Consequently, the cell strainer was flushed with PBS containing 1% FCS. Cells were kept at 4 • C at all times. Blood was collected through heart puncture in a heparin tube. Whole blood was incubated with erythrocyte lysis buffer (0.155 M NH 4 Cl; 10 mM KHCO 3 ; 1 mM EDTA; pH 7.2-7.6) before antibody staining for flow cytometry. Lung tissue was cut into pieces and enzymatically digested for 30 min at 37 • C with 750 U/mL Collagenase Type I (Worthington, LS004194) and 0.31 mg/mL DNase I (Roche, 10104159001). Single cells were separated over a Percoll (GE Healthcare, 17-0891-01) density gradient following the manufacturer's protocol before proceeding with hematopoietic fraction for flow cytometry. Peritoneal cells were obtained by flushing the peritoneal cavity with PBS. Blood values were measured using a Scil Vet abc Plus machine (Horiba Medical, Montpellier, France) using the pre-installed mouse research program.

LCMV Infections
In order to induce a representative multilineage challenge including the red cell lineage, mice were infected intraperitoneally with 1 × 105 plaque-forming units (PFU) of LCMV Armstrong obtained from EVAg, European Virus Archive. After 8 days, the mice were sacrificed and their organs were processed according to the methods above.

Immunohistochemistry
Mouse BM was formalin-fixed in 10% neutral buffered formalin for 48 h, embedded in paraffin, sectioned and stained with hematoxylin and eosin (H&E). Immunohistochemistry (IHC) was performed as previously described [10]. All slides were digitally processed using the Aperio ScanScope (Aperio, Vista, CA, USA) and captured using ImageScope software version 12.0.0 (Aperio).

Flow Cytometry
Cells were incubated with antibodies for 30 min at 4 • C as listed in Supplementary Table S1 (for mouse tissue) and Table S2 (for human tissue). When specified, human PBMCs were incubated with 100 µg/mL Nanogam, broad spectrum human normal immunoglobulin (Prothya Biosolutions, Amsterdam, The Netherlands) for FcR-blocking as control for isotype staining. Dead cells were excluded by using the fixable near-IR live/dead staining kit (Thermo Fisher Scientific (Waltham, MA, USA), L10119). A Foxp3/Transcription Factor Staining Buffer Set (eBioscience (San Diego, CA, USA), 00-5523-00) was used according to the manufacturer's specifications for intracellular staining. Flow cytometry was performed on an FACS Symphony (BD Biosciences, Haryana, India), and data was analyzed using FlowJo V10 (Tree star) software. Cell sorting was performed on an FACS Aria III (BD Biosciences), and on-site analysis was performed using BD FACSDiva 8.0.1. software.

Cytospins
Mouse E-cadherin + ;IgE + cells were spun down at 800 rpm on an object glass using a Shandon Cytospin II (Thermo Scientific), air dried and fixed in methanol for 5 min. Samples were stained with May-Grünwalds eosine-methylene blue solution (Merck, Rahway, NJ, USA) for 10 min and Giemsa dilution for 20 min.

Single Cell RNA Sequencing
E-cadherin + ;cKIT + and Ecadherin + ;cKIT − cells were sorted and used at a 2x single-cell suspension at final conc. 1000 cells/µL in 1 × PBS containing 0.04% weight/volume BSA (400 µg/mL). Cells were loaded into a 10X Genomics Chromium Controller for capture and barcoding following the 10X Genomics Protocol (Chromium Single Cell 3 Reagent Kits v3, CG000183 Rev A), as described by the manufacturer (10X Genomics). See supplementary information for more details.
The molarity of the Single Cell 3 Gene Expression libraries were quantified on a 2100 Bioanalyzer Instrument following the Agilent Technologies Protocol (Agilent DNA 7500 kit, G2938-90024), as described by the manufacturer (Agilent Technologies, Santa Clara, CA, USA). The Single Cell 3 Gene Expression libraries were diluted to 10 nM and pooled to create one Illumina sequencing library pool. The molarity of the Illumina sequencing library pool was quantified by qPCR following the Roche Protocol (KAPA Library Quantification Kit Illumina ® Platforms, KR0405), as described by the manufacturer. Single-cell RNA sequencing was performed on the Illumina Nextseq 550 system (Illumina). The samples were sequenced on a NextSeq High with a base composition of 28-8-56. Sample 1 had a total of 265 million reads and sample 2 had a total of 72 million reads. Downstream analysis was performed using Cellranger version 3.0.2 with mm10 as a reference genome. In total, 6632 and 3745 cells were picked up for sample 1 and 2 with a median of 3.3 K and 1.7 K detected genes per cell, an average of 40 K and 19 K reads per cell, and median UMI counts per cell of 13 K and 4.5 K, respectively. The output data of Cellranger has been used for further in-depth analysis. In-depth analysis was performed within R using the Seurat package (version 3).
Genes that were expressed in less than 40 cells were removed from the individual datasets which resulted in 6632 and 3745 cells. Individual datasets have also been filtered for the highest and lowest 1% number of genes and highest 1% number of RNA counts in order to remove outliers from the dataset. A total of 12,395 and 10,412 genes were detected in these samples. After these filtering steps, the datasets were merged, which resulted in a total of 10,141 (6279 and 3662) cells and 12,462 genes. Cells have not been filtered for mitochondrial expression because the expression levels of mitochondrial genes were very low. Normalization, scaling and PCA dimensionality reduction of the data have been performed using standard Seurat parameters. To identify clusters, the TSNE algorithm for dimensionality reduction has been used. t-SNE clusters were automatically detected by Seurat using the FindNeighors/FindClusters functions with 15 dimensions and the run t-SNE function with a resolution of 1.

Statistics
GraphPad Prism 6 software was used to apply a standard two-sided Student's t-test (paired) for pairwise comparisons. p < 0.05 values were considered statistically significant.

E-cadherin Is Predominantly Expressed by Basophil/Mast Cell Precursors in Mouse BM
To define the role of E-cadherin in the HSPC lineage of mice, we generated Vav-iCre;Cdh1 F/F :mTmG mice by crossing the previously described Vav-iCre transgenic mice with Cdh1 F/F ;mTmG mice ( Figure 1A) [10,24]. Vav-iCre was used to induce somatic inactivation of floxed Cdh1 expression within the hematopoietic cell lineage. Expression of Cre can accordingly be defined by monitoring GFP fluorescence, which is induced upon Cre-driven recombination of the inserted mTmG reporter gene [24,25]. Vav-iCre;mTmG mice were developed as control animals and used to characterize E-cadherin protein expression in the hematopoietic cell lineage.
As E-cadherin predominantly localizes at the plasma membrane, we used flow cytometry to examine the distribution of E-cadherin within the hematopoietic cell lineage in BM of Vav-iCre;mTmG mice. Hematopoietic Cre recombinase activity was verified by correlation of GFP-expression with CD45 (Supplementary Figure S1E). E-cadherin expression was detected in~0.6% of GFP + hematopoietic cells, compared to a background signal of~0.1% in the Vavi-Cre;Cdh1 F/F ;mTmG and Vavi-Cre;;mTmG mouse measured with isotype control antibody ( Figure 1B-D, Supplementary Figure S1A-D). Hereby we confirm the specificity of E-cadherin detection and suggest successful KO of Cdh1 expression ( Figure 1C,D). Remarkably, as we included CD31 (encoding PECAM-1) staining to define the endothelial compartment (CD45 − CD31 + ), we noticed that E-cadherin + hematopoietic cells could be distinguished by high expression of CD31 within CD45 + marked hematopoietic compartment ( Figure 1E). In line with the observed accumulation of Cdh1 mRNA within the basophil/mast cell precursor lineage in mouse BM, approximately 60-80% of E-cadherin + hematopoietic cells reflected a basophil/mast cell signature, as defined by cell-surface protein expression of IgE and FcεR ( Figure 1F) [21]. E-cadherin thereby marks the entire population of basophil and mast precursors (CD45 + IgE + CD45 -KIT -) in mouse BM. The low percentage of E-cadherin + hematopoietic cells in mouse BM, as detected by flow cytometry, was consistent with the sporadic detection of E-cadherin-expressing cells in mouse BM sections by immunohistochemistry (IHC; Figure 1G). Additionally, Ecadherin is not coexpressed with lymphoid (CD3, CD4, CD8, B220), myeloid (Gr-1, CD11b) or erythroid (CD71, Ter119) lineage (LIN) markers ( Figure 1G Cytospins of E-cadherin + IgE + mouse BM cells show, as previously described for mouse BM and blood-derived basophils, polymorphic nuclei and dark granules, confirming their basophilic character ( Figure 1H) [26]. The remaining fraction of E-cadherin + cells (~20% corresponding to 0.1% of total hematopoietic BM cells), however, did not reflect a basophildefined cell-surface signature but expressed intermediate-to high-cell-surface expression of cKIT ( Figure 1K). These cells comprise a fraction of progenitor cell-enriched LIN -KIT + (LK) population and therefore may have progenitor potential ( Figure 1L). Notably, the low number of detected E-cadherin + hematopoietic cells in mouse BM is also in line with the predominant detection of Cdh1 mRNA expression in the rare population of basophil/mast cell progenitors in mouse BM. Taken together, these results confirm that E-cadherin is predominantly expressed by the basophil/mast cell lineage in mouse BM. Additionally, we found that E-cadherin may also be expressed by more immature HPCs. E-cadherin expressing HPCs in mice are not fully committed to the basophil/mast cell lineage.
To define the relationship between E-cadherin + cKIT + hematopoietic cells and E-cadherin + cKIT − defined basophil/mast cell precursors, we sorted both populations from the BM of wild-type mice and performed single-cell (sc) RNA-seq analysis (Supplementary Figure S2A). The obtained two-dimensional representation revealed that Ecadherin + basophil/mast precursors clustered as one major branch, which separates from the E-cadherin + cKIT + cellular compartment (Figure 2A). A large fraction of E-cadherin + cKIT + cells were found to coexpress Mcpt8 and Prss34 ( Figure 2B,C). Prss34 is predominantly expressed by basophils and at low levels by mast cells in the peritoneum [27]. Additionally, Mcpt8 is used to distinguish the basophil lineage, as it is not expressed by mast cells in mice, although its expression is also detected at low levels in granulocyte/macrophage progenitors (GMPs) [27][28][29][30][31]. Altogether, these findings indicate that the majority of E-cadherin + cKIT + hematopoietic cells preferentially act as precursor cells of the basophil/mast cell lineage.
However, the presence of two additional smaller branches suggests that E-cadherin + cKIT + cells are not fully committed to the basophil lineage, as these branches express erythroid (Car1, Hba-a1, Gypa, and Tfrc Figure 2D-F, Supplementary Figure S2F) and monocyte (Elane) cell lineage markers ( Figure 2G). Additionally, we identified a cell cluster with high-cKIT expression that lacked distinguished lineage markers (cluster #6) and therefore may resemble non-committed early HPCs ( Figure 2I).

E-cadherin Is Dispensable for Basophil Formation in Mice
Since E-cadherin mostly marks basophil precursors in mouse BM, we next defined whether E-cadherin expression is required for the generation of basophils. We therefore measured the percentage of basophil precursors in BM hematopoietic cells as well as matured basophils in circulation and peripheral organs such as the spleen and lungs in VaviCre;Cdh1 F/F ;mTmG mice by flow cytometry (Figure 3A,B). Basophils (IgE + B220 -) from these organs were confirmed to be negative for cKIT cell-surface expression, which discriminates basophils from mast cells (Supplementary Figure S3A-D). Basophils in VaviCre;Cdh1 F/F ;mTmG mice completely lacked cell-surface expression of E-cadherin, suggesting a complete KO of E-cadherin in the basophil lineage, as compared to control mice ( Figure 3C-F). However, basophil precursor numbers in BM (~0.4%) as well as the generation of maturated basophils in the spleen (~0.1%), blood (~0.3%) and lungs (~0.3%) of VaviCre;Cdh1 F/F ;mTmG mice were not hampered as compared to control animals ( Figure 3B). E-cadherin is therefore dispensable for the formation and circulation of basophil/mast cell precursors in mice. Furthermore, white blood cell counts, and the percentages of lymphocytes, monocytes and granulocytes did not differ between control VaviCre;mTmG and VaviCre;Cdh1 F/F ;mTmG mice ( Figure 3G-J).
As a small fraction of E-cadherin + cKIT + cells displayed expression of erythroid transcription markers ( Figure 2D-F), as well as erythroid progenitor potential in colony formation assays ( Figure 2J-L), we additionally examined erythropoiesis in VaviCre;Cdh1 F/F ;mTmG mice to address the control of E-cadherin over mouse erythropoiesis. Loss of hematopoietic E-cadherin, however, does not alter (CD71 + Ter119 + ) erythroblasts percentage in BM (Supplementary Figure S3G). In mice, the spleen is an important site for extramedullary regulation of erythropoiesis; erythropoiesis in the spleen was also examined. However, hematopoietic-specific loss of E-cadherin did not alter erythroblast percentage in the spleen (Supplementary Figure S3H) [32]. Additionally, as also specified in Figure 1I, there was no E-cadherin cell-surface expression detected in the erythroid lineage in BM or the spleen (Supplementary Figure S3I,J).
As well as steady-state conditions, we investigated the erythroid response to anemia, stress erythropoiesis. We induced anemia by LCMV infection, causing hemolysis and/or INF-γ mediated inflammation [33,34].

Cdh1 Expression Distinguishes Human from Mouse Hematopoiesis
We here identified that E-cadherin mostly marks the mast/basophil lineage in mice. In contrast, E-cadherin has been reported to mainly mark erythroblasts in human BM [21,22]. This poses the question of to what extent E-cadherin expression is conserved between the hematopoietic lineages of mice and humans. We data mined two previously published scRNA seq datasets to compare the distribution of E-cadherin expression in HSPCs between mice and humans [35,36]. For extensive methodology comparison between mouse and human hematopoiesis by scRNA seq on HSPC populations we refer to Pellin et al. [36]. In both mice and humans, HSPC differentiation follows a hierarchical tree-like continuum of states, in which branches terminate in committed cell lineages [36]. Strikingly, the distribution of E-cadherin expression within the two-dimensional representation of the HSPC cell lineage is not conserved between mice and humans ( Figure 4A,B). CDH1 expression in human HSPCs selectively marks precursors of the erythroid lineage and is hardly expressed in the basophil/mast cell lineage ( Figure 4A). In contrast, in mouse HSPCs, Cdh1 is mostly expressed by precursors of the basophil/mast cell lineage, in agreement with our findings ( Figure 4B). Overall, whereas E-cadherin expression is highly conserved in epithelia of mice and humans, the distribution of E-cadherin expression within the HSPC lineage is not consistently conserved. We additionally tested E-cadherin cell-surface expression on human basophil-enriched CD123 + HLA-DR -IgE + cell population derived from human peripheral blood mononuclear cells (PBMCs) ( Figure 4C, Supplementary Figure S4A,B) [37]. However, we did not detect E-cadherin membrane protein expression on this enriched population of human basophils, in contrast to human erythroblasts, which we included as a positive control ( Figure 4C, Supplementary Figure S4A-C). This corresponds with the human scRNAseq database, where E-cadherin mRNA is hardly detected in the basophil/mast cell lineage of HSPCs ( Figure 4A). The prominent E-cadherin expression on basophil precursors and circulating basophils in mice is therefore not conserved in humans.
We also examined CDH1 expression in 100 defined human leukemic cell lines using DSMZcellDive, an online tool from the Leibniz institute, and identified that the human leukemic cell line KU-812 displays high CDH1 expression ( Figure 4D) [32]. This is of particular interest as KU-812 cells have been characterized as displaying characteristics of late HPCs that are committed to the basophil lineage [38]. Intriguingly, early characterization of this cell line revealed that KU-812, as well as having basophil characteristics, also expressed erythroid defined markers, as defined by α-Globin and GYPA protein expression [38]. We confirmed these findings by examining the expression of erythroid markers (e.g., TFRC (encoding CD71), GYPA and HBA1 (encoding α-globin) using DSMZcellDive ( Figure 4D,E). These findings are of particular interest as the basophil lineage separates prior erythroid lineage formation in both mice and humans ( Figure 4A,B). KU-812 therefore may reflect transformed HPCs with bipotent lineage formation potential towards the basophil and erythroid cell lineages. Overall, the obtained results underscore a distinguished role for E-cadherin signaling in mouse basophils compared to human erythroid progenitors.

Discussion
In this study, we compared human with mouse hematopoiesis to examine whether mice can be used to investigate the function of E-cadherin in human hematopoiesis. We defined E-cadherin expression as mostly marking the basophil/mast cell lineage in murine BM, which is in agreement with previous findings [21,39]. E-cadherin expression also marks a small fraction of late myeloid HPCs that are able to reconstitute the myeloid and erythroid cell lineage in vitro under conditions that favor myeloid or erythroid outgrowth.
In contrast to that of the mouse, the human basophil lineage does not express detectable mRNA or protein levels of E-cadherin. E-cadherin, therefore, may display a mouse-specific regulator of basophil behavior, as also seen for Mcpt8, which also marks the basophil lineage in mouse but is absent in human basophils [28]. Alongside proteases, mouse and human basophils also display a distinct cytokine receptor profile, further underscoring the differences between basophil regulation in mice and humans [40][41][42].
Within the mouse basophil lineage, the molecular function of E-cadherin remains to be determined. However, despite the small sample size, we found E-cadherin not to be essential in driving basophil formation in mice. Instead, its function may underlie the immune response of basophils. Alongside homophilic interactions, E-cadherin also facilitates the formation of heterophilic interactions as described for T-cell and NK-cell associated ligands CD103 and KLRG1 [43][44][45][46][47][48][49][50][51]. E-cadherin-CD103/KLRG1 interactions mostly act as an anti-inflammatory regulatory signal and thereby may enable basophils to attenuate T-cell activity at the site of inflammation [49,50,52,53]. Additionally, basophils have been found to promote not only inflammatory responses, but also anti-inflammatory signals in the context of autoimmune disease, contact dermatitis and colitis [54][55][56]. Basophilic E-cadherin, therefore, may control the interplay of basophils with immune components to control their inflammatory response during inflammation. Future research should therefore address E-cadherin function under specific basophilic challenges such as parasite infections or allergies and regulation of its transcription.
Although we were unable to detect E-cadherin cell-surface expression on isolated human basophils from peripheral blood, human basophils may still upregulate E-cadherin expression at the site of inflammation. We identified that human-derived KU-812 cells, characterized as transformed late HPCs committed to differentiating the basophil lineage, also express E-cadherin as well as erythroid-defined proteins [38]. In mouse BM, we revealed that E-cadherin marks a subset of late HPCs that are also able to differentiate into the basophil lineage as well as into the erythroid lineage. Together these findings may imply that human BM likewise comprises a subset of late bifunctional HPCs that is marked by E-cadherin expression from which the leukemic KU-812 cell line upon transformation originated.
As well as distinguishing between the expression of E-cadherin in the basophilic lineages of mice and humans, our study also underscored the marked difference of Ecadherin expression within the erythroid lineages of mice and humans. Whereas E-cadherin mostly marks the erythroid lineage in human BM, its expression is absent within the erythroid lineage in mouse BM. More than two decades ago, human erythroblasts were already found to express E-cadherin, although, its regulation of erythropoiesis has never been resolved [22].
Whereas the regulation of mouse and human erythropoiesis is mostly resemblant under steady-state conditions, the homeostatic/erythropoietic response to anemia, also known as stress erythropoiesis, is differently controlled [57,58]. Stress erythropoiesis in mice is mainly driven by activating extramedullary erythropoiesis in the spleen; whereas, stress erythropoiesis in humans is predominantly (but not restrictively) activated in the BM [57,58]. Stress erythropoiesis in the mouse spleen is well-characterized, yet the molecular mechanisms underlying stress erythropoiesis in human BM are mostly elusive [59][60][61][62]. Although E-cadherin is not conserved in mouse splenic or BM erythroblasts, and neither controls stress erythropoiesis in mice, it may in fact involve the regulation of stress erythropoiesis in human BM, in which molecular regulation is mostly elusive.
Overall, hematopoietic E-cadherin reveals evolutionary discrepancies between mice and humans. This is also in line with a marked variation in transcriptome composition during terminal erythropoiesis between mice and humans [63,64]. One could therefore question if the mouse can be a representative model for human hematopoietic E-cadherin function and more general human erythropoiesis. This, therefore, also calls into question to what extent alternative animal models (e.g., zebrafish, rat, rabbit or monkey) can be used to model E-cadherin signaling in human erythropoiesis. Rats, for example, in contrast to mice, mimic stress erythropoiesis that occurs in humans by strongly expanding erythropoiesis in their BM [57].
Furthermore, it remains to be defined by what means E-cadherin expression is differently regulated in hematopoiesis in humans and mice. Our evaluation of defined regulators of E-cadherin expression (e.g., SOX2, ESR2, TCF7, GATA-2, GATA-3, GATA-4, SMARCA4, SMAD3, SOX9, Runx1, EOMES, PPAR), by data mining single-cell RNA-seq profiles of mouse and human hematopoiesis, did not reveal distinct expression patterns [36]. Yet, erythroid-defined transcription factors have been identified as displaying an altered chromatin occupancy in mice and humans [65]. This thereby implies that, even when expression levels of transcription regulators in mouse and human erythropoiesis are similar, their transcriptional outcome may differ. This could also explain why E-cadherin expression is not conserved within the human basophil lineage that, similarly to mice, is characterized by high GATA-2 expression, which is implicated in driving E-cadherin expression in the mouse basophil lineage [21]. Differences in E-cadherin expression between human and mouse hematopoiesis can also be induced by differences in promoter methylation, as methylation of the promoter region of CDH1 is well-known to downregulate E-cadherin expression [66,67].
Overall, we conclude that hematopoietic E-cadherin harbors both evolutionary conserved and species-specific functions for which models need to be carefully chosen.  Informed Consent Statement: Human material was obtained after informed consent was given in accordance with the Declaration of Helsinki, the Dutch National and Sanquin Internal Ethic Boards. All mice were maintained in the animal facility of the Netherlands Cancer Institute (NKI). Animal experiments were conducted according to institutional and national guidelines.

Data Availability Statement:
We will deposit our raw single-cell RNAseq datasets in publicly available repositories. The datasets and materials analyzed during the current study are available from the corresponding author upon reasonable request.