{"paper_id":"2444846b-e06c-4306-a5a9-b750425bd5e8","body_text":"1 \nSelf-organized hemanoids derived from human iPSCs create a niche that pro-\nduces definitive extraembryonic hematopoiesis. \nAfrim Avdili 1, Martina Auer 1, Dagmar Brislinger 2, Dagmar Kolb 2,3, Gerit Moser 2, Andreas \nReinisch1,4, Gerald Hoefler 5,6, Claudia Bernecker 1, Julia Fuchs 2,7, Julia Feichtinger 2, Peter \nSchlenke1, Isabel Dorn1,8. \n \n1Department of Blood Group Serology and Transfusion Medicine, Medical University of Graz, \nGraz, Austria. \n2Division of Cell Biology, Histology and Embryology, Gottfried Schatz Research Center for \nCell Signaling, Metabolism and Aging, Medical University of Graz, Graz, Austria. \n3Core Facility Ultrastructural Analysis, Medical University of Graz, Graz, Austria. \n4Division of Hematology, Department of Internal Medicine, Medical University of Graz, \n Graz, Austria. \n5Diagnostic and Research Institute of Pathology, Medical University of Graz, Graz, Austria. \n6Lung Research Center, Medical University of Graz, Graz, Austria \n7Institute of Biomechanics, Graz University of Technology, Graz, Austria \n8Lead contact  \n*Correspondence: isabel.dorn@medunigraz.at \nSummary \nManufacturing red blood cells (RBCs) from hu man induced pluripotent stem cells (iPSCs) \ncan improve our understanding of embryonic eryt hropoiesis, foster innovative treatments for \nRBC-related diseases, and ultimately address clinical blood supply shortages. However, \nexisting systems face low efficiency, enucleation failure, and uncertainty about the develop-\nmental wave of cultured RBCs. We successfully used self-organized hemanoids to improve \niPSC-derived RBC generation. Based on the hypothesis that cellular interactions and 3D \norganization promote hematopoietic cell fate, we aimed to thoroughly characterize \nhemanoids. We visualized the spatiotemporal emergence of hematopoiesis by generating a \nCD43-GFP reporter iPSC line. Imaging and spatial transcriptomics analysis provided de-\ntailed insight into the hemanoid architecture , identifying stromal cells and hepatoblasts as \npotential erythropoiesis-supportive elements. The developmental stage mirrors \nextraembryonic hematopoiesis. Given the difficult ies of accessing these early stages in vivo, \nour system offers a platform not only for further  clinical translation but also for exploring hu-\nman embryonic blood wave dynamics.  \nKeywords \nRed blood cell, erythropoiesis, hematopoiesis, iPSC, embryonic, yolk sac, fetal liver, CD43-\nreporter, STEM, Spatial transcriptomics. \n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted May 8, 2026. ; https://doi.org/10.64898/2026.05.05.722134doi: bioRxiv preprint \n\n2 \nIntroduction \nCulturing human RBCs ex vivo from induced pluripotent stem cells (iPSCs) has enormous \npotential to expand diagnostic and therapeutic op tions for RBC-related diseases and to ad-\ndress the increasing shortage in clinical blood supply. 1,2 Furthermore, it offers a unique op-\nportunity to deepen our limited understanding of human developmental erythropoiesis. Early \ndevelopmental stages are not readily accessible in humans, and animal models, like mice, \ndiffer in key aspects.   \nThe first blood cells originate in the yolk sac (YS), supplying the developing embryo with \nnutritional, metabolic, and oxygen support. Within the first 18 days post-conception (dpc) \n(Carnegie stage (CS) 7-8), mesenchymal cells adjacent to the endoderm differentiate into \nhematopoietic cells (HC) and become enveloped by endothelial cells, which later form the \nYS vascular plexus.\n3–5 This first wave of “extraembryonic primitive hematopoiesis“ generates \nRBCs, megakaryocytes, and macrophages. The YS is also the origin of the second wave, \ntermed “extraembryonic definitive hematopo iesis”, generating erythr o-myeloid progenitors \n(EMPs) from the endothelium through  a gradual process known as endothelial-to-\nhematopoietic transition (EHT)  (~28-35 dpc, CS 13-15). With the onset of fetal circulation, \nEMPs leave primitive blood islands and migrate from the YS to the FL, where they produce a \nbroader spectrum of HCs, including granuloc ytes, monocytes, and mast cells. Both \nextraembryonic waves are ultimately replaced by  a third wave of intraembryonic hematopoi-\nesis originating in the aorto-gonad-mesonephr on (AGM) region (~30-32 dpc, CS14). AGM-\nderived hematopoietic stem cells (HSCs), possessing self-renewing properties and en-\nhanced lymphoid potential, initially colonize the FL alongside EMP-derived cells, before mi-\ngrating into the bone marrow (BM) to sustain lifelong hematopoiesis. 5–7 The different waves \nof human hematopoiesis overlap temporally and spatially throughout development. Prelimi-\nnary exploration of cellular morphology and cell surface marker expression has yielded no \nclear markers for assigning individual cells to a specific wave. A distinction is likely possible \nbased on differences in gene expression profiles, such as HSC-specific signature genes or \nglobin genes.8–13 Primitive RBCs are characterized by  expression of embryonic hemoglobins \n(embHb) Gower I ( ζ2ε2) and Gower II ( α2ε2). While EMP- and HSC-derived erythropoiesis in \nthe FL synthesize predominantly fetal hemoglobin (HbF, α2γ2), BM-derived cells produce \nadult hemoglobin (HbA, α2β2).4 \nSince the discovery of iPSCs, several culture systems have been developed to model hu-\nman erythropoiesis. To simulate the complex in  vivo situation, researchers use extensive \ncytokine stimulation in combination with digestion and purification steps. Despite significant \nprogress in recent years, established systems re main severely limited by low cellular yields \nand a failure to achieve terminal enucleation (below 10%).\n2,14,15 In addition, it remains un-\n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted May 8, 2026. ; https://doi.org/10.64898/2026.05.05.722134doi: bioRxiv preprint \n\n3 \nclear which developmental wave iPSC-deriv ed cultured RBCs (cRBCs) represent. Because \ncellular interactions are essential for cell fate decisions, organoid-based systems have been \nestablished to differentiate pluripotent cells  more effectively into various cell types. 16–19 As \nthe environment may also substantially impact hematopoietic development  20,21, our group \ndeveloped a simplified erythropoiesis model based on the formation and maintenance of \nself-organized 3D complexes, termed „hemanoids“. 22,23 Minimal cytokine stimulation  (inter-\nleukin-3 (IL-3), stem cell factor (SCF), and erythropoietin (EPO)) 24–27 results in the continu-\nous release of HCs from hemanoids into the cu lture supernatant over several weeks, which \ncan be further differentiated into RBCs.  High expansion and enucleation rates (up to 60%) \nalready enabled us to confirm that cRBCs exhibit morphological, biomechanical, and blood-\ngroup antigen expression profiles compar able to those of cord blood-derived \nreticulocytes.22,23 Interestingly, by altering cytokine stimulation, the system can produce mac-\nrophages and granulocytes and is further scalable in stirred bioreactors.28,29 \nDespite these promising results, the underlying erythropoiesis supporting mechanisms re-\nmain unclear. For further improvement and meaningful  future application of the system, it is \ncrucial to comprehend the tissue architecture of hemanoids, the spatiotemporal development \nof hematopoiesis inside hemanoids, and to clarify which developmental wave the generated \nerythropoiesis corresponds to. To achieve these objectives, here we generated a CD43-GFP \nreporter iPSC line and tracked hematopoietic emergence within hemanoids. We investigated \nthe hemanoids architecture by immunohistochemistry (IHC) and scanning transmission elec-\ntron microscopy (STEM). Additionally, we performed spatial transcriptomics (ST) analysis to \ncorrelate histological characteristics with t he transcriptional profile. We identified stromal \nelements and hepatoblasts as potential hematopo iesis-supportive interaction partners. Re-\nsults argue for a developmental stage of EMP-derived hematopoiesis. The hemanoid model \nthus offers a unique platform for modeling def initive extraembryonic erythropoiesis spanning \nthe undifferentiated iPSC stage through the enucleated RBC stage. Because in vivo access \nto these stages is limited, our model bridges existing gaps and offers valuable insights into \nembryonic erythropoiesis, potentially serving as a platform for future clinical applications.  \n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted May 8, 2026. ; https://doi.org/10.64898/2026.05.05.722134doi: bioRxiv preprint \n\n4 \nResults  \nHemanoids enable hematopoietic and erythroid differentiation of human iPSCs \nWe initially applied three iPSC lines 30–32 as biological replicates to our culture system ( Fig-\nure 1 ).22,23 After about five days of cytokine stimulation (phase I), self-organized three-\ndimensional hemanoids were established, cons isting of an adhesive stromal layer, spherical \nstructures, cell-dense areas, and, in individua l cases, macroscopically detectable red is-\nlands. Throughout culturing, the size of the hemanoids increased significantly (Figure S1A). \nFrom approximately day 14 onwards, hemanoids continuously released CD43+ HCs (purity \n> 95% measured by flow cytometry) into the supernatant ( Figure S1B), reaching a cumula-\ntive number of 3.1×10 6 ± 1.7×106 CD43+ cells over 5 weeks (n=6, mean ± SD; per one well \nof a six-well plate containing 1-2 hemanoids) . Thereafter, the potential of hemanoids to re-\nlease HCs was exhausted. For further erythroi d differentiation, released CD43+ cells were \nrepeatedly collected and subjected to RBC differentiation 33 over an additional 18 days \n(phase II). Cells showed homogenous maturation into 99% GPA+/Band3+ erythroid cells 34 \n(Figure S1C). On day 18, all cells expressed hemoglobin, and the percentage of enucleated \ncells exhibiting the typical RBC morphology reached 39.1% ± 16.4% ( Figures S1D-E ). A \nmean cumulative expansion of 3,673 ± 2,037-fo ld was observed during the 18-day erythroid \ndifferentiation phase ( Figure S1F). These results are comparable to those reported in our \nprevious study22, despite being derived from different iPSC sources. They, therefore, confirm \nthe robustness and reproducibility of the hemanoid system. The advantages of this system \nare i) minimal handling, ii) low cytokine supplementation (only SCF, EPO, IL-3), iii) the con-\ntinuous release of a pure population of CD43+ HCs into the supernatant, and iv) enhanced \nexpansion and enucleation of erythroid cells. We  hypothesize that these benefits arise from \na specialized microenvironment and cellular interactions within the hemanoids. \nSpatiotemporal emergence of CD43+ hematopoietic cells  \nBecause CD43 (leukosialin) is considered the first specific pan-hematopoietic cell-surface \nmarker during human pluripotent stem cell differentiation\n35, we generated a CD43-GFP fluo-\nrescent reporter iPSC line (CD43R-iPSC) to  monitor the emergence and subsequent distri-\nbution of HCs within hemanoids. Using CRISPR/Cas9 gene editing, we targeted the AAVS1 \nsafe harbor locus on chromosome 19 to introduce the CD43_GFP vector construct through \nadeno-associated virus serotype 6 (AAV6)-mediated homology-directed repair (HDR)36–39, as \nshown in Figure S2.  Successfully manipulated cells from the iPSC lines PEB-AL#6 30 and  \nCM1#132 were clonally expanded. Precise CD43 reporter knock-in into the AAVS1 locus was \nconfirmed by PCR genotyping (Figure S2E). The CD43R-iPSCs showed comparable hema-\ntopoietic and erythroid potential as their parental  cell lines, as demonstrated by the number \n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted May 8, 2026. ; https://doi.org/10.64898/2026.05.05.722134doi: bioRxiv preprint \n\n5 \nof CD43+ cells released into the supernatant, hematopoietic colony-forming potential in sem-\nisolid media, and erythroid differentiation capacity (Figure S3).  \nTo identify the earliest CD43-GF P+ cells within hemanoids, we conducted time-lapse mi-\ncroscopy starting on day 5 of hematopoietic specification (phase I) ( Figure 2A ). The first \nGFP-expressing cells emerged on day 6, as confirmed by three independent experiments. \nCD43-GFP+ cells initially appeared in specific  locations within cell-dense areas and subse-\nquently expanded from there for approximately 5 weeks. GFP+ cells migrated within \nhemanoids, before being released as single cells into the supernatant ( Figures 2B-D). Flow \ncytometry characterization following cell-surface CD43 staining confirmed the correlation \nbetween endogenous CD43 gene expression and GFP expression from the inserted CD43-\nreporter construct (Figures 2E and 2F). Before their release, GFP+ HCs strongly adhered to \nthe hemanoid surrounding stromal layer ( Figure 2G). This adhesion was resistant to disrup-\ntion, e.g., by extensive washing steps.  \nThe heterogeneous tissue organization of the hemanoid reveals its complexity \nHemanoids consistently form blister-like structures, often more than one, filled with fluid \n(Figure 3A). In addition, they develop a stromal cell layer that extends beyond the complex \nand mediates adhesion of the hemanoid to the culture plate ( Figure 3B ). In our previous \nstudy22, we observed that both features are crucial for successful hematopoietic induction. In \ncontrast to the pure HC population released into the supernatant, we found only 49.6% ± \n8.8% CD43+ HCs within the hemanoids, measured by flow cytometry after enzymatic diges-\ntion (Figure 3C). Hematoxylin/eosin (HE)-stained tissue sections obtained between 16 and \n28 days of hematopoietic specification (phase I) showed morphological features comparable \nto those of ectodermal (neural crest tube-like  structures), mesodermal (stromal tissue and \nblood cells), and endodermal origin (glandular structures) ( Figures 3D and S4A). Especially \nat early stages (<20 days), CD43+ HCs were primarily found within VE-Cadherin+ (CD144) \nvessel-like structures that resembled YS blood islands morphologically. Vessels were sur-\nrounded by vimentin+/β -laminin+ mesenchymal-like cells (Figures 3E and S4B-C). CD163+ \nmacrophages were the only HC type predominantly located outside blood vessels and within \nthis stromal compartment ( Figure S4D). A notable difference in older hemanoids (day 28) \nwas the predominance of hematopoiesis outside blood vessels and within the mesenchymal \ncompartment ( Figures 3F and S4B ). The hematopoietic compartment itself contained \nerythroid precursor cells, granulocytes, monocytes, megakaryocytes, platelets, and mast \ncells (Figures 3G-H) as confirmed by antibody-mediated staining (Figures S4D-H ). CD61+ \nmegakaryocytes exhibited a small cell size (~20 µm) typical of their juvenile maturation \nstage\n40 (Figure S4E). A striking feature was the predominance of eosinophilic granulocytes \n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted May 8, 2026. ; https://doi.org/10.64898/2026.05.05.722134doi: bioRxiv preprint \n\n6 \n(Figures 3G and S4F ). Although all HC types were detectable in day 16 hemanoids, mast \ncells and granulocytes became more abundant in older hemanoids.  \nTo extend our imaging to the nanometre leve l, we performed STEM analysis of tissue sec-\ntions (Figures 4 and S5). Ultrastructural analysis confirmed our IHC findings, showing blood \nislands with various HC types lined by endothelial cells in day 17 hemanoids ( Figures 4A-C \nand S5A). The surrounding stromal compartment was characterized by a loose network of \nmesenchymal-like cells, fibroblasts, collagen fibers, and individual tissue macrophages ( Fig-\nure 4D). By day 28, the hemanoids exhibited a significant transformation; the stromal tissue \nhad become highly organized, with a dense network of collagen fibers produced by an in-\ncreased number of stromal cells, yet hematopoiesis had declined ( Figure S5B). Blood is-\nlands became less distinct and showed signs of di sruption, while HCs were primarily located \nwithin the stroma ( Figure S5C). Already by day 17, STEM revealed cell-cell interactions be-\ntween erythroid precursors and other cells, suggesting metabolite exchange and mutual in-\nfluence (Figure S5D ). The outer surface of the hemanoid was covered by a conspicuous \nepithelial layer of endodermal origin (AFP+, ß-laminin+, HLA-G-, CK7-) ( Figure S5E). This \nlayer was further characterized by a basement  membrane, intracellular vesicles, and cell \nextensions at the outer surface. Cells were partially connected via desmosomes.  \nSpatial transcriptomics confirmed tissue across the three germ layers and distinct \nhematopoietic populations.  \nTo further extend our results to a molecular level and specify the developmental wave of \nhemanoid-derived hematopoiesis, we performed a 10X Visium spatial transcriptomics (ST) \nanalysis of day 16 and day 28 hemanoids ( Figure 5A). Day 16 was chosen based on previ-\nous results showing an initial increase in CD43+/CD45+ HCs at this culture day. 41 Day 28 \nwas selected as a later stage in hematopoietic development, prior to exhaustion of the sys-\ntem. Ten to eighteen hemanoids from PEB-AL#6 iPSCs were pooled to cover the capture \narea of the ST slide (6.5 mm x 6.5 mm, Figure S6A). We initially focused on evaluating the \nday 16 data. RNA expression data from 1,002 spots in the capture area were categorized \ninto 10 clusters and annotated using Azimuth with reference datasets for human embryonic \ndevelopment42 and tissue-specific marker genes ( Figures 5B-C and S6A ). Erythroid-\nmegakaryocyte progenitors (EryMK), erythroid pr ecursors (Erythroid), myeloid precursors \n(Myeloid), and stromal cells (Stroma) repr esented mesoderm-derived tissue. Hepatoblast-\nlike cells (Hepatoblasts), intestinal/bronchoalveolar epithelial-like cells (Epithelial), and a \ncluster enriched in endodermal gene expression (Endo) represented the endodermal tissue, \nwhile neuroprogenitors (Neuro) and photoreceptor cells (PRC) were categorized as ecto-\ndermal-derived clusters. Figure 5D  shows the expression of two representative marker \ngenes per annotated cluster, such as NES and MAP6 in neuroprogenitors (ectoderm), \n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted May 8, 2026. ; https://doi.org/10.64898/2026.05.05.722134doi: bioRxiv preprint \n\n7 \nSERPINA1 and FGB in hepatoblasts (endoderm), and HAND1/2 in stromal cells (mesoderm) \n(Figure 5D ). One cluster could not be assigned (undef ined). It likely contains transcripts \nfrom a mixed cell population, as also visible in the heat map ( Figure S6B).  The contribution \nof more than one cell type to individual clusters was expected, given the 55 µm spatial reso-\nlution (spot size) achievable with the 10X Visium system.  \nWhen all clusters were examined for a hematopoietic signature, this was mainly detectable \nin the EryMK cluster and, to a lesser extent, in the Myeloid and Erythroid clusters ( Figure \n5E). The EryMK and Myeloid clusters expressed common hematopoietic genes RUNX1, \nSPN (CD43), and GATA2 . The EryMK cluster expressed genes relevant for platelet-\nmediated hemostasis, like PF4 (platelet factor 4), PPBP (pro-platelet basic protein), and GP9 \n(glycoprotein IX Platelet), but also  erythroid-specific globin genes ( HBG1, HBE1, HBA1), \nGYPA (glycophorin A), SLC4A1 (Band3), and GATA1 ( Figures 5E-F). Upregulated genes \nwithin the Erythroid cluster encode essential RBC components, including globin genes \n(HBG1, HBG2, HBE1, HBZ, HBA1/2), PKLR (pyruvate kinase), ANK1 (ankyrin), and \nSLC4A1. Based on the gene expression profile, the Erythroid cluster may encompass the \nmore differentiated erythroid population, wher eas the more immature progenitor population \ncontributes to the EryMK cluster. The Myeloid cluster expressed the macrophage-associated \ngenes MRC1, CD33, and FCGR2A (Figure 5E ), as well as granulocyte-associated genes \nPRG2, MMP9, IL6R, AIF1, and SLPI  (Figure 5F). Feature plots illustrating expression pat-\nterns for selected hematopoietic marker genes are shown in Figure S7. Gene set enrich-\nment analysis (GSEA) revealed enrichment of the “response to oxygen” and “platelet activa-\ntion” pathways in EryMK. Myeloid progenitors were enriched for “Leukocyte activation” and \n“Adaptive immune response” pathways ( Figure S8A). Clusters were further validated using \nGene Ontology (GO) analysis. In Enriched pathways in EryMK play a crucial role in erythro-\ncyte differentiation and blood coagulation. In the Myeloid cluster, pathways were associated \nwith the regulation of immune effector processes, leukocyte migration, and activation ( Fig-\nure S8B).  \nTranscriptional profiles indicate a developmental stage that mirrors definitive \nextraembryonic hematopoiesis. \nMorphologically detectable red islands already indicate hemoglobin production within day 16 \nhemanoids. ST data confirmed the expression of globin genes in the Erythroid and the \nEryMK cluster ( Figures 6A-B). Both clusters expressed HBZ and HBA1/2 from the alpha-\nglobin locus and embryonic HBE1 and fetal HBG1/HBG2 from the ß-globin locus, indicative \nof the synthesis of embHb (Gower I and Gower II) and HbF. Definitive HBB was barely de-\ntectable. Whereas primitive RBCs primarily express embHb, the coexpression of embHb and \nHbF is a typical pattern in EMP-derived erythropoiesis\n4 and was confirmed on a protein level \n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted May 8, 2026. ; https://doi.org/10.64898/2026.05.05.722134doi: bioRxiv preprint \n\n8 \nby IHC staining of hemanoids ( Figure S9A ). Although hemanoids expressed SOX6 and \nBCL11A, as repressors of embHb and HbF 43–45,  we did not observe their significant \nupregulation in one of the hematopoietic clusters (Figures 6C and S9B-C). \nGATA1, KLF1, TAL1, LMO2, and LDB1, as core erythroid TFs 46,47, were expressed in the \nEryMK cluster ( Figure 6C). Additional MYB  expression confirms an advanced EMP stage, \nas MYB is not expressed in the primitive program. 48,49 Moreover, we compared our dataset \nwith a set of marker genes ( RUNX1, HOXA9, MLLT3, MECOM, HLF, SPINK2 ), recently de-\nscribed by Calvanese et al. 9, to distinguish intraembryonic AGM-derived HSCs from their \nextraembryonic progenitors. Although we observe d expression of all six genes across the \nEryMK and the Myeloid cluster, there was no clear match of all markers to a single \nsubcluster ( Figures 6C and S9B-C ). Enhanced lymphopoiesis as a hallmark of AGM-\nderived hematopoiesis was not detectable. Hematopoietic clusters showed minor expression \nof CD3 (T-lymphocytes), but lacked expression for MS4A1 (CD20) or CD19 (B-lymphocytes) \n(Figures 5E and S7). IHC confirmed individual CD3+ cells, whereas CD20+ B cells were \nabsent (Figure S9D). Therefore, we could not confirm a cell population comparable to AGM-\nderived hematopoiesis. The development of AGM-derived hematopoiesis is significantly in-\nfluenced by WNT signaling (controlling early mesodermal patterning) 50, NOTCH signaling \n(essential for arterial hemato-vascular development) 51,52, and HOX pathways. 9,12 Genes in-\nvolved in WNT signaling, NOTCH genes (NOTCH 1–4 ), NOTCH ligands ( JAG1/2, DLL1, 3, \nand 4), and target genes (HEY1/2, and HES1-5) were not significantly expressed in hemato-\npoietic clusters. HOX gene expression was also sparse, although the EryMK and myeloid \nclusters indicate low HOXA9 and HOXA10 expression. In contrast to their low expression in \nthe hematopoietic clusters, WNT, NOTCH, and HOX pathways were significantly \nupregulated in the Neuro cluster (Figure 6D).  \nThe transcriptional profile of hemanoids overlaps with that of human YS and FL.  We \nset out to determine whether day 16 hemanoids show similarities with the human YS and FL, \nas sources of primitive and EMP-derived HCs. We integrated our ST data with two publicly \navailable scRNA-seq datasets on the developing human YS at CS 10/11 53 and 17 54 using \nHarmony.55 After single-cell data processing and filtering, we retained gene expression data \nfrom 7,545 cells from CS 17 and 10,893 cells from CS 10/11. UMAP visualization reveals a \nsimilar expression pattern among certain clusters ( Figure S10A). We further integrated our \nST data with two datasets on human FL development: one from CS 20 and 23 53, and the \nother from CD45+ isolated FL cells from post-conception weeks (wpc) 8–16. 11 Our data also \nshow overlap with cell populations during human FL development (Figure S10B).  \n \n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted May 8, 2026. ; https://doi.org/10.64898/2026.05.05.722134doi: bioRxiv preprint \n\n9 \nDay 28 hemanoids show a reduced hematopoietic potential and no further induction \nof AGM-derived definitive hematopoiesis \nSince prolonged culturing of hemanoids might induce AGM-derived hematopoiesis, ST \nanalysis was extended to day 28 hemanoids. RNA expression data derived from 1,349 spots \nof the capture area were categorized into 7 clusters and annotated using Azimuth with refer-\nence datasets for human embryonic development\n42 (Figures S11A-B ). The mesodermal \ntissue was represented by an erythroid cluster (Erythroid), a cluster enriched for myeloid and \nmegakaryocyte gene expression (Hemato), a stromal cluster (Stroma), and a separate myo-\nfibroblast cluster (Myofibroblast). Hepatoblas t-like cells (Hepatoblast) and neuroprogenitors \n(Neuro) represented endodermal and ectodermal tissue. One cluster could not be assigned \n(Undefined). Interestingly, the hematopoietic co mpartment's contribution to the overall gene \nexpression profile decreased compared with day 16 (9% versus 23%), whereas the Neuro \n(37% versus 21%) and Stroma (21%  vs 8%) contributions increased. Figure S11C shows \nthe expression of three marker genes per cl uster. A hematopoietic gene expression profile \nwas primarily detected in the Hemato and Erythroid clusters ( Figure S11D). Both expressed \nembryonic and fetal globin genes in the absence of ß-globin. In line with this, upregulation of \nBCL11A and SOX6  by HCs was not detectable ( Figures S11E-F). Investigation of hemato-\npoiesis-related signaling pathways NOTCH, WNT, and HOX, and expression of signature \ngenes for AGM-derived HSCs\n9 gave no evidence of further induction of AGM-derived defini-\ntive hematopoiesis in day 28 hemanoids (Figures S11F-G). Transcriptome profiles of day 16 \nand 28 hemanoids were further integrated and batch-corrected using Harmony. 55 Unsuper-\nvised UMAP clustering grouped the cell populations from the two samples into 9 clusters. \nDay 16 and day 28 samples displayed a comparable pattern of gene expression. However, \nthe overall expression of hematopoiesis-related genes was reduced in day 28 samples. In \nparticular, the expression of early hematopoi etic markers CD34, SPN, RUNX1, and GATA2 \nwas very low, whereas connective tissue genes like COL3A1 and neuroprogenitor-related \ngenes like NES and SOX11 remained consistently expressed (Figures 6E-F and S10C).  \nStroma cells and hepatoblasts provide a supportive niche for early hematopoiesis \nMicroscopic evaluation of the ST tissue sections shows that already by day 16, hemanoids \nare capable of producing mature RBCs without  undergoing the erythroid differentiation step \n(phase II) of our protocol ( Figure 7A). This highlights the potential of hemanoids to create a \nsupportive environment for erythroid developm ent. In addition to the obvious interaction be-\ntween HCs and stromal cells, ST analysis revealed a close proximity to hepatoblast-like \ncells. Cells with a hepatoblast-like gene expression profile were often located near primitive \nblood islands ( Figure 7B). We analyzed gene expression patterns within these clusters, fo-\n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted May 8, 2026. ; https://doi.org/10.64898/2026.05.05.722134doi: bioRxiv preprint \n\n10 \ncusing specifically on genes encoding growth factors, adhesion molecules, and their coun-\nterparts.  \nThe Hepatoblast cluster showed upregulated gene expression (e.g., SERPINA1 (Alpha-1 \nAntitrypsin), HNF4A (Hepatocyte nuclear factor 4-alpha), and ACSL5 (Acyl-CoA \nSynthetase)) and enrichment of pathways directly linked to liver cell metabolism ( Figures \n7B-C and S12 ). In addition to genes involved in lipid metabolism (e.g., APOA, APOB, and \nACSL5), expression of genes involved in iron and Vitamin B12 metabolism was observed, \nsuch as FTL (Ferritin-L chain, involved in iron storage, TF  (transferrin), and CUBN (cubilin, a \nvitamin B12 receptor complex). The Hepatoblast cluster further expressed IGF2 (insulin-like \ngrowth factor 2) and, to a minor extent, EPO ( Figures 7C-D and S12 ). We found \nupregulation of genes encoding various ECM proteins, including vitronectin ( VTN), \nfibronectin 1 ( FN1), fibrinogens ( FGA, FGB, FGG ), and laminins β 3 ( LAMB3) and C1 \n(LAMC1) (Figure 7E). Recent reports on primary human YS hematopoiesis 10,56, have identi-\nfied an interaction between endodermal vitronectin and αvβ1-integrin, or the integrin subu-\nnits alpha2b, beta3, and beta5 on HCs, and between fibronectin and α4ß1- and αvß1-\nintegrin on HCs. We confirmed upregulation of all the corresponding integrin-coding genes in \nthe EryMK cluster ( ITGA2B, ITGA4, ITGA5, ITGAV, ITGB 1, ITGB3, and ITGB5, as well as \nGFI1B and NFE2, involved in ITGB3 signaling57) (Figure 7F).  \nIn the Stromal cluster, the transcriptional profile and GO enrichment analysis confirmed in-\nvolvement in “Extracellular matrix organization” and “connective tissue development” (Figure \nS13A-B). Upregulated genes include those enc oding various collagens, fibronectin (FN1 ), \nlaminin-ß2 (LAMB2), and, to a lesser extent, fibrillin ( FBN1) (Figures 7E and S13C-E). Sev-\neral top upregulated genes regulate cellular development ( DSC3, GATA6, LMCD1), and  \ncardiac and angiogenic differentiation  (HAND1/2, TBX20, TNNT2, OLFML3). We further ob-\nserved increased activity in TGF-ß pathways, including TGFB2, BMP5 (encoding a secreted \nTGFß ligand58), FMOD (coding fibromodulin, which regulat es TGF-ß activity by sequestering \nTGF-ß in the ECM 59), and TGFBI (TGF-ß-induced protein that influences cell adhesion). \nNotably, the EryMK cluster showed a significant rise in TGFB1 expression (Figure 7D, S13B \nand S13F), while myeloid cells expressed FBN1 (Fibrillin), which controls TGF-ß bioavaila-\nbility and interacts with α5ß1- and αvß3-integrins (Figure 7E).60,61 \nIn human BM and FL, erythroid maturation occurs in close contact with CD163+ macro-\nphages in erythroblastic islands (EBI). Interact ion is mediated by a4ß1-integrin, EMP, and \nICAM4 on RBCs and VCAM1, EMP, and α vß3-integrin on macrophages. 62–65  Although we \nfound no obvious morphological correlates of EBIs, the myeloid cluster expressed CD163 \n(confirmed by IHC staining), MAEA (coding EMP), MRC1, VCAM1, and SIGLEC1, in line \nwith the phenotype of EBI macrophages. In addition to ITGA4  and ITGB1 (coding integrin \n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted May 8, 2026. ; https://doi.org/10.64898/2026.05.05.722134doi: bioRxiv preprint \n\n11 \nα4β1-integrin), the EryMK cluster also expressed ICAM4  as a potential interaction partner \n(Figure S13G).  \n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted May 8, 2026. ; https://doi.org/10.64898/2026.05.05.722134doi: bioRxiv preprint \n\n12 \nDiscussion  \nUsing an integrated imaging and transcriptomic approach, we obtained detailed insights into \nself-organized hemanoids that facilitate improved RBC generation from human iPSCs. The \ncomposition of the hemanoid is highly heterogeneous, comprising ectodermal , mesodermal, \nand endodermal tissue. The surface was covered by an epithelial layer of endodermal origin \nthat may function in the exchange of nutrients, metabolites, and water between the \nhemanoid and the culture medium. Others reported trophoblast-like features. 66 Although \nmorphology suggested this direction, the lack  of CK7 and HLA-G expression prevented con-\nfirmation in our study. 67 The established CD43-GFP reporter iPSC lines revealed the onset \nof hematopoiesis within hemanoids on day 6 of cytokine stimulation. This aligns with a report \non remodeling EHT ex vivo, describing the appearance of initial CD43+ cells on days 6-8. 35 \nCD43+ cells expanded from their area of origin and migrated within hemanoids before being \nreleased into the supernatant from day 14 onward. Hematopoiesis was initially  organized in \nblood islands morphologically resembling the YS vascular plexus. 68 Macrophages were the \nonly HC type in the extravascular connective ti ssue, which might support the assumption \nthat primitive macrophages, unlike primitive RBCs and megakaryocytes, originate from a \nmonopotent progenitor. 69,70 In older hemanoids, t he endothelial barrier was disrupted, and \nHCs became evenly distributed between the vess el rudiments and the connective tissue in \nparallel to their continuous release into the supernatant. We speculate that the growing HC \nmass within blood islands and the increasing mechanical pressure disrupt the endothelial \nlayer, rather than the transendothelial migration of hematopoiesis. Further investigation is \nneeded, such as using an endothelial reporter alongside the CD43 reporter, to clarify how \nHCs migrate from blood islands into connective tissue and whether similar mechanisms may \nbe relevant in vivo, such as the transition of YS-derived EMPs to the FL. During continuous \nculturing, the stromal compartment became highly organized, while hematopoiesis de-\ncreased, aligning with a reduced release of HCs into the supernatant. Our observations \nclosely overlap with recent data from the human YS, showing that the hematopoietic-to-\nstromal cell ratio decreases from young (CS10) to old (CS22) YS. The authors suggested \nthat a loss of stromal support between 6-8 wpc in the YS leads to apoptosis and depletion of \nremaining hematopoiesis through terminal differentiation.\n10 Despite the contribution of vari-\nous cell types to the hemanoid, exclusively CD43+ HCs were released into the supernatant. \nWe hypothesize that developing HCs lose thei r adherent properties and can emigrate from \nthe hemanoid scaffold, while the framework of  non-hematopoietic cells remains captured \nwithin. CD43 itself has been proposed to mediate anti-adhesive properties of HCs. 35 Com-\nbining the CD43R-iPSC line with approaches to di srupt different receptor-ligand interactions \n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted May 8, 2026. ; https://doi.org/10.64898/2026.05.05.722134doi: bioRxiv preprint \n\n13 \n(described below) may shed light on the adhesion capacities of extraembryonic HCs during \ntheir maturation.  \nThe presence of monocytes, mast cells, and granulocytes, along with expression of \nMYB49,71,72, and fetal globin genes, indicates the presence of a more advanced hematopoiet-\nic wave that at least corresponds to EMP-derived hematopoiesis. 3,4 Protein  analysis con-\nfirmed about 70% HbF and 20% embHb in hemanoid-derived cRBCs. 22  Further induction of \nAGM-derived hematopoiesis appears to be limited. Adult β -globin and increased expression \nof negative regulators of embHb and HbF (SOX6 and BCL11A) were not detectable. 4,43–45 \nLymphoid potential was restricted to a few CD3+ cells, and we could not confirm expression \nof recently described signature genes for AGM-derived HSCs 9 within a single hematopoietic \ncluster. Overall, our results suggest that predominantly an intermediate EMP program is pre-\nsent in day 16 and day 28 hemanoids, rather than a definitive AGM-derived program capable \nof producing cells resembling in vivo-generated HSCs. The development of AGM-derived \nhematopoiesis is significantly influenced by WNT signaling\n50, NOTCH signaling 51,52, and \nHOX pathways.9,12 The experimental conditions used in our study do not specifically target \nthese pathways. We suggest that hematopoie sis in the hemanoid system becomes exhaust-\ned after 5 weeks because i) AGM-derived hematopoiesis is not induced, ii) a complete FL \nenvironment is absent, and iii) support from the microenvironment diminishes. It would be \nvery interesting to see if AGM-derived hematopoiesis could be induced by modifying HOX \npathways, as recently described.73 \nOur study aimed to identify niche factors that influence hematopoietic cell fate. ST data re-\nvealed expression of distinct adhesion molecules on HC populations. Besides inter-\nhematopoietic cell contacts, we identified hepatoblasts and stromal cells as potential interac-\ntion partners. During human embryogenesis, the FL acts as the second major site of hema-\ntopoiesis, with crosstalk between hematopoietic and FL cells.\n74 In vivo, liver rudiments de-\nvelop as a diverticulum from the floor of the embryonic gut around 21 dpc (CS10). 5 Unex-\npectedly, we consistently identified clusters  of hepatoblasts near blood islands. Based on \ntheir transcriptional profile, they might suppor t hematopoiesis by providing growth factors \n(IGF-2, EPO) and their involvement in iron, lipid, and vitamin B12 metabolism. We recently \ndemonstrated the importance of lipids for RBC differentiation. 33 EPO expression by endo-\ndermal cells of the human YS or FL has been reported, 10,24,75 as has IGF-2 production by FL \ncells, and its importance for HSC expansion. 76,77 Interestingly, hepatoblasts contribute to the \nproduction of ECM (vitronectin, fibronectin, fibrinogen, and the laminin-ß3 chain). While \nlaminin mediates cell attachment, migration, and tissue organization during \nembryogenesis78–80, the expression of vitronectin and fibr onectin in FL is discussed to con-\ntribute to FL colonization by YS EMPs and HSCs. 81 Moreover, studies on primary human YS \n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted May 8, 2026. ; https://doi.org/10.64898/2026.05.05.722134doi: bioRxiv preprint \n\n14 \nhematopoiesis suggest that endodermal vitronectin and YS fibronectin interact specifically \nwith distinct integrins on HCs 10,56 and thereby modify HSC function, expand the HSC pool, \nand contribute to long-term HSC quiescence. 10,82,83 We confirmed gene expression for all \ncorresponding integrin subunits in the EryMK cluster, which might indicate similar mecha-\nnisms in hemanoid-derived hematopoiesis.   \nIn the hemanoid system, 3D-organization and stromal cell formation are essential for HC \nproduction.22 The CD43R-iPSC line revealed strong attachment of HCs to the stromal com-\npartment, consisting of MSCs, fibroblasts, and various ECM molecules. Based on ST data, \nand in line with published data on YS hematopoiesis 10, attachment of HCs to the ECM might \ninvolve the collagen receptor CD36, and  the fibronectin receptors α 4β 1- and α vβ 1-integrin. \nThe supportive role of MSCs or ECM in erythroid differentiation is well established, as they \nare used to enhance hematopoietic or  erythroid growth in culture. 82,84 Several top DEGs in \nthe Stromal cluster are known to regulate cellular development and differentiation. Interest-\ningly, we observed upregulated TGF-ß pathways across different cell types, including stro-\nmal cells (e.g. TGFB2, BMP5 58, FMOD59), myeloid cells (Fibrillin, modulating TGF-ß \nbioviability60,61), and ERyMK progenitors ( TGFB1).  TGF-ß2 regulates cell growth, migration, \nand differentiation during embryogenesis. As TGF-ß signaling also regulates a wide range of \nbiological processes in HSCs, 85 TGF-ß pathways might influence hematopoiesis and eryth-\nropoiesis inside hemanoids. TGF- β 1 production by megakaryocytes, with supportive effects \non terminal RBC differentiation, has been reported.86,87  \nFuture Perspectives. In this study, we explored the enormous potential of iPSCs as a \nsource for in vitro modeling of human hematopoietic and erythroid development. We demon-\nstrated that hemanoids reflect human YS-der ived extraembryonic erythropoiesis, spanning \nthe undifferentiated iPSC to the enucleated RBC stage, and provided insight into tissue or-\nganization that might affect RBC development. Since extraembryonic hematopoiesis of hu-\nman origin is not available for repeated experiments due to ethical concerns and physical \ninaccessibility of embryonic material, reproducib le culture systems provide an important tool \nfor studying the earliest physiological stages of hematopoiesis. Unlike other established sys-\ntems, the hemanoid system not only produces er ythroid precursors but also generates enu-\ncleated RBCs in sufficient quantities for further f unctional studies. Therefore, it fills a gap by \noffering insight into the structure and function of the earliest embryonic RBCs. This includes \ntheir membrane composition, blood group antigen expression, biomechanical properties, and \noxygen-binding capacities.  Such analyses are crucial for evaluating the potential of iPSC-\nderived cRBCs for clinical applications in transfusion medicine. While cRBCs closely resem-\nble native cells, observed minor differences ma y arise from their different developmental \norigins rather than cultural conditions. The characteristics of iPSC-derived RBCs might be \nadvantageous for transfusions in preterm infant s, as the oxygen-binding capacities of em-\n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted May 8, 2026. ; https://doi.org/10.64898/2026.05.05.722134doi: bioRxiv preprint \n\n15 \nbryonic and fetal hemoglobin may protect them from cell-toxic effects caused by high oxygen \nlevels and free radical injury. 88 Because preterm infants often require only small RBC vol-\numes (less than 10 mL), producing sufficient amounts of cRBCs becomes feasible.  The suc-\ncessful expansion of 3D-hemanoids in a bioreactor has recently been demonstrated. 28,29,89 \nUtilizing autologous iPSCs can further advanc e personalized medicine approaches, includ-\ning CRISPR/Cas9-based genome editing to model or treat hematological disorders. Conse-\nquently, the hemanoid system may serve as a platform for future clinical translation to study \nRBC diseases or for de novo RBC production. Resources provided by our study, including \nthe CD43R-iPSC line, refined protocols for ST analysis of small spheroids, ultrastructural \nSTEM images, and ST data, will support future research in the vital field of developmental \nhematopoiesis.  \nLimitations of the study  \nOur study is limited by the resolution of the Visum 10X ST system. This constrains more \nprecise identification of cell clusters and the detection of rare events such as the emergence \nof HSCs. To obtain single-cell-level information and more accurately define cell populations \nand their characteristics, a scRNA-seq analysis  of the hemanoids is planned for a future \nstudy. The cell interaction mechanisms identified in this study need to be confirmed in further \nresearch and examined for their functional importance.  \n \nResource availability \nLEAD contact \nRequests for further information and resources should be directed to and will be fulfilled by \nthe lead contact, Isabel Dorn (isabel.dorn@medunigraz.at). \nMaterials availability \nThe CD43-GFP reporter iPSC line generated in th is study is available upon request from the \nlead contact with a completed materials transfer agreement. Usage must comply with the \nethics approval on which the patient's consent was based and exclude commercial use.  \nData and code availability \nThe ST data have been deposited in NCBI GEO as GSE324601 and are publicly available \nas of the date of publication. STEM images have been deposited in Dryad as [Dataset DOI: \n10.5061/dryad.69p8cz9hz] and are publicly available as of the date of publication.  \nThis paper analyzes existing, publicly available scRNA-seq data, accessible at E-MTAB-\n11673, GEO: GSE144024, GEO: GSE144024, and E-MTAB-7407. \n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted May 8, 2026. ; https://doi.org/10.64898/2026.05.05.722134doi: bioRxiv preprint \n\n16 \nThis paper does not report original code.  \nAny additional information required to reanalyze the data reported in this paper is available \nfrom the lead contact upon request.   \nAcknowledgments \nWe thank M. Sundl for assistance with immunohistochemistry and sample preparation for ST \nanalysis; K. Hingerl for assistance with STEM analysis; the Core Facility Bioimaging (M. \nAbsenger) for live-cell imaging support; the Core Facility Molecular Biology (B. Gallé and N. \nSchweintzger) for ST sample processing; S. Trajanoski for his input on ST data analysis; \nand E. van den Akker for providing Sani-003A-iPSCs. This research was funded by the Aus-\ntrian Science Fund (FWF), Grant-DOI 10.55776/I6572 to I.D. and 10.55776/PAT9611123 to \nG.M. Figures were created with BioRender. \n \nAuthor contributions \nI.D. and A.A. designed the study; M.A. and A.A. performed iPSC culturing and ex vivo eryth-\nropoiesis; A.A. generated the GFP reporter iPSC line; M.A., D.B., G.H., and J.F. performed \nimmunohistochemistry; D.K., M.A., and I.D. performed STEM analysis; A.A., M.A., G.M., and \nI.D. designed and performed Spatial Transcriptomi cs analyses; A.A. performed the bioinfor-\nmatics analysis; A.R., P.S., J.F. and I.D. monitored and supervised the study; C.B. and P.S. \nperformed project administration and provided resources; I.D. and A.A. analyzed and inter-\npreted the experiments, and wrote the original manuscript; all authors reviewed and edited \nthe final version of the manuscript.  \n \nDeclaration of interests: The authors declare no competing interests. \n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted May 8, 2026. ; https://doi.org/10.64898/2026.05.05.722134doi: bioRxiv preprint \n\n17 \nFigure titles and legends \nFigure 1. Hemanoid formation and erythroid differentiation of human iPSCs (see also \nFigure S1). (A) Erythroid differentiation through 3D-hemanoids. Left: Undifferentiated iPSCs; \nmiddle: self-organized Hemanoid; right: Electron microscopy image of hemanoid-derived \ncultured RBCs (scale bars: 200 µm, 750 µm, 10 µm).  (B) Illustration of the cell culture work-\nflow. Days -5 to 0: Induction of germ layer formation by embryoid body (EB) formation.  \nPhase I (days 0 – up to day 49):  Hematopoietic specification / hemanoid formation in \nAPEL™ medium supplemented with EPO, SCF, and IL-3. Starting around day 14, \nhemanoids continuously released CD43+ HCs into the supernatant. Phase II (+ 18 days): \nErythroid differentiation of cells harvested between days 14 and 49 from the hemanoid su-\npernatant (dashed red line) over an additional 18 days. ( C) Left and middle: Representative \nbrightfield images of a three-week-old hemanoid generated from PEB-AL#6 iPSCs, showing \nspherical structures, cell-dense areas with red islands, and an adherent stromal cell layer \n(Primovert Zeiss, 4x, scale bar: 500 µm). Right: Magnification of the rectangular area, repre-\nsenting red islands (black arrow), parts of the stromal layer (white arrow), covered by single \ncells released into the supernatant (scale bar: 100 µm).  \n \nFigure 2. Emerging hematopoiesis inside hemanoids generated from CD43-GFP re-\nporter iPSC lines (see also Figures S2 and S3) . (A) Representative time-lapse microscopy \nimages of a PEB-AL#6_CD43R iPSC-derived hemanoid expressing CD43 tagged with GFP. \nAfter 6 days and 15 hours in culture, the first GFP+ cells were detected (scale bar: 200 µm, \nmagnification 50 µm). (B and C) Representative fluorescence microscopy images obtained \nbetween weeks 2 and 4 of a hemanoid from CM1#1_CD43R iPSCs and PEB-AL#6_CD43R \niPSCs (scale bar: 500 µm). Red hemoglobin-pos itive areas are CD43-negative, consistent \nwith CD43 downregulation during terminal RBC maturation. ( D) Release of GFP+ single \ncells from a PEB-AL#6_CD43R hemanoid into the supernatant. Shown are microscopic \nmagnifications of the rectangular areas (scale bars: 500 µm, 300 µm, and 50 µm). (E  and F) \nSingle cells released from CM1#1_CD43R and PEB-AL#6_CD43R hemanoids were stained \nagainst CD43 (APC) to confirm endogenous CD43 expression of GFP+ cells. ( G) 4-weeks \nold hemanoids derived from PEB-AL#6_CD43R_iPSCs (top) and CM1#1_CD43R_iPSCs \n(bottom). GFP+ HCs demonstrate close contact with the stromal cell layer (scale bars: 750 \nµm and 300 µm).  \n \nFigure 3. Immunohistochemistry-based characterization of hemanoids. ( A) Repre-\nsentative images of two different hemanoids. Whit e arrows indicate blistered, fluid-filled are-\n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted May 8, 2026. ; https://doi.org/10.64898/2026.05.05.722134doi: bioRxiv preprint \n\n18 \nas (scale bar: 1mm). ( B) Brightfield images of two PEB-AL#6 hemanoids (3 and 4.5 weeks). \nBlack arrows indicate the stroma layer that attaches them to the tissue culture plate (scale \nbar: 750 µm). (C) Percentage of cells stained positive for hematopoietic cell surface markers \nafter enzymatic dissociation of hemanoids, and analyzed by flow cytometry (n=6; 3 biological \nreplicates; mean ± SD). ( D) HE-stained hemanoid section showing morphology of mesoder-\nmal (Mes) tissue with hematopoietic areas (red dashed line), ectodermal structures (Ect) \n(yellow dashed line), and endodermal glandular stru ctures (End, black dashed line) (scale \nbar: 200 µm). ( E and F) HE-stained tissue sections of two different hemanoids. Shown are \ncross-sections and magnifications, co-stained by  horseradish peroxidase for CD43 (hemato-\npoiesis), CD144 (VE-cadherin, endothelial cells), and vimentin (mesenchymal cells). In (E), \nCD43+ HCs are found within CD144+ vessel-like structures (black arrows), surrounded by \nvimentin+ mesenchymal-like cells. In (F), CD43+ HCs are found both inside (black arrows) \nand outside the CD144+ vessel, within a network of vimentin+ cells. ( G and H)  Brightfield \nimages of HE-stained hemanoids (100X oil, sca le bar: 20 µm), showing hematopoietic areas \ncontaining erythroid cells, myeloid cells, and thrombocytes. (G) HCs within the stromal net-\nwork, with a predominance of eosinophils. (H) small vessel (black arrows) containing, e.g., \neosinophils and platelets, surrounded by myeloid cells. ( a Mesenchymal cell; b Eosinophilic \ngranulocyte; c Neutrophilic granulocyte; d Erythroid precursor cell; e Mast cell; f Platelets; g \nMonocyte; h Myeloid precursor). See Figure S4  for confirmation of cell types by specific \nantibody staining. \n \nFigure 4. STEM analysis of a hemanoid (day 17) derived from PEB-AL#6 (see also Fig-\nure S5). (A) The identical hemanoid is shown i) as an adherent hemanoid in the dish, ii) after \nfixation, sectioning, and toluidine blue staining, and iii) analyzed by STEM (diameter: 812 \nµm). ( B) Magnifications from A, from left to right: Vessel containing HCs; HCs within the \nvessel; endothelial cell. ( C) Magnifications from B: Different HC types inside vessels (scale \nbar: 2 µm, platelets 1µm). ( D) Cells and fibers of the stromal compartment (scale bar: 2 µm, \ncollagen fibers 500 nm).  \n \nFigure 5. Spatial transcriptomics (ST) analysis and cluster annotation of day 16 \nhemanoids (see also Figures S6-S8) . (A) Schematic overview of the ST workflow. ( B) Uni-\nform Manifold Approximation and Projection  (UMAP) visualization of ST-sequencing data. \nColors indicate the 10 gene-expression-based clusters. ( C) Percentage of spots contributing \nto each of the 10 identified clusters. (D) Dot plot showing the expression frequency (dot size) \nand the expression level (color intensity) of two marker genes in each cluster. ( E) Dot plot \nshowing the expression of canonical marker genes for the hematopoietic lineage (Gran \n(Granulocytes), M ϕ (Makrophages), Mast (Mast cells), Lymph (Lymphoid cells), MK \n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted May 8, 2026. ; https://doi.org/10.64898/2026.05.05.722134doi: bioRxiv preprint \n\n19 \n(Megakaryocytes), Hemat (Hematopoietic cells). ( F) Volcano plots showing upregulated and \ndownregulated genes (log2-fold change) in the EryMK, Myeloid, and Erythroblast clusters \ncompared to mean expression in all other clusters.  \n \nFigure 6. Developmental wave of hemanoid-derived hematopoiesis. (A) Dot plot show-\ning the expression of genes encoding the alpha and beta chains of embryonic, fetal, and \nadult hemoglobin in day 16 hemanoids. ( B) Globin gene expression overlay on Visium spots \nfrom a section of day 16 hemanoids (Loupe br owser projection, log2 transformed UMI \ncounts). ( C) Expression of signature genes for AGM-derived HSCs\n9 and of core erythroid \ntranscription factors 8 in day 16 hemanoids (see also Figure S9). ( D) Expression of genes \nrelated to WNT signaling, Notch signaling, and HOX gene clusters in day 16 hemanoids. ( E)  \nUMAP visualization of integrated ST data sets from day 16 and day 28 hemanoids, using \nHarmony (see Figures S10C and S11 for day 28 ST analysis). ( F) Expression of cell-type-\nspecific marker genes in the integrated day 16 (pink dots) and day 28 (green dots) data sets \nfrom (E): Erythroid (Ery), megakaryocytes (MK), hematopoietic (Hemato), hemogenic endo-\nthelium (HE), lymphoid (Lympho), neuro progenitors (Neuro), Stroma, endodermal (Endo). \nDot size represents gene-expression frequency , and color intensity indicates expression \nlevels. \n \nFigure 7. Cellular interactions of HCs (ST analysis of day 16 hemanoids) (see also Fig-\nures S12 and S13). (A) HE-stained hemanoid section on the ST slide showing terminal ma-\ntured RBCs (black arrows) (scale bar: 50 µm). ( B) Morphology of hepatoblasts (left, dotted \nlines) located near blood islands. Hepatoblast-specific marker gene expression overlay on \nVisium spots (Loupe browser projection). ( C) Volcano plot showing upregulated and \ndownregulated genes (log2 fold change) in the Hepatoblast cluster. ( D) Dot plot showing the \nexpression of genes encoding growth factors. ( E) Expression of genes encoding ECM com-\nponents. ( F) Expression of genes encoding adhesion molecules. ( G) Graphical illustration \nsummarizing day 16 ST results regarding the ex pression of growth factors (bold), adhesion \nmolecules, and ECM (Italic) in the EryMK, Myeloid, Hepatoblast, and the Stroma cluster. \n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted May 8, 2026. ; https://doi.org/10.64898/2026.05.05.722134doi: bioRxiv preprint \n\n20 \nMethods \nHuman material and cell lines \nThe study was approved by the local ethics commi ttee at the Medical University of Graz in \nline with the Declaration of Helsinki (27-165ex14/15). Three human iPSC lines derived from \nerythroblasts were used (UBTi001-A (PEB-AL#6) 30; CM1 32, Sani-003A 31) and cultured at \n37°C and 5% CO2 in Stem MACS iPS Brew XF medium (#130-104-368, Miltenyi Biotech) on \n6-well tissue culture plates coated with Matr igel® (#354277, Corning). Cells were mechani-\ncally split every 6-7 days and supplemented with 10 µM Rock Inhibitor (#130-106-538, \nMiltenyi Biotech). Human K-562 cells (ACC-10, DSMZ) were cultured in RPMI-1640 medium \n(#11875093, Gibco) supplemented with 10% fetal bovine serum (FBS) (#S0615, Biochrom) \nand 1% Penicillin Streptomycin (PS) (#15070063, Gibco). HEK293T cells (#300189, CLS) \nwere maintained in DMEM high-glucose (#D5671, Sigma-Aldrich) with 10% FBS, 1% PS, \nand 25 mM HEPES (#15630056, Gibco). \nHemanoid formation and erythroid differentiation of iPSCs \nHematopoietic and erythroid differentiation of iPSCs was performed as recently described \n22,23 and illustrated in Figure 1. For EB generation, iPSC colonies were detached from the \ntissue culture well using 1mg/mL collagenase type IV (#17104019, Gibco). Cell aggregates \nwere seeded on ultra-low-binding plates (#15277905, Nunclon Sphera, Thermo Scientific) \nand cultivated for 5 days in hESC medium without bFGF.\n90 Thereafter, spherical EBs were \ntransferred onto six-well tissue culture plates in STEMdiff™ APEL™ 2 medium (#5270, \nStemCell Technologies), supplemented with 5% Protein-Free Hybridoma Medium (#12040-\n077, ThermoFisher Scientific), 100 ng/mL SCF (#300-07, Peprotech), 5 ng/mL IL-3 (#200-\n03, Peprotech), and 3 U/mL EPO (Erypo, Janssen Biologics B.V.). The medium was \nchanged weekly. Within a few days, EBs adhered to the tissue culture plate and formed self-\norganized 3D structures, termed hemanoids. The size and the morphology of the hemanoids \nwere assessed by microscopy (EVOS M5000, ThermoFisher Scientific). Hemanoids were \ncharacterized at different maturation stages  by flow cytometry and immunohistochemistry. \nFor flow cytometry characterization, hemanoids were digested into a single-cell suspension \nusing 0.4 IU/mL Collagenase B (#11088815, Roche) (2 h at 37°C, 5% CO2) and Cell disso-\nciation buffer (#13151014, Gibco) (10 min at RT) as described.\n91  Single cells released from \nhemanoids into the supernatant were repeatedly harvested to determine cell counts, charac-\nterize them by flow cytometry, and assess hematopoietic colony formation in semisolid me-\ndia.  \nFor erythroid differentiation, released single cells were repeatedly harvested from the super-\nnatant and cultured for 18 days in an established erythroid differentiation protocol.\n33 Culture \n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted May 8, 2026. ; https://doi.org/10.64898/2026.05.05.722134doi: bioRxiv preprint \n\n21 \nmedium consisted of Iscove’s Modified Dulbecco’s Medium (IMDM) (#FG0465, Biochrom) \ncontaining 5% Octaplas® LG (Octapharma), 10 µg/mL insulin (#91077C, Sigma-Aldrich), 330 \nµg/mL transferrin (#T101-5, BBI Solutions), 1% PS, and from day 8 onwards 4mg/dL choles-\nterol-rich lipids (#L4646, Sigma-Aldrich). Cells were stimulated as follows: Day 0 to day 8: \n100 ng/mL SCF, 5 ng/mL IL-3, 3 U/mL EPO, and 10\n-6 M hydrocortisone (OHC) (#H2270, \nSigma-Aldrich); Day 8 to day 11: 100ng/mL SCF and 3 U/mL EPO; day 11 to day 21: 3 U \n/mL EPO. Cell numbers and cell vitality were counted in a Malassez counting chamber after \nstaining with trypan blue (#T8154, Sigma-Aldric h). Hematopoietic and erythroid differentia-\ntion were monitored by flow cytometry and microscopic evaluation of cytospin preparations \nafter staining with May-Gruenwald-Giemsa (#102103, Hemafix, Biomed) and neutral \nbenzidine (#D9143, o-Dianisidine, Sigma-Aldrich) for the detection of hemoglobin. At least \n300 cells were enumerated under the microscope (Axioscope, Zeiss). In some experiments, \nday 18 cells were filtered through an Acrodisc WBC syringe filter (#AP-4851, Pall Corpora-\ntion) to obtain the pure enucleated portion of cultured RBCs.  \nFlow cytometry \nFlow cytometry analysis was performed on a CytoFLEX\n® flow cytometer (Beckman Coulter) \nusing the CytExpert software 2.4. The following antibodies were used to stain the cells \nthroughout hematopoietic and erythroid differentiation:  CD34-PE (#A07776, Beckman Coul-\nter), CD43-APC (#560198, BD Biosciences), CD45-PC7 (#IM3548, Beckman Coulter,), \nCD45-FITC (#IM3454808, BD Biosciences), CD36-FITC (#B49201, Beckman Coulter), \nCD235a-FITC (#B49206, Beckman Coulter), CD49d-APC (#B01682, Beckman Coulter), \nCD71-PE (#555537, BD Biosciences), and CD233 (Band3)-PE (#9439PE, IBGRL). Dead \ncells were excluded by 4',6-Diamidino-2-phenylindol (DAPI) (#D3571, ThermoFisher Scien-\ntific) staining. Graphs were partially generated using FlowJo (TreeStar, v10). \nColony formation  \nSingle cells released from the hemanoid into the supernatant were collected and plated in \ntriplicate (2,500 cells/dish) on MethoCult (#H84434, StemCell Technologies) coated 35mm \ndishes (#27100, StemCell Technologies). After 10-14 days, the dishes were scored using \nlight microscopy (Primovert; Zeiss). Colonies we re classified into burst-forming unit-erythroid \n(BFU-E), colony-forming unit-erythroid (CFU-E), colony-forming unit-\ngranulocyte/erythrocyte/monocyte/megakaryocyte (CFU-GEMM), colony-forming unit-\nmacrophage (CFU-M), and colony-forming unit-granulocyte/macrophage (CFU-GM). \n \n \n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted May 8, 2026. ; https://doi.org/10.64898/2026.05.05.722134doi: bioRxiv preprint \n\n22 \nImmunohistochemistry   \nHemanoids were washed 2x with Dulbecco’s Phosphate Buffered Saline (DPBS) (#14190-\n094, Gibco™), fixed for 1 hour in 4% paraformaldehyde (#1.04005.1000, Merck Millipore), \nand embedded in paraffin with Excelsior™ (Thermo Fisher). Paraffin-embedded hemanoids \nwere sectioned (10µm sections) using a Thermo  ScientificTM rotation microtome HM355s \n(Fisher Scientific). Before staining, antigen retrieval was performed in a microwave for 2 x \n20-minute cycles in citrate buffer (pH 6). UltraVison™ Quanto Detection System HRP (#TL-\n015-QHD, Epredia™) was used for antibody det ection using the Horseradish peroxidase \n(HRP) system. Briefly, after being washed 3x with PBS, the slides were incubated for 10 \nminutes with UltraVision™ Hydrogen Peroxidas e Block, washed 3x again with PBS, and \nincubated for 5 minutes with UltraVision™ Protein. Primary antibodies, rabbit anti-human \nCD43  (#MA5-16339, ThermoFisher Scientific), recombinant rabbit anti-human CD144 (VE-\nCadherin) (#MAB-44374, ThermoFisher Scientific), rabbit anti-human HBE1 (#PA5-106357, \nThermoFisher Scientific), rabbit anti-human Laminin beta-1 (#PA5-27271, Thermo Fisher), \nmouse anti-human vimentin (#MAB3400, Merck, Millipore), rabbit anti-human AFP \n(#145501-AP, Proteintech), mouse anti-human HBG1 (#66168-1-Ig, Proteintech), mouse \nanti-human CD68 (#14-0688-82, Thermo Fisher Scientific), mouse anti-human CD163 (#DB \n045, DB Biotech), mouse anti-human CD14 (#60253, PTGlab, Proteintech), mouse anti-\nhuman 235a (#M0819, Dako, rabbit anti-human EPX (#ab238506, Abcam), rabbit anti-\nhuman myeloperoxidase (#GA511, Dako), mouse anti-human mast cell tryptase (#IR640, \nDako), mouse anti-human Integrin beta 3 (#ab9509, Abcam), mouse anti-human CD20 \n(#GA604, Dako), rabbit anti-human CD3 (#GA503, Dako), mouse anti-human CK7 (#MS-\n1352-P, ThermoFisher Scientific),  mouse anti-human HLA-G (#557577; BD Biosciences) \nwere diluted in antibody diluent (#TA-125-ADQ, Epredia™) according to manufacturer’s in-\nstructions and the slides were incubated for 60 minutes. Thereafter, the slides were incubat-\ned for 10 minutes with Primary Antibody Amplifier Quanto, washed again 3x, and incubated \nfor 10 minutes in the dark with HRP Polymer Quanto (light sensitive). Coloring was per-\nformed for 5 minutes using a mixture of DAB Quanto Chromogen and Substrate (one drop of \nchromogen in 1 ml substrate) (brown color) or for 10 minutes using 4 drops of AEC Sub-\nstrate (red color) (#ab64252, Abcam). Counterstaining to identify the tissue morphology was \nperformed using modified hematoxylin (H&E, #8947.1, Roth) for 2 minutes. The whole stain-\ning process was performed in a humidified chamber at RT. IHC slides were scanned using a \ndigital slide scanner (Slideview VS200, Ol ympus, Tokyo, Japan) equipped with an LED \nsource (Excelitas Technologies, X-Cite Xy lis, Mississauga, Canada) and a CMOS camera \n(2304\n/i3 ×/i3 2304, ORCA-Fusion C14440-20UP, 16-bit, Hamamatsu, Japan). Analysis soft-\nware for scanned slides was Olympus OlyVIA 3.4.1.  \n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted May 8, 2026. ; https://doi.org/10.64898/2026.05.05.722134doi: bioRxiv preprint \n\n23 \n \nScanning Transmission Electron Microscopy (STEM) \nHemanoids were fixed in in 0.1M cacodylate buffer (2.14% w/v Dimethylarsinic acid sodium \nsalt trihydrate (#820670, Sigmal-Aldrich) in Aqua bidest, pH 7.4) supplemented with 2.5% \nglutaraldehyde (#16200, Electron Microscopy Sciences) and 2% paraformaldehyde \n(#1.04005.1000, Merck Millipore) for 3h, and then post-fixed in 2% osmium tetroxide \n(#19110, Electron Microscopy Sciences) for 2h at RT. After dehydration in a graded series of \nethanol (50%-100%), tissues were infiltrated in propylene oxide (#149620010, ThermoFisher \nScientific) for 1h, followed by stepwise infiltration in TAAB Embedding Resin (TER) (#T004, \nTAAB Laboratories Equipment Ltd., UK): 50% v/v TER in propylene oxide for 3h at RT, fol-\nlowed by 66% v/v TER (overnight, 4°C), and finally pure TER (3 h, 45°C). Embedded tissues \nwere transferred to embedding molds (#10590, PELCO) and polymerized for 48h at 60°C. \nSemithin sections (1µm) were cut with glass knives (#7890-04, Leica Microsystems) and \nstained with Toluidin blue solution (1% w/v Dinatriumtetraborate (Sigma-Aldrich, #106306) \nand 1% w/v Toluidine blue (#R1727, Agar Scientific) in Aqua Bidest). Slides were assessed \nusing a BX41 light microscope (Olympus). Ultrathin sections (70 nm) were cut with a UC 7 \nUltramicrotome (Leica Microsystems, Austria) and a diamond knife (#2302, Diamond Knife \nDiATOME  Sciences Services), placed on piol oform-covered grids (#R1275 Agar Scientific) \n(#G200H-Cu and G2010Cu, Sciences Services), and stained with 1% platinum blue (EMS, \nUSA, #22407) for 15 min and 3% lead citrate (#16707235, Leica Microsystems) for 5 min. \nElectron micrographs were taken using a Tecnai G2 transmission electron microscope \n(Thermo Fisher Scientific, Netherlands) wi th a Gatan Ultrascan 1000 charge-coupled device \n(CCD) camera (-20°C; acquisition software: Digital Micrograph, Ametek Gatan, Germany; \nand Serial EM). The acceleration voltage was 120 kV. To image large areas of hemanoids at \nhigh resolution, scanning transmission electron microscopy (STEM) imaging mode on a \nfield-emission scanning electron microscope (ZEISS FE-SEM Sigma 500) with an accelera-\ntion voltage of 15 kV, in combination with ATLAS TM (version 5.2.2.15, ZEISS), was used.   \nPlasmid-AAV design and cloning  \nThe AAV vector plasmid was cloned into the pAAV-MCS plasmid (#240071, Agilent Tech-\nnologies) containing inverted terminal repeats from AAV serotype 2 (AAV2), with a maximal \npacking capacity of 4,7 kb. The donor plasmid was assembled by standard Gibson assembly \n(Table S1) of the NotI HF (#R3189S, NEB) linearized plasmid backbone using NEBuilder® \nHiFi DNA Assembly Mastermix (#E2621L, NEB). The constructed plasmid contains the right \n(RHA) and left homology arms (LHA), each 300 bp, homologous to the DNA flanking the \nspCas9 cut site ( AAVS1_sgRNA: GGGGCCACUAGGGACAGGAU ), 5’ splice acceptor to \n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted May 8, 2026. ; https://doi.org/10.64898/2026.05.05.722134doi: bioRxiv preprint \n\n24 \nensure the exact splicing after transcription, T2A a self-cleaving peptide, puromycin-resistant \ncassette with bovine growth hormone (bGH) polyadenylation signal, followed by a long  \n(2179 bp) CD43 promotor region which drives the expression of the fused green fluorescent \nprotein (GFP) and the SV40 polyadenylation signal.  \nAAV production \nHuman embryonic kidney 293T cells (HEK293T cells) were expanded to a total of at least \n130 million cells in DMEM high glucose (Sigma-Aldrich) supplemented with 4 mM L-\nglutamine (#G7513, Sigma-Aldrich), 1mM S odium pyruvate (#11360070, ThermoFisher), \n10% FBS (Biochrom), 1% PS (Gibco), 25 mM HEPES (Gibco), and 1 mM Sodium butyrate \n(#B5887, Sigma-Aldrich). 13 million cells were seeded per individual 15 cm dish one day \nbefore transfection. At about 70–80% confluency, cells were transfected using 5 µg/mL \npolyethyleneimine (#23966, Polysciences). For transfection of a total of 10 plates, 60 μ g \nAAV donor plasmid (pAAV6_AAVS1_CD43_GFP) and 220 μ g helper plasmid pDGM692 (was \na gift from David Russell, #110660, Addgene) were mixed with PEI in Opti-MEM (#3798570, \nGibco). The mixture was incubated at RT for 15 minutes, then added dropwise to the media \nand carefully swirled. Cells were incubated in a humidified 37°C incubator for 72 hours. \nAAV6 viral particles were harvested and purified using the AAVpro Purification Kit (#6666, \nTakara) following the manufacturer’s instructions. Viral particles were stored in aliquots at -\n80°C till further use. The copy number/µl was determined by ddPCR (QX200, Biorad). \nElectroporation of iPSCs \nAfter digestion with Accutase (#T8154, Sigma-Aldrich), the iPSC single-cell suspension was \nelectroporated using the Lonza 4D Nucleofector (program CA-137) and the P3 Primary Cell \nNucleofection Kit (#V4XP-3024, Lonza). We have electroporated as few as 300,000 cells per \ncondition using the electropor ation strips holding 20 µl. The RNP complex (Cas9 + sgRNA \nmixed in a 1:2.5 molar ratio, Cas9 Nuclease V3, #1081059, Gibco) was prepared at 25°C for \n15 minutes before electroporation and scaled down based on the number of electroporated \ncells. After electroporation, the cells were  transduced with 5,000 – 10,000 AAV vector ge-\nnomes/cell and incubated at 37°C and 5% CO\n2. Cells were seeded as single cells into six-\nwell plates containing pre-warmed antibiotic-f ree XF media (Miltenyi). After 48h of incuba-\ntion, small colonies were scored under the mi croscope. Puromycin selection was initiated \nafter cells reached about 40-50% confluency. Puromycin (#A11138-03, Gibco) was used at \nconcentrations ranging from 0.1 µg/mL to 0.5 µg/mL, depending on the iPSC line. \nPuromycin-resistant cells were picked and cl onally expanded before further approval of suc-\ncessful gene targeting.  \n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted May 8, 2026. ; https://doi.org/10.64898/2026.05.05.722134doi: bioRxiv preprint \n\n25 \nCharacterization of iPSCs edited clones \nGenomic DNA was purified from the clones using QuickExtract ™  DNA Extraction Solution \n(#QE09050, Epicentre). Briefly, the mixture of cells and extract solution 25 µl was vortexed \nfor 15 seconds, incubated at 65°C for 6 minutes, vortexed for 15 seconds, and incubated \nagain at 98°C for 2 minutes. The DNA was amplified using an in-out PCR (one primer within \nthe introduced DNA sequence, the other primer outside of the homology arms) ( Table S2). \nThe amplicon was Sanger-sequenced (Eurofins Genomics) to confirm the knock-in at the \nAAVS1 harbor locus. \nLive cell imaging and immunofluorescence microscopy \nThe emergence of hematopoietic CD43-GFP+ cells within adherent hemaloids was ob-\nserved by live-cell imaging using a Nikon HCS Ti2 Eclipse (Celesta V2) microscope. \nHemanoids were analyzed at different time points. Image acquisition settings were optimized \nfor GFP detection. The time frame was set to one image every 10, 30, or 60 minutes for 24 \nor 72 hours. Samples were imaged with a 20x objective (NA 0.75), and excitation was pro-\nvided by a 488 nm laser. Emission was collected using MXR10018 1\nst 4000 KG Phometrics \nBSI Express CMOS camera/PMT. Images were acquired using NIS Elements C software \nwith Jobs (v5.42.07). Additional observation of CD43-GFP+ cells was done with an EVOS \nM5000 microscope (ThermoFisher Scientific). \nSpatial transcriptomics sample processing and sequencing \n10 to 18 hemanoids generated from the PEB-AL#6 iPSC line were pooled on days 16 and \n28, respectively (step II). The hemanoids were washed 2x with PBS (Gibco), fixed for 60 \nminutes using 4% paraformaldehyde, an d embedded in paraffin using Excelsior™ \n(ThermoFisher Scientific). For spatial transcriptomics, the hemanoids after RNA quality as-\nsessment (DV200) were cut into 5 µm thick sections using a rotation microtome HM355s \n(ThermoFisher), and mounted within the capture areas of the Visum Spatial Gene Expres-\nsion Slide (#PN-1000189, 10X Genomics). The slide was dried at 40°C on a heat plate. Tis-\nsue deparaffinization, H&E staining, imaging, and decrosslinking were performed according \nto the 10X Visium Spatial Gene Expression for FFPE guideline (10X Genomics, \nDeparaffinization, H&E Staining, Imaging & Decrosslinking, CG000409 Rev C). H&E images \nwere taken using a Leica Aperio ScanScope\n®  AT imaging system at x40 magnification and \nthe Aperio ImageScope software (v12.4.6.5003). Probe extension and library construction \nsteps using the Visium FFPE Reagent kit (PN-1000361) and the Visum Human \nTranscriptome Probe kit (#PN1000363) followed the 10X user guide Visium Spatial Gene \nExpression Reagent Kits for FFPE (CG000407 | Rev D). Tissue slides had an average \nDV200 of 28%. The coverage area was estimated with the Loupe Browser v7 (10X Ge-\n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted May 8, 2026. ; https://doi.org/10.64898/2026.05.05.722134doi: bioRxiv preprint \n\n26 \nnomics, Pleasanton, California, U.S.). Sequencing was performed with the recommended \nread mode: read 1: 28 cycles; i7 index read: 10 cycles; i5 index read: 10 cycles; and read 2: \n50 cycles on an Illumina NextSeq2000. \nProcessing of spatial RNA sequencing reads \nAfter sequencing, the reads were aligned to the human genome (hg38) using the Space \nRanger pipeline (v4.0.1, 10X Genomics) with default parameters. Space Ranger was also \nused to align paired histology images with the positions of mRNA capture spots on the \nVisium slides. The raw UMI count matrix, images, spot-image coordinates, and scale factors \nwere imported to the Seurat R package (v5.1.0) 93 for downstream data processing. In brief, \nwe first performed quality control (QC) to remove low-quality spots based on metrics includ-\ning total UMI counts, the number of detected genes, and the percentage of mitochondrial \ngene expression. Spots with unusually low or high gene counts, low UMI counts, or high \nmitochondrial content were excluded from further analysis. Following QC, we used \nSCTransform to normalize and scale the data and identify variable genes. Dimensionality \nreduction was performed using RunPCA, and the first 20 principal components were used \nfor downstream analyses. We applied FindNeighbors to these components, followed by \nFindClusters to cluster the ST spots at a resolution of 1.0. Finally, we used RunUMAP on the \nsame 20 principal components to visualize the data in two dimensions. Differentially ex-\npressed genes (DEGs) for each cluster were  identified using the FindAllMarkers or \nFindMarkers functions in Seurat with defaul t parameters, comparing gene expression within \neach cluster to all remaining clusters. DEG analysis was performed using a pairwise Wilcox-\non Rank-Sum test between spots within each cluster and all other spots in the dataset. The \nDotPlot function was used to illustrate the expression pattern of selected genes for different \ncell types or conditions.  \nCell type annotation and label transfer were performed using the Azimuth package v0.5.0\n94 \nwith the fetal development reference data 42, followed by manual annotation using known \nmarker genes for clusters that showed additional diversity in gene signatures. \nWe used Harmony v1.2.0\n55 to integrate and batch-correct the ST data of FFPE samples \n(days 16 and 28) with published scRNA-seq datasets. Namely, the scRNA-seq data from \nyolk sac (E-MTAB-11673 54, and GEO: GSE144024 53), and fetal liver (GEO: GSE144024 53, \nand E-MTAB-740711). Prior to integration, each dataset was individually preprocessed using \nthe Seurat pipeline, including normalization, identification of highly variable features, and \nscaling. After quality control and preprocessing, the datasets were merged into a single Seu-\nrat object. Following the merge, we repeated the standard analysis steps: we performed \nprincipal component analysis (PCA) with RunPCA, constructed a shared nearest-neighbor \ngraph using FindNeighbors, identified clusters using FindClusters, and visualized the inte-\n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted May 8, 2026. ; https://doi.org/10.64898/2026.05.05.722134doi: bioRxiv preprint \n\n27 \ngrated data using RunUMAP. Gene ontology (GO) enrichment analysis for the selected cell \nclusters was performed in R, using the enric hGO function from the clusterProfiler v4.6.2 95 \npackage and the “org.Hs.eg.db” database. G ene set enrichment analysis (GSEA) was per-\nformed on gene lists identified by the FindMarkers  function as statistically significant. The \ngseGO function from the clusterProfiler package v4.6.2 95 with default parameters was used. \nThe selected pathways were visualized using the R package ggplot2 v4.0.1. 96 For compari-\nson, we have also utilized the fgsea v1.32.497 package in R. \nStatistics \nData are presented as mean ± standard devia tion (SD) unless otherwise stated. Raw data \nwere tested for normality of distribution, and statistical analyses were performed using a two-\ntailed unpaired t-test, a two-tailed Mann–Whitney test, a Wilcoxon rank sum test, a two-way \nANOVA for the line graphs, and a Kruskal–Wallis test with multiple comparison tests, de-\npending on the dataset. GraphPad Prism 10.4.1 (GraphPad Software, San Diego, CA, USA) \nor R (R-Studio, R 4.2.0) was used for statistical analyses. \n \nSupplemental information \nThe Supplemental information (Document S1) includes Figures S1 - S13, Tables S1 and S2, \nand supplemental references.  \n \n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted May 8, 2026. ; https://doi.org/10.64898/2026.05.05.722134doi: bioRxiv preprint \n\n28 \nReferences \n1. Peyrard T, Bardiaux L, Krause C, Kobari L, Lapillonne H, Andreu G, Douay L. 2011 \nJul; Banking of pluripotent adult stem ce lls as an unlimited source for red blood cell  \nproduction: potential applications for alloimmunized patients and rare blood \nchallenges. Transfus Med Rev. 25(3):206–16. doi:10.1016/j.tmrv.2011.01.002 \n2. Gunawardena N, Chou ST. 2025; Generation of red blood cells from induced \npluripotent stem cells. Curr Opin Hematol. 31(16):115–21. \ndoi:10.1097/MOH.0000000000000810. \n3. Ditadi A, Sturgeon CM, Keller G. 2016; A view of human haematopoietic development \nfrom the Petri dish. Nat Rev Mol Cell Biol. 18(1):56–67. doi:10.1038/nrm.2016.127 \n4. Palis J. 2024 Aug; Erythropoiesis in the mammalian embryo. Exp Hematol. \n136:104283. doi:10.1016/j.exphem.2024.104283 \n5. Ivanovs A, Rybtsov S, Ng ES, Stanley EG, Elefanty AG, Medvinsky A. 2017 Jul 1; \nHuman haematopoietic stem cell development: from the embryo to the dish. \nDevelopment. 144(13):2323–37. doi:10.1242/dev.134866 \n6. Tavian M, Coulombel L, Luton D, Clemente HS, Dieterlen-Lièvre F, P έ ault B. 1996; \nAorta-Associated CD34+ Hematopoietic Cells in the Early Human Embryo. Blood. \n87(1):67–72. doi:https://doi.org/10.1182/blood.V87.1.67.67 \n7. Tavian M, Hallais MF, Péault B. 1999 Feb; Emergence of intraembryonic \nhematopoietic precursors in the pre-liver human  embryo. Development. 126(4):793–\n803. doi:10.1242/dev.126.4.793 \n8. Palis J. 2014; Primitive and definitive erythropoiesis in mammals. Front Physiol. 5 \nJAN(January):1–9. doi:10.3389/fphys.2014.00003 \n9. Calvanese V, Capellera-Garcia S, Ma F, Fares I, Liebscher S, Ng ES, Ekstrand S, \nAguadé-Gorgorió J, Vavilina A, Lefaudeux D, et al. 2022; Mapping human \nhaematopoietic stem cells from haemogenic endothelium to birth. Nature. \n604(7906):534–40. doi:10.1038/s41586-022-04571-x \n10. Goh I, Botting RA, Rose A, Webb S, Engelbert J, Gitton Y, Stephenson E, Londoño \nMQ, Mather M, Mende N, et al. 2023; Yolk sa c cell atlas reveals multiorgan functions \nduring human early development. Science (80- ). 381(6659). \ndoi:10.1126/science.add7564 \n11. Popescu DM, Botting RA, Stephenson E, Green K, Webb S, Jardine L, Calderbank \nEF, Polanski K, Goh I, Efremova M, et al. 2019. Decoding human fetal liver \nhaematopoiesis. Vol. 574, Nature. 365-371 p. doi:10.1038/s41586-019-1652-y \n12. Dou DR, Calvanese V, Sierra MI, Nguyen AT, Minasian A, Saarikoski P, Sasidharan \nR, Ramirez CM, Zack JA, Crooks GM, et al. 2016; Medial HOXA genes demarcate \nhaematopoietic stem cell fate during human development. Nat Cell Biol. 18(6):595–\n606. doi:10.1038/ncb3354 \n13. Merryweather-Clarke AT, Tipping AJ, Lamikanra AA, Fa R, Abu-Jamous B, Tsang \nHP, Carpenter L, Robson KJH, Nandi AK, Roberts DJ. 2016; Distinct gene expression \nprogram dynamics during erythr opoiesis from human induced pluripotent stem cells \ncompared with adult and cord blood progenitors. BMC Genomics. 17(1):1–20. \ndoi:10.1186/s12864-016-3134-z \n14. Focosi D, Amabile G. 2018; Induced Pluripotent Stem Cell-Derived Red Blood Cells \nand Platelet Concentrates: From Bench to Bedside. Cells. 7(1):2. \ndoi:10.3390/cells7010002 \n15. Lee E, Sivalingam J, Lim ZR, Chia G, Shi LG, Roberts M, Loh YH, Reuveny S, Oh \nSKW. 2018 Dec 1; Review: In vitro generation of red blood cells for transfusion \nmedicine: Progress, prospects and ch allenges. Biotechnol Adv. 36(8):2118–28. \ndoi:10.1016/J.BIOTECHADV.2018.09.006 \n16. Lancaster MA, Knoblich JA. 2014 Jul; Organogenesis in a dish: modeling \ndevelopment and disease using organoid  technologies. Science. 345(6194):1247125. \ndoi:10.1126/science.1247125 \n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted May 8, 2026. ; https://doi.org/10.64898/2026.05.05.722134doi: bioRxiv preprint \n\n29 \n17. Artegiani B, Hendriks D. 2025 Feb; Organoids from pluripotent stem cells and human \ntissues: When two cultures meet  each other. Dev Cell. 60(4):493–511. \ndoi:10.1016/j.devcel.2025.01.005 \n18. Hofer M, Lutolf MP. 2021; Engineering organoids. Nat Rev Mater. 6(5):402–20. \ndoi:10.1038/s41578-021-00279-y \n19. Dardano M, Kleemiß F, Kosanke M, Lang D, Wilson L, Franke A, Teske J, Shivaraj A, \nde la Roche J, Fischer M, et al. 2024; Blood-Generating Heart-Forming Organoids \nrecapitulate co-development of the human hematopoietic system and the embryonic \nhearto Title. Nat Cell Biol. doi:10.1038/s41556-024-01526-4 \n20. Richard C, Drevon C, Canto PY, Villain G, Bollérot K, Lempereur A, Teillet MA, \nVincent C, Rosselló Castillo C, Torres M, et al. 2013 Mar; Endothelio-mesenchymal \ninteraction controls runx1 expression and modulates the  notch pathway to initiate \naortic hematopoiesis. Dev Cell. 24(6):600–11. doi:10.1016/j.devcel.2013.02.011 \n21. Frenz-Wiessner S, Fairley SD, Buser M, Goek I, Salewskij K, Jonsson G, Illig D, Zu \nPutlitz B, Petersheim D, Li Y, et al. 2024; nature methods Generation of complex \nbone marrow organoids from human induced pluripotent stem cells. Nat Methods |. \n21:868–81. doi:10.1038/s41592-024-02172-2 \n22. Bernecker C, Ackermann M, Lachmann N, Rohrhofer L, Zaehres H, Araúzo-Bravo \nMJ, van den Akker E, Schlenke P, Dorn I. 2019 Dec 1; Enhanced Ex Vivo Generation \nof Erythroid Cells from Human Induced Pluripotent Stem Cells in a Simplified Cell \nCulture System with Low Cytokine Support. Stem Cells Dev. 28(23):1540–51. \ndoi:10.1089/scd.2019.0132 \n23. Bernecker C, Matzhold EM, Kolb D, Avdili A, Rohrhofer L, Lampl A, Trötzmüller M, \nSinger H, Oldenburg J, Schlenke P, et al. 2022; Membrane Properties of Human \nInduced Pluripotent Stem Cell-Derived Cu ltured Red Blood Cells. Cells. 11(16). \ndoi:10.3390/cells11162473 \n24. Malik J, Kim AR, Tyre KA, Cherukuri AR, Palis J. 2013 Nov; Erythropoietin critically \nregulates the terminal maturation of murine and human  primitive erythroblasts. \nHaematologica. 98(11):1778–87. doi:10.3324/haematol.2013.087361 \n25. Koury MJ, Bondurant MC. 1990 Apr; Erythropoietin retards DNA breakdown and \nprevents programmed death in erythroid  progenitor cells. Science. 248(4953):378–\n81. doi:10.1126/science.2326648 \n26. Goodman JW, Hall EA, Miller KL, Shinpock SG. 1985 May; Interleukin 3 promotes \nerythroid burst formation in “serum-free” cultures without  detectable erythropoietin. \nProc Natl Acad Sci U S A. 82(10):3291–5. doi:10.1073/pnas.82.10.3291 \n27. Sui X, Krantz SB, Zhao ZJ. 2000 Jul; Stem cell factor and erythropoietin inhibit \napoptosis of human erythroid  progenitor ce lls through different signalling pathways. \nBr J Haematol. 110(1):63–70. doi:10.1046/j.1365-2141.2000.02145.x \n28. Ackermann M, Kempf H, Hetzel M, Hesse C, Hashtchin AR, Brinkert K, Schott JW, \nHaake K, Kühnel MP, Glage S, et al. 2018 Nov; Bioreactor-based mass production of \nhuman iPSC-derived macrophages enables  immunotherapies against bacterial \nairway infections. Nat Commun. 9(1):5088. doi:10.1038/s41467-018-07570-7 \n29. Varga E, Brandsma E, Juarez-Garza BE, Ramlal RPE, Karrich JJ, Laurent A, Chavli \nA, Paskel R, Fu K, Flavell RA, et al. 2025 Oct; Large-Scale Production of Transfusion-\nReady Red Blood Cells From Induced  Pluripotent Stem Cells. Adv Sci (Weinheim, \nBaden-Wurttemberg, Ger. 12(38):e04725. doi:10.1002/advs.202504725 \n30. Avdili A, Rohrhofer L, Auer M, Bernecker C, Schlenke P, Dorn I. 2022; Generation of \nthe human erythroblast-derived iPSC line UBTi001-A. Stem Cell Res. \n64(August):102910. doi:10.1016/j.scr.2022.102910 \n31. Varga E, Hansen M, Wüst T, von Lindern M, van den Akker E. 2017 Dec; Generation \nof human erythroblast-derived iPSC line using episomal reprogramming  system. \nStem Cell Res. 25:30–3. doi:10.1016/j.scr.2017.10.001 \n32. Jamil MA, Singer H, Al-Rifai R, Nüsgen N, Rath M, Strauss S, Andreou I, Oldenburg \nJ, El-Maarri O. Molecular Analysis of Fetal and Adult Primary Human Liver Sinusoidal \nEndothelial Cells: A Comparison to Other Endothelial Cells. Int J Mol Sci Artic. \n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted May 8, 2026. ; https://doi.org/10.64898/2026.05.05.722134doi: bioRxiv preprint \n\n30 \ndoi:10.3390/ijms21207776 \n33. Bernecker C, Köfeler H, Pabst G, Trötzmüller M, Kolb D, Strohmayer K, Trajanoski S, \nHolzapfel GA, Schlenke P, Dorn I. 2019; Cholesterol Deficiency Causes Impaired \nOsmotic Stability of Cultured Red Blood Cells. Front Physiol. 10. \ndoi:10.3389/fphys.2019.01529 \n34. 2022 Sep; Caulier A, Sankaran VG. Molecular and cellular mechanisms that regulate \nhuman  erythropoiesis. Blood. 2022;139(16):2450-2459. Blood. 140(12):1451. \ndoi:10.1182/blood.2022017227 \n35. Vodyanik MA, Thomson JA, Slukvin II. 2006; Leukosialin (CD43) defines \nhematopoietic progenitors in human embryonic stem cell differentiation cultures. \nBlood. 108(6):2095–105. doi:10.1182/blood-2006-02-003327 \n36. Ogata T, Kozuka T, Kanda T. 2003; Identification of an Insulator in AAVS1, a \nPreferred Region for Integration of Adeno-Associated Virus DNA. J Virol. \n77(16):9000–7. doi:10.1128/jvi.77.16.9000-9007.2003 \n37. Mali P, Yang L, Esvelt KM, Aach J, Guell M, DiCarlo JE, Norville JE, Church GM. \n2013; RNA-guided human genome engineering via Cas9. Science (80- ). \n339(6121):823–6. doi:10.1126/science.1232033 \n38. Liu Z, Chen O, Wall JBJ, Zheng M, Zhou Y, Wang L, Ruth Vaseghi H, Qian L, Liu J. \n2017; Systematic comparison of 2A peptides for cloning multi-genes in a polycistronic \nvector. Sci Rep. 7(1):1–9. doi:10.1038/s41598-017-02460-2 \n39. Song L, Li X, Jayandharan GR, Wang Y, Aslanidi G V., Ling C, Zhong L, Gao G, \nYoder MC, Ling C, et al. 2013; High-Efficiency Transduction of Primary Human \nHematopoietic Stem Cells and Erythroid Lineage-Restricted Expression by Optimized \nAAV6 Serotype Vectors In Vitro and in a Murine Xenograft Model In Vivo. PLoS One. \n8(3):1–12. doi:10.1371/journal.pone.0058757 \n40. Levine R, Hazzard K, Lamberg J. 1982; The significance of megakaryocyte size. \nBlood. 60(5):1122–31. doi:10.1182/blood.v60.5.1122.1122 \n41. Ackermann M, Haake K, Kempf H, Kaschutnig P, Weiss AC, Nguyen AHH, Abeln M, \nMerkert S, Kühnel MP, Hartmann D, et al. 2021; A 3D iPSC-differentiation model \nidentifies interleukin-3 as a regulator of  early human hematopoietic specification. \nHaematologica. 106(5):1354–67. doi:10.3324/haematol.2019.228064 \n42. Cao J, O’Day DR, Pliner HA, Kingsley PD, Deng M, Daza RM, Zager MA, Aldinger \nKA, Blecher-Gonen R, Zhang F, et al. 2020; A human cell atlas of fetal gene \nexpression. Science,370(6518): eaba7721. doi:10.1126/science.aba7721 \n43. Xu J, Sankaran VG, Ni M, Menne TF, Puram R V, Kim W, Orkin SH. 2010; \nTranscriptional silencing of g -globin by BCL11A involves long-range interactions and \ncooperation with SOX6. Genes Dev. 24(8):783–98. doi:10.1101/gad.1897310. \n44. Sankaran VG, Xu J, Ragoczy T, Ippolito GC, Walkley CR, Maika SD, Fujiwara Y, Ito \nM, Groudine M, Bender MA, et al. 2009; Developmental and species-divergent globin \nswitching are driven by BCL11A. Nature. doi:10.1038/nature08243 \n45. Mcgrath KE, Frame JM, Fromm GJ, Koniski AD, Kingsley PD, Little J, Bulger M, Palis \nJ. 2011; A transient definitive erythroid lineage with unique regulation of the-globin \nlocus in the mammalian embryo. Blood. 117(17):4600–8. doi:10.1182/blood-2010 \n46. Hattangadi SM, Wong P, Zhang L, Flygare J, Lodish HF. 2011 Dec 8; From stem cell \nto red cell: regulation of erythropoiesis at mu ltiple levels by multiple proteins, RNAs, \nand chromatin modifications. Blood. 118(24):6258–68. doi:10.1182/blood-2011-07-\n356006 \n47. Nuez B, Michalovich D, Bygrave A, Ploemacher R, Grosveld F. 1995 May; Defective \nhaematopoiesis in fetal liver resulting from inactivation of the EKLF  gene. Nature. \n375(6529):316–8. doi:10.1038/375316a0 \n48. Hoeffel G, Chen J, Lavin Y, Low D, Almeida FF, See P, Beaudin AE, Lum J, Low I, \nForsberg EC, et al. 2015; C-Myb+ Erythro-Myeloid Progenitor-Derived Fetal \nMonocytes Give Rise to Adult Tissue-Resident Macrophages. Immunity. 42(4):665–\n78. doi:10.1016/j.immuni.2015.03.011 \n49. Mucenski ML, McLain K, Kier AB, Swerdlow SH, Schreiner CM, Miller TA, Pietryga \n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted May 8, 2026. ; https://doi.org/10.64898/2026.05.05.722134doi: bioRxiv preprint \n\n31 \nDW, Scott WJJ, Potter SS. 1991 May; A functional c-myb gene is required for normal \nmurine fetal hepatic  hematopoiesis. Cell. 65(4):677–89. doi:10.1016/0092-\n8674(91)90099-k \n50. Sturgeon CM, Ditadi A, Awong G, Kennedy M, Keller G. 2014; Wnt signaling controls \nthe specification of definitive and primit ive hematopoiesis from human pluripotent \nstem cells. Nat Biotechnol. 32(6):554–61. doi:10.1038/nbt.2915 \n51. Uenishi GI, Jung HS, Kumar A, Park MA, Hadland BK, McLeod E, Raymond M, \nMoskvin O, Zimmerman CE, Theisen DJ, et al. 2018; NOTCH signaling specifies \narterial-type definitive hemogenic endothelium from human pluripotent stem cells. Nat \nCommun. 9(1). doi:10.1038/s41467-018-04134-7 \n52. Ditadi A, Sturgeon CM, Tober J, Awong G, Kennedy M, Yzaguirre AD, Azzola L, Ng \nES, Stanley EG, French DL, et al. 2015; Human definitive haemogenic endothelium \nand arterial vascular endothelium represent distinct lineages. Nat Cell Biol. 17(5):580–\n91. doi:10.1038/ncb3161 \n53. Wang H, He J, Xu C, Chen X, Yang H, Shi S, Liu C, Zeng Y, Wu D, Bai Z, et al. 2021; \nDecoding Human Megakaryocyte Development. Cell Stem Cell. 28(3):535–549.e8. \ndoi:10.1016/j.stem.2020.11.006 \n54. Stephenson E. 2022. Human fetal yolk sac scRNA-seq data (sample ID: F158 for \nHaniffa Lab; 16099 for HDBR).  \n55. Korsunsky I, Millard N, Fan J, Slowikowski K, Zhang F, Wei K, Baglaenko Y, Brenner \nM, Loh P ru, Raychaudhuri S. Fast, sensitive and accurate integration of single-cell \ndata with Harmony. Nat Methods. doi:10.1038/s41592-019-0619-0 \n56. Chao Y, Xiang Y, Xiao J, Zheng W, Ebrahimkhani MR, Yang C, Wu AR, Liu P, Huang \nY, Sugimura R. 2023; Organoid-based single-cell spatiotemporal gene expression \nlandscape of human embryonic development and hematopoiesis. Signal Transduct \nTarget Ther. 8(1):2022–4. doi:10.1038/s41392-023-01455-y \n57. Beauchemin H, Shooshtarizadeh P, Vadnais C, Vassen L, Pastore YD, Möröy T. \n2017; Gfi1b controls integrin signaling-dependent cytoskeleton dynamics and \norganization in megakaryocytes. Haematologica. 102(3):484–97. \ndoi:10.3324/haematol.2016.150375 \n58. Detmer K, Walker AN. 2002 Jan; Bone morphogenetic proteins act synergistically with \nhaematopoietic cytokines in  the differentiation of haematopoietic progenitors. \nCytokine. 17(1):36–42. doi:10.1006/cyto.2001.0984 \n59. Zheng Z, Granado HS, Li C. 2023 Feb; Fibromodulin, a Multifunctional Matricellular \nModulator. J Dent Res. 102(2):125–34. doi:10.1177/00220345221138525 \n60. Bax D V, Bernard SE, Lomas A, Morgan A, Humphries J, Shuttleworth CA, \nHumphries MJ, Kielty CM. 2003 Sep; Cell adhesion to fibrillin-1 molecules and \nmicrofibrils is mediated by alpha 5  beta 1 and alpha v beta 3 integrins. J Biol Chem. \n278(36):34605–16. doi:10.1074/jbc.M303159200 \n61. Jovanovic J, Takagi J, Choulier L, Abrescia NGA, Stuart DI, van der Merwe PA, \nMardon HJ, Handford PA. 2007 Mar; alphaVbeta6 is a novel receptor for human \nfibrillin-1. Comparative studies of  mo lecular determinants underlying integrin-rgd \naffinity and specificity. J Biol Chem. 282(9):6743–51. doi:10.1074/jbc.M607008200 \n62. Li W, Wang Y, Zhao H, Zhang H, Xu Y, Wang S, Guo X, Huang Y, Zhang S, Han Y, et \nal. 2019. Identification and transcriptome analysis of erythroblastic island \nmacrophages.  \n63. Chasis JA, Mohandas N. 2008; Erythroblas tic islands: niches  for erythropoiesis. \ndoi:10.1182/blood-2008 \n64. Hom J, Dulmovits BM, Mohandas N, Blanc L. 2015 Dec; The erythroblastic island as \nan emerging paradigm in the anemia of inflammation. Immunol Res. 63(1–3):75–89. \ndoi:10.1007/s12026-015-8697-2 \n65. Jacobsen RN, Perkins AC, Levesque JP. 2015; Macrophages and regulation of \nerythropoiesis. Curr Opin Hematol. 22(3):212–9. \ndoi:10.1097/MOH.0000000000000131 \n66. Tamaoki N, Siebert S, Maeda T, Ha NH, Good ML, Huang Y, Vodnala SK, Haro-Mora \n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted May 8, 2026. ; https://doi.org/10.64898/2026.05.05.722134doi: bioRxiv preprint \n\n32 \nJJ, Uchida N, Tisdale JF, et al. 2023; Self-organized yolk sac-like organoids allow for \nscalable generation of multipotent hematopoietic progenitor cells from induced \npluripotent stem cells. Ce ll Reports Methods. 3(4):100460. \ndoi:10.1016/j.crmeth.2023.100460 \n67. Fuchs J, Nonn O, Daxboeck C, Groiss S, Moser G, Gauster M, Lang-Olip I, Brislinger \nD. 2021 Dec 1; Automated Quantitative Image Evaluation of Antigen Retrieval \nMethods for 17 Antibodies in Placentation and Implantation Diagnostic and Research. \nMicrosc Microanal. 27(6):1506–17. doi:10.1017/S1431927621012630 \n68. Ferkowicz MJ, Yoder MC. 2005 Sep; Blood island formation: longstanding \nobservations and modern interpretations. Exp Hematol. 33(9):1041–7. \ndoi:10.1016/j.exphem.2005.06.006 \n69. McGrath KE, Frame JM, Palis J. 2015 Dec; Early hematopoiesis and macrophage \ndevelopment. Semin Immunol. 27(6):379–87. doi:10.1016/j.smim.2016.03.013 \n70. Bertrand JY, Jalil A, Klaine M, Jung S, Cumano A, Godin I. 2005; Three pathways to \nmature macrophages in the early m ouse yolk sac. Blood. 106(9):3004–11. \ndoi:10.1182/blood-2005-02-0461 \n71. Palis J, Robertson S, Kennedy M, Wall C, Keller G. 1999; Development of erythroid \nand myeloid progenitors in the yolk sac and embryo proper of the mouse. \nDevelopment. 126(22):5073–84. doi:10.1242/dev.126.22.5073 \n72. Tober J, McGrath KE, Palis J. 2008 Mar 1; Primitive erythropoiesis and \nmegakaryopoiesis in the yolk sac are independent of c-myb. Blood. 111(5):2636–9. \ndoi:10.1182/blood-2007-11-124685 \n73. Ng ES, Sarila G, Li JY, Edirisinghe HS, Saxena R, Sun S, Bruveris FF, Labonne T, \nSleebs N, Maytum A, et al. 2025; Long-term engrafting multilineage hematopoietic \ncells differentiated from human induced pluripotent stem cells. Nat Biotechnol. \n43(8):1274–87. doi:10.1038/s41587-024-02360-7 \n74. Soares-da-Silva F, Peixoto M, Cumano A, Pinto-do-Ó P. 2020; Crosstalk Between the \nHepatic and Hematopoietic Systems During Embryonic Development. Front Cell Dev \nBiol. 8(July):1–20. doi:10.3389/fcell.2020.00612 \n75. Sugiyama D, Kulkeaw K, Mizuochi C, Horio Y, Okayama S. 2011 Jul; Hepatoblasts \ncomprise a niche for fetal liver erythropoiesis through cytokine  production. Biochem \nBiophys Res Commun. 410(2):301–6. doi:10.1016/j.bbrc.2011.05.137 \n76. Chou S, Lodish HF. 2010 Apr; Fetal liver hepatic progenitors are supportive stromal \ncells for hematopoietic  stem cells. Proc Natl Acad Sci U S A. 107(17):7799–804. \ndoi:10.1073/pnas.1003586107 \n77. Baker J, Liu JP, Robertson EJ, Efstratiadis A. 1993 Oct; Role of insulin-like growth \nfactors in embryonic and postnatal growth. Cell. 75(1):73–82.  \n78. Klees RF, Salasznyk RM, Vandenberg S, Bennett K, Plopper GE. 2007; Laminin-5 \nactivates extracellular matrix production and osteogenic gene focusing in human \nmesenchymal stem cells. Matrix Biol. 26(2):106–14. doi:10.1016/j.matbio.2006.10.001 \n79. Nakashima Y, Kariya Y, Yasuda C, Miyazaki K. 2005 Nov; Regulation of cell adhesion \nand type VII collagen binding by the beta3 chain  short arm of laminin-5: effect of its \nproteolytic cleavage. J Biochem. 138(5):539–52. doi:10.1093/jb/mvi153 \n80. Elkhal A, Tunggal L, Aumailley M. 2004 Jun; Fibroblasts contribute to the deposition \nof laminin 5 in the extracellular  matrix. Exp Cell Res. 296(2):223–30. \ndoi:10.1016/j.yexcr.2004.02.020 \n81. Sugiyama D, Kulkeaw K, Mizuochi C. 2013 Feb; TGF-beta-1 up-regulates extra-\ncellular matrix production in mouse hepatoblasts. Mech Dev. 130(2–3):195–206. \ndoi:10.1016/j.mod.2012.09.003 \n82. Zhang P, Zhang C, Li J, Han J, Liu X, Yang H. 2019; The physical microenvironment \nof hematopoietic stem cells and its emerging roles in engineering applications. Stem \nCell Res Ther. 10(1):1–13. doi:10.1186/s13287-019-1422-7 \n83. Shen J, Zhu Y, Zhang S, Lyu S, Lyu C, Feng Z, Hoyle DL, Wang ZZ, Cheng T, Zack \nWang CZ. 2021; Vitronectin-activated \nα vβ 3 and α vβ 5 integrin signalling specifies \nhaematopoietic fate in human pluripotent stem cells. Cell Prolif. 54. \n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted May 8, 2026. ; https://doi.org/10.64898/2026.05.05.722134doi: bioRxiv preprint \n\n33 \ndoi:10.1111/cpr.13012 \n84. Neildez-Nguyen TMA, Wajcman H, Marden MC, Bensidhoum M, Moncollin V, \nGiarratana MC, Kobari L, Thierry D, Douay L. 2002; Human erythroid cells produced \nex vivo at large scale differentiate into red blood cells in vivo. Nat Biotechnol. \n20(5):467–72. doi:10.1038/nbt0502-467 \n85. Blank U, Karlsson S. 2015 Jun; TGF-\nβ  signaling in the control of hematopoietic stem \ncells. Blood. 125(23):3542–50. doi:10.1182/blood-2014-12-618090 \n86. Gao X, Lee HY, da Rocha EL, Zhang C, Lu YF, Li D, Feng Y, Ezike J, Elmes RR, \nBarrasa MI, et al. 2016 Dec 8; TGF- β  inhibitors stimulate red blood cell production by \nenhancing self-renewal of BFU-E erythr oid progenitors. Blood. 128(23):2637–41. \ndoi:10.1182/blood-2016-05-718320 \n87. Xie Y, Bai H, Liu Y, Hoyle DL, Cheng T, Wang ZZ. 2015; Cooperative Effect of \nErythropoietin and TGF- β  Inhibition on Erythroid Development in Human Pluripotent \nStem Cells. J Cell Biochem. 116(12):2735–43. doi:https://doi.org/10.1002/jcb.25233 \n88. Pellegrino C, Stone EF, Valentini CG, Teofili L. 2024; Fetal Red Blood Cells: A \nComprehensive Review of Biological Properties and Implications for Neonatal \nTransfusion. Cells. 13(22):1843. doi:10.3390/cells13221843 \n89. Ackermann M, Rafiei Hashtchin A, Manstein F, Carvalho Oliveira M, Kempf H, \nZweigerdt R, Lachmann N. 2022 Feb; Continuous human iPSC-macrophage mass \nproduction by suspension culture in stirred  tank bioreactors. Nat Protoc. 17(2):513–\n39. doi:10.1038/s41596-021-00654-7 \n90. Dorn I, Klich K, Arauzo-Bravo MJ, Radstaak M, Santourlidis S, Ghanjati F, Radke TF, \nPsathaki OE, Hargus G, Kramer J, et al. 2015; Erythroid differentiation of human \ninduced pluripotent stem cells is independent of donor cell type of origin. \nHaematologica. 100(1):32–41. doi:10.3324/haematol.2014.108068 \n91. Cerdan C, Hong SH, Bhatia M. 2007 Oct; Formation and hematopoietic differentiation \nof human embryoid bodies by  suspension and hanging drop cultures. Curr Protoc \nStem Cell Biol. Chapter 1:Unit 1D.2. doi:10.1002/9780470151808.sc01d02s3 \n92. Gregorevic P, Blankinship MJ, Allen JM, Crawford RW, Meuse L, Miller DG, Russell \nDW, Chamberlain JS. 2004 Aug; Systemic delivery of genes to striated muscles using \nadeno-associated viral  vectors. Nat Med. 10(8):828–34. doi:10.1038/nm1085 \n93. Hao Y, Stuart T, Kowalski MH, Choudhary S, Hoffman P, Hartman A, Srivastava A, \nMolla G, Madad S, Fernandez-Granda C, et al. 2024; Dictionary learning for \nintegrative, multimodal and scalable sing le-cell analysis. Nat Biotechnol. 42(2):293–\n304. doi:10.1038/s41587-023-01767-y \n94. Hao Y, Hao S, Andersen-Nissen E, Mauck WM, Zheng S, Butler A, Lee MJ, Wilk AJ, \nDarby C, Zager M, et al. 2021; Integrated analysis of multimodal single-cell data. Cell. \n184(13):3573–3587.e29. doi:10.1016/j.cell.2021.04.048 \n95. Wu T, Hu E, Xu S, Chen M, Guo P, Dai Z, Feng T, Zhou L, Tang W, Zhan L, et al. \n2021; clusterProfiler 4.0: A universal enric hment tool for interpreting omics data. \nInnovation. 2(3):100141. doi:10.1016/j.xinn.2021.100141 \n96. Wickham H. 2009. ggplot2: Elegant Graphics for Data Analysis. Springer New York; \ndoi:10.1007/978-0-387-98141-3 \n97. Korotkevich G, Sukhov V, Sergushichev A. 2023 [cited 2024 Jun 21]; fgsea: Fast \nGene Set Enrichment Analysis. :1–29. doi:10.1101/060012 \n \n \n \n \n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted May 8, 2026. ; https://doi.org/10.64898/2026.05.05.722134doi: bioRxiv preprint \n\n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted May 8, 2026. ; https://doi.org/10.64898/2026.05.05.722134doi: bioRxiv preprint \n\n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted May 8, 2026. ; https://doi.org/10.64898/2026.05.05.722134doi: bioRxiv preprint \n\n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted May 8, 2026. ; https://doi.org/10.64898/2026.05.05.722134doi: bioRxiv preprint \n\n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted May 8, 2026. ; https://doi.org/10.64898/2026.05.05.722134doi: bioRxiv preprint \n\n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted May 8, 2026. ; https://doi.org/10.64898/2026.05.05.722134doi: bioRxiv preprint \n\n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted May 8, 2026. ; https://doi.org/10.64898/2026.05.05.722134doi: bioRxiv preprint \n\n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted May 8, 2026. ; https://doi.org/10.64898/2026.05.05.722134doi: bioRxiv preprint","source_license":"CC-BY-4.0","license_restricted":false}