An interactive resource of molecular signalling in the developing human haematopoietic stem cell niche

ABSTRACT The emergence of definitive human haematopoietic stem cells (HSCs) from Carnegie Stage (CS) 14 to CS17 in the aorta-gonad-mesonephros (AGM) region is a tightly regulated process. Previously, we conducted spatial transcriptomic analysis of the human AGM region at the end of this period (CS16/CS17) and identified secreted factors involved in HSC development. Here, we extend our analysis to investigate the progression of dorso-ventral polarised signalling around the dorsal aorta over the entire period of HSC emergence. Our results reveal a dramatic increase in ventral signalling complexity from the CS13-CS14 transition, coinciding with the first appearance of definitive HSCs. We further observe stage-specific changes in signalling up to CS17, which may underpin the step-wise maturation of HSCs described in the mouse model. The data-rich resource is also presented in an online interface enabling in silico analysis of molecular interactions between spatially defined domains of the AGM region. This resource will be of particular interest for researchers studying mechanisms underlying human HSC development as well as those developing in vitro methods for the generation of clinically relevant HSCs from pluripotent stem cells.

This study provides a valuable resource for researchers investigating the mechanisms underlying human HSC development, and for those developing in vitro methods for generating clinically relevant HSCs from pluripotent stem cells.By identifying stage-specific changes in molecular signalling during HSC development, this resource could aid in the development of more precise in vitro models for studying human haematopoiesis.

CS13 to CS14 as a key checkpoint in HSC generation
Ventral polarisation of signalling at the appearance of definitive HSCs (CS13/14).
We investigated the spatial molecular landscape of the AGM region at CS13, immediately before HSC generation, and at CS14, corresponding to the onset of definitive HSC generation (Ivanovs et al., 2011).We took transverse 10µm sections from the mid trunk of CS13 and CS14 embryos (n=2 embryos for each stage) using anatomical landmarks as guidelines to ensure the same region was taken across embryos.Sections were taken between the caudal appearance of the midgut loop and the caudal appearance of the liver and duodenum, before the bifurcation of the Ao .
Within this region, the Ao contains the most and largest IAHCs (Tavian et al., 1996;Tavian et al., 1999).For CS13, with LCM we subdissected concentric dorsal and ventral domains around the Ao (Fig. 1A).These domains included: i) ventral and dorsal 'inner' domains (VI and DI, respectively) comprising the Ao luminal layer (endothelial lining and adjacent mesenchymal cells) and ii) ventral and dorsal 'outer' domains (VO and DO, respectively) comprising sub-aortic mesenchyme reaching dorsally to the notochord and ventrally to the gonadal epithelium.At this early stage, the sub-aortic mesenchyme represents a thin layer, so the Ao endothelial floor is in close vicinity to the gonadal epithelium outlining the urogenital ridges.Therefore, due to the potential impact on IAHC/HSC development we also sub-dissected iii) gonadal-epithelium into the proximal domain (GEP) adjacent to the ventro-lateral wall of Ao and distal domain (GED), diverging from the Ao and descending ventrally (Fig. 1A).By CS14, the gonads become significantly distanced from the Ao floor due to thickening of the subaortic mesenchyme.Therefore, for CS14, urogenital ridges were excluded from analysis and the following 6 domains were subdissected: i) ventral and dorsal 'inner' (VI and DI), ii) 'mid' (VM and DM) and iii) 'outer' (VO and DO), which comprise the Ao endothelial lining, sub-aortic mesenchyme/stroma and more distant outer mesenchyme/stroma, respectively (Fig. 1B).
Transcriptome libraries from each of these spatial domains were generated and RNA-seq was performed.Concurrently, for both CS13 and CS14, we used consecutive sections immunostained for Development • Accepted manuscript CDH5 (VE-Cadherin) and RUNX1 (Runt-related transcription factor 1) to validate the presence of ventrally localized CDH5 + RUNX1+ IAHCs (Fig. S1B).
Emergence of definitive HSCs is accompanied by sharp increase in complexity of ventralised signalling.
For deeper analysis of changes in the molecular milieu in AoV during CS13CS14 transition, we focused on the ventral luminal lining where IAHC/HSCs emerge: CS14_VI vs CS13_VI and deeper stromal layers: CS14_VM vs CS13_VO (these domains are roughly equivalent in terms of size and distance from the Ao lumen) (Fig. 2A).We identified significant changes in expression of secreted VM vs CS13 VO) (Fig. 2A).Gene Set Enrichment Analysis (GSEA) identified in CS14_VI attenuation of Development • Accepted manuscript cell cycle related pathways such as G2M Checkpoint, E2F Targets and MYC Targets V1/V2 (Fig. 2B, C).
This decrease in cell proliferation can be explained by gradual endothelial maturation which may mask the potential increase in proliferation of the smaller fraction of maturing HSPCs (Batsivari et al., 2017;Canu et al., 2020).
We also detected in VI during CS13CS14 transition a strong activation of the Epithelial Mesenchymal Transition (EMT) Pathway (Fig. 2B, S2B).Given the mechanical similarities between EMT and EHT, upregulation of EMT genes could contribute to HSC development (Ottersbach, 2019).
We also observed an increase in the hypoxia pathway during this transition, consistent with previous reports implicating hypoxia in EHT (Imanirad et al., 2014) (Fig. 2B).These changes, which were also observed in the deeper ventral sub-endothelial layers, suggest a qualitative change in the HSC niche and consolidation of signalling through AoV layers, as evidenced by the shared pathways between the luminal and underlying stromal layers (Fig. S2C).Indeed, 6 out of 9 ventrally enriched pathways are shared by both CS13_VI and CS13_VO and 16 out of 26 ventrally enriched pathways are shared by both CS14_VI and CS14_VM (Fig. 2, S2C, Supplementary File 1).Full GSEA pathway analysis for both CS14_VI vs CS13_VI and CS14_VM vs CS13_VO pairwise comparisons can be found in Supplementary File 1.Our findings suggest complex and dynamic endothelial-stromal interactions involved in the regulation of HSC development, which are explored in detail in section 4.

Hallmark Pathway patterns
To gain insight into underlying signalling dynamics over the CS13-CS17 time window, we focused on the Ao inner luminal layer (VI) where EHT takes place.To this end, we combined our current CS13_VI and CS14_VI LCM-Seq datasets described above with CS16_VI and CS17_VI LCM-Seq datasets published previously (Crosse et al., 2020).We identified 6 major dynamic patterns of gene expression in the VI across the CS13-CS17 time window (Fig. 3A).Of particular interest was Group 2 comprising a broad array of genes with monotonically rising expression, which correlates with IAHC/HSC production across this period.Within this group, 'TGF Beta Signaling', 'Il6 Jak Stat3 Signaling', 'EMT' and 'Hypoxia' pathways increase up to CS16/17.Conversely, Group 3 genes show monotonic decrease of expression from CS13 to CS17, including pathways identified above during the AoV CS13CS14 transition, such as 'Oxidative phosphorylation'.Progressive downregulation of cell proliferation related pathways 'E2F Targets' and 'Myc Targets V1/V2' may be associated with gradual cell maturation of endothelial cells (Fig. 2B and Fig. 3).Of interest, during CS16CS17 transition, group 2 genes demonstrate a reverse trend (CS16 n=3, CS17 n=4): the lowest CS16 expressing genes become highly expressed at CS17, whereas the highest expressing genes drop their expression (Fig. 3A), which might manifest homeostatic stabilization of molecular signalling at the end of IAHC/HSC formation.
Full lists of genes within each expression pattern can be found in Supplementary File 2.
Overexpression of HOXA TFs has been used to enhance multi-potency and self-renewal of human haematopoietic progenitors (Doulatov et al., 2013;Dou et al., 2016;Sugimura, 2019).Recently, HOXA9 has been included in the molecular signature of the human AGM-derived HSCs (Calvanese et al., 2022).Therefore, our analysis of HOX genes is focused on the inner compartments of the Ao, VIthe site of HSC emergence, and DI.We found that all genes of the HOXA group are expressed in VI with different dynamics across the CS13 -CS17 window (Fig. 3B-C, S3A-B).The expression of majority of HOXA genes peaks at CS16 in line with increased detectability of HSC at this stage (Ivanovs et al., Development • Accepted manuscript 2011).Amongst all HOXA genes, the aforementioned ones: HOXA5, HOXA7, HOXA9, HOXA10 and additionally HOXA3 showed the highest expression levels.(Figure S3C).Stage-specific ventral enrichment of consecutive HOXA genes, HOXA10, HOXA11 and HOXA13, was observed at CS13, CS14 and CS17 respectively (Figure S3C).
Similar to HOXA, various HOXB genes are also broadly expressed in the AGM region (Fig. S3Ai and Aii).

Evolvement of the HSC molecular signature
We next sought to explore the expression dynamics of transcription factors RUNX1, HOXA9, MLLT3, MECOM and HLF, which were reported to be enriched in human AGM-derived HSCs (Calvanese et al., 2022).We found that expression of these TFs is high and peaks at CS16 before tapering off at CS17 (except RUNX1), potentially marking the end of the HSC temporal window (Fig. 3D).Of them, HLF behaves distinctively: its initially low expression raises steeply during CS13CS14 transition suggesting its early involvement in HSC formation (Fig. 3D).Notably, HLF is considered to be the most specific within the AGM-derived HSC signature (Calvanese et al., 2022).
The analysis of dorso-ventral polarisation revealed fluctuations in differences of expression of these genes between VI and DI within the range of + 1.25 Log2 fold change interval.Of them, only one, HLF showed dramatic ventral enrichment during CS13-CS14 transition (Figure 3C), again pointing out at its potential role in early IAHC/HSC development.Later, significant HLF ventral polarisation, along with RUNX1, was observed at CS17, whereas HOXA9 expression tended to go down (Fig. 3C).

Development • Accepted manuscript
Stage-specific molecular interactions in AoV during IAHC/HSC formation.
To better characterize potential signalling interactions between spatial domains , Network Analysis Toolkit for Multicellular Interactions, NATMI (Hou et al., 2020) was used to predict ligand -receptor interactions for LCM-Seq CS13, CS14 and previously published LCM-Seq CS16 datasets (Crosse et al., 2020).The full set of interactions between spatial domains for each dataset can be interactively explored at https://medvinsky-lab.github.io/hsc-niche-explorer/.A heatmap for predicted (potential) ligand-receptor interactions by specificity edges (see methods) between LCM-Seq spatial domains for each embryo stage is shown in Fig. 4A.For each stage, we have focused predominantly on signalling towards receptors in the VI as the site of HSC emergence.
At CS13, VI is predicted to receive the highest number of signalling ligands from itself and DI, its dorsal counterpart (Fig. 4A).(Specific ligand-receptor interactions will be further labelled as follows: ligandreceptor).Network analysis of the predicted top ligand-receptor interactions shows a cluster of adhesion molecules signalling within CS13_VI including integrins (ITGAM, ITGAL) and ICAM1-3 (all edge weights > 0.2) (Fig. 4B).There are also several cytokinereceptor interactions within VI including CCL28  CCR10 which has a described role in promoting proliferation of primitive human HSPC (Karlsson et al., 2013).By contrast to VIVI, the predicted DIVI molecular interactions (Fig. S4A), do not include integrins and have fewer cytokines in line with the paucity of proinflammatory signalling described above for the CS13_DI (Fig. 1D).Although ventral signalling is pivotal for HSC development, the dorsal domain of the Ao may also contribute into this process as shown for the mouse (Souilhol et al., 2016a).For example, NODAL and INHBA known to be involved in mesoderm patterning, fate specification and epithelial-mesenchymal transition (Green et al., 1992;Perea-Gomez et al., 2002;Shen, 2007;Yu et al., 2021) are secreted by CS13_DI.Despite having relatively few predicted interactions with CS13_VI (Figure 4A), the GEP, which is initially adjacent to the ventral floor of the Ao might be a critical source of BMP4 for developing HSCs at this early stage (Figure S4B).
Distancing of the GEP from the Ao at later stages may attenuate the impact of BMP4, which may be necessary for completion of HSC maturation as shown in the mouse (Souilhol et al., 2016a).
At CS14, the VI maintains a high number of predicted interactions within itself (VI ligandVI receptor) including again a cluster of adhesion molecules but now centred around selectins (SELL, SELP) rather than integrins (Fig. 4A,B).Such temporal changes in adhesion dynamics during CS13CS14 might underlie cell transitions and movements during EHT, including such processes as the potential rolling Development • Accepted manuscript of HSPCs along the Ao endothelium and trans-endothelial monocyte/macrophage migration (Perlin et al., 2017;Mariani et al., 2019).Notably, PROC (Protein C), has predicted interactions with its 3 receptors PROCR (EPCR), THBD and TEK within CS14_VI.PROCR is an important marker for mouse pre-HSC and human fetal liver HSCs (Zhou et al., 2016;Subramaniam et al., 2019).Given the functional importance of PROCR in HSC (Gur-Cohen et al., 2016;Chagraoui et al., 2019), the spike of PROC signalling within CS14_VI might contribute to the onset of IAHCs and development of definitive HSCs.
Another contrast to CS13 is that at CS14 the highest number of signalling ligands being received by the VI are predicted to derive from the VO suggesting its impact on initiation of IAHC/HSC specification (Fig. 4A).However, since VO is separated from the VI by the VM layer, impactful VOderived molecules are likely to be low molecular weight diffusible proteins (usually below 45kDa) capable of migrating significant distances.Among these CS14_VO-derived ligands is LIF, which can These data were organised in two separate datasets visualised by UMAP: (1) endothelial/haematopoietic and (2) mesenchymal/epithelial populations of the Ao proximal stromal layers (Fig. 5A).Clustering analysis followed by marker identification determined population identities.Within the endothelial/haematopoietic dataset the endothelial CD144 (VE-cad)-positive cells were split into two major types expressing classical arterial markers GJA5, GJA4 (Cl.3) and venous markers NR2F2; APLNR (Clusters, Cl.1 & Cl.6) (Fig. 5B).However, this arterial-venous split is not clearly demarcated due to promiscuous expression of these and other markers (Fig 5C ), similar to our previous observation for CS16 (Crosse et al., 2020).For convenience, Cl.3 will be further termed "arterial" and Cl.1 & Cl.6 "venous", based on predominant expression of corresponding markers.

Signalling networks across the HSC developmental niche
Sections 1 and 2 highlighted signalling pathways potentially involved in HSC generation owing to their ventralised enrichment.Using GSEA of the single cell datasets, these pathways were resolved to discrete populations (Fig. 6A).Some of them, such as 'IL2 Stat5 signaling', 'Wnt Beta Catenin', 'TGF Beta' (all Cl.3) and 'Hypoxia' (Cl.resolved ligand-receptor pairs of interest within the above pathways which may mediate interactions between cell clusters (Fig. 6B).For example, EDN1 supporting HSC emergence in the human embryo (Crosse et al., 2020) is part of the ventrally enriched 'TNFA signalling via NFKB pathway' in the CS14 AoV (Fig. 1C).Analysis of predicted relationships in CS14 AoV showed signalling of EDN1 predominantly from arterial Cl.3 towards mesenchymal Cl.4 & Cl.9 via receptor EDNRA and back to itself as well as towards the endothelial Cl.1 via receptor EDNRB (Fig. 6B), recapturing our previous finding for CS16 (Crosse et al., 2020).Similarly, the Wnt antagonist SFRP1 downregulated in the CS14 AoV compared to CS13 AoV (Fig. 2A), was predicted to have two directions of signalling through alternate receptors.Namely, predominantly stromal derived (Cl.0, 4, 8 & 9) SFRP1 is predicted to signal towards arterial Cl.3 and HSPC Cl.7 through FZD6, and signal back to the same stromal compartments in an autocrine manner through FZD2 (Fig. 6B).
NATMI was also used to predict ligand-receptor interactions between populations in an unbiased way.
Venous Cl.6 (followed by arterial Cl.3, and HSPC Cl.7), consistently had the highest number of receptors with predicted ligand pairs from all populations across the dataset (Fig. 6C).This is We also found that mesenchymal CL.9 secretes CXCL12 predictively targeting the arterial Cl.3 via CXCR4 receptor (Fig. 6D).This is of interest since CXCL12 is involved in maintenance and function of HSCs in vivo (Lapidot and Kollet, 2002;Wright et al., 2002;Sugiyama et al., 2006).
Ligand-receptor pairs were also predicted between endothelial Cl.3 & Cl.6 and HSPC Cl.7 in vivo (Fig. 6C, E), which may reflect involvement of autocrine signalling during EHT and support previous reports on the role of endothelium in the AGM niche (Butler et al., 2010;Heck et al., 2020;Hadland et al., 2022).Interactions of endothelial Cl.3 & Cl.6 with HSPC Cl.5 are also predicted but less prominent.
CD44 which is upregulated in CS14_VI (Fig. 1C) and marks EHT (Oatley et al., 2020)  Our analysis revealed an interesting feature of Notch signalling in the AoV.Notch signalling is essential for HSC development (Kumano et al., 2003;Hadland et al., 2004;Gama-Norton et al., 2015;Uenishi et al., 2018), but needs to be downregulated by the end of HSC maturation in the AGM region, which may be mediated, at least partly, by Notch ligands with different strength (Bigas and Espinosa, 2012;Souilhol et al., 2016b).We found that Notch ligands DLL4 and DLL1 are expressed separately by arterial and venous endothelia (Cl.3 and Cl.6 respectively) (Fig. 6E), raising the possibility that HSC maturation is driven by differential spatially and temporally defined contacts with these endothelial compartments.NATMI-predicted signalling interactions between all populations in this dataset can be interactively explored at https://medvinsky-lab.github.io/hsc-niche-explorer/ .

DISCUSSION
Embryonic HSC precursors emerge within the complex signalling milieu of the AGM region in the Ao (Medvinsky and Dzierzak, 1996;de Bruijn et al., 2002).Significant progress has been made in identifying key genetic regulators of HSC development in model organisms (Ciau-Uitz et al., 2016;Weijts et al., 2021).Due to ethical reasons and scarcity of samples, HSC development in humans has been much less studied than in model organisms.Progress in analysis of mechanisms underlying human HSC development is further hampered due to differences with model organisms as well as failure to establish human AGM cultures capable of recapitulating HSC development (Easterbrook et al., 2019).As a result, attempts to derive bona fide HSCs from pluripotent stem cells without genetic reprogramming have been unsuccessful (Sugimura et al., 2017;Fidanza and Forrester, 2021).Analysis of the data-rich resource generated in this study, revealed a plethora of molecular processes with a potential role in HSC development that may serve to fill the gaps in our understanding.
The evidence that IAHC and HSC development in the Ao is ventralised, sparked interest in spatially polarised expression of genes in the AGM region (Tavian et al., 1996;Marshall et al., 2000;Taoudi and Medvinsky, 2007;Peeters et al., 2009;Ivanovs et al., 2014).The spatial asymmetry of IAHC/HSC development may result from differential signals emitted from the surrounding stroma (Souilhol et al., 2016a;McGarvey et al., 2017;Crosse et al., 2020) and also from differences in origins of the dorsal Development • Accepted manuscript and ventral portions of the Ao (Pardanaud et al., 1996).Combined spatial and single cell transcriptomics proved to be a powerful approach for analysing HSC development.It led to identification of factors supporting HSC development, such as the ADM-RAMP2 ligand-receptor pair between the niche and emerging HSPC (Yvernogeau et al., 2020), Endothelin 1 secreted molecule (Crosse et al., 2020) and of the molecular signature enriched in developing human HSC (Calvanese et al., 2022).
Here, based on understanding that HSC development in the mouse model is step-wise over a temporal window (Taoudi et al., 2008;Rybtsov et al., 2011;Rybtsov et al., 2014;Zhou et al., 2016;Hadland et al., 2022), we have captured a dynamic representation of polarized signalling throughout the human CS13-CS17 HSC developmental period.We had a particular interest in CS13CS14 transition, when IAHC and definitive HSC emerge first.Significant dorso-ventral polarisation was detectable already at CS13, prior to emergence of definitive IAHC/HSCs, which is reflected in upregulation of respective markers (e.g.MYB, PROCR, HLF, CD44 and MECOM, of which HLF is an important member the molecular signature of developing HSCs) (Calvanese et al., 2022).
Simultaneously, we observed a striking increase in signalling complexity in AoV, likely reflecting both vascular remodelling and EHT.One of the upregulated signalling pathways was EMT, which may contribute to EHT via an integral role of molecules such as CD44 (Xu et al., 2015;Oatley et al., 2020).
Overall, significant increase in complexity of signalling suggests that CS13CS14 transition is a key check-point potentially underlying appearance of IAHCs and definitive HSCs.Dynamic changes within the VI compartment during the HSC developmental window included expression waves of integrins, selectins and ephrins observed sequentially at CS13, CS14 and CS16 stages.We do not have a clear picture of migration pathways of mammalian HSCs and adhesion mechanisms underlying their mobility, although β1 integrins were shown to be essential for colonisation of the embryonic liver (Hirsch et al., 1996).Selectins act as substrates for HSPC rolling and migration in the human and zebrafish (Greenberg et al., 2000;Esain et al., 2015) and ephrins mediate haematopoietic cells interactions with adult bone marrow stroma (Ting et al., 2010;Kwak et al., 2016).How dynamic changes in adhesion impact IAHC formation/ HSC maturation and migration behaviour requires further elucidation.
We also looked into dynamic expression of HOX genes during HSC emergence as these genes play different important roles during haematopoietic development and differentiation (Bhatlekar et al., 2018).We found that all genes of HOXA, B and C groups are expressed in the VI during CS13-CS17 at different levels and with different dynamics.What the role of this complex spatiotemporal pattern of HOX genes expression in HSC development remains an open question.Our resource enables various hypotheses to be raised and tested.
Endothelial cells give rise to HSCs and progenitor cells and concurrently are an important component of the HSC niche, both in the embryo and in the adult (Kenswil et al., 2021;Hadland et al., 2022;Yu et al., 2022).Our combined spatial and single cell transcriptomics analysis revealed heterogeneity in the composition of the Ao endothelium with some overlap between arterial and venous molecular characteristics, in line with our previous report for CS16 (Crosse et al., 2020).HSPCs in the Ao were represented by two distinct populations, which differed strikingly by proliferative state and proinflammatory signature.Our analysis supports a broadly accepted view on lineage relationship of HSPCs with the arterial endothelium (Clements and Traver, 2013;Park et al., 2018).However, we also observed a strong predicted relationship with the venous-like endothelium, as we have previously

Human embryonic material
Human embryonic samples of Carnegie Stages 13 -14 were provided by the MRC Centre for Reproductive Health and by the Joint MRC / Wellcome (MR/R006237/1) Human Developmental Biology Resource (https://www.hdbr.org/).This study was approved by the Lothian Research Ethics Committee.The embryos were obtained immediately after elective termination of pregnancy for which each patient gave informed consent in writing.Embryos were either used immediately as fresh tissue or flash frozen in Optimal Cutting Temperature (OCT) compound and stored at −80°C.

Laser Capture Microdissection
Human embryos embedded in OCT stored at −80°C were equilibrated to −24°C and sectioned in a caudal-to-rostral direction using a cryotome FSE cryostat (Thermo Scientific).Frequent checks under Development • Accepted manuscript the microscope verified the level reached along the rostral-caudal axis as defined by anatomical landmarks.For both CS13 and CS14, transverse 10µm sections were taken between the most caudal appearance of the midgut loop and the caudal appearance of the liver (Fig. S1A).Once the appropriate level had been reached cryosections were transferred onto nuclease-free polyethylene napthalate (PEN) membrane slides (Zeiss).At intervals, a sister section would be transferred to a SuperFrost slide for future validation of ventral IAHCs by immunohistochemical analysis.For visualisation during LCM, sections were stained using a rapid Haematoxylin and Eosin staining protocol -3 minutes (min) in 70% ethanol, 1 min H2O, 4 min Mayer's Haematoxylin Solution (Sigma-Aldrich), 2 min tap H2O, 15 s Eosin Y (Sigma-Aldrich), 1 min 70% ethanol, 1 min 90% ethanol, 3 mins 100% ethanol.All H 2 O was treated with diethyl pyrocarbonate (DEPC) and all reagents were precooled in ice except 100% ethanol which was room temperature.
The laser capture microscope used was the PALM microbeam (Zeiss).The microscope and surrounding area were sprayed down with RNaseAWAY ® (Sigma-Aldrich).Sections were viewed and microdissected in brightfield using a 10X objective.The microdissected regions were collected into the caps of AdhesiveCap 500 opaque (Zeiss) 500μl PCR tubes.15μl lysis buffer (Nichterwitz et al., 2016) (0.2% Triton X-100 (Sigma-Aldrich), 2U/μl Rnase inhibitor (Takara), Phosphate Buffer Solution) was added directly on top of the dissected tissue in the tube caps and the tubes were closed in an inverted position.

Immunofluorescence and confocal imaging
Embryo sections, obtained as described above, were fixed in cold 4% PFA (Sigma Aldrich) for 10 minutes and stained as follows: 3x wash in PBS, 5 min each.10 min permeabilisation in PBS/0.5% Triton X-100 (Sigma Aldrich).2x 5 min PBS wash.30 min PBS/10% FCS protein block.Overnight incubation with a primary antibody diluted in PBS/2% FCS.2x 5 min PBS wash.secondary antibodies.Negative controls (no primary antibodies) were also included.
For 10X single cell sequencing the dorsal aorta was dissected from a CS13 human embryo and dissociated into single cells in 1mg/ml Collagenase-Dispase (Roche) and 0.12 mg/ml of DNase I

LCM-Seq Transcriptome Analysis
Sequencing read quality was evaluated using FastQC, which assessed Phred quality score, adaptor contamination, GC content, and duplicate levels.Illumina adaptor sequences were removed using Flexbar (Dodt et al., 2012).The alignment of reads to the human reference genome hg38 (Ensembl version 85) was performed using STAR (Dobin et al., 2013) and SAMtools (Li et al., 2009) was employed to sort and index the aligned reads.The read fragments per gene were counted and a matrix of reads per gene was generated using BEDtools' multicov function.
In R, the tool DESeq2 (Love et al., 2014) was used for DGE analysis to determine fold changes in expression between different spatial domains while controlling for per embryo batch effects.The p.
value was a Wald test statistic for differential expression between two domains.The likelihood ratio test was used to model gene dynamics across CS13-CS17 ventral domains.DEGreport was used to cluster genes differentially expressed along the CS13-CS17 temporal axis into groupings based on patterns of expression (Pantano, 2023).P values were adjusted for by multiple hypothesis correction with the Benjamini and Hochberg method to produce adjusted p. values (p.adj).Genes were considered significant with a p.adj < 0.05.R packages EnhancedVolcano, pheatmap and ggplot2 were used to make plots visualizing gene expression.Genes were determined to encode secreted factors, receptors or transcription factors with an unbiased approach using Uniprot: KW-0964, GO: 0038023 and GO: 0003700 terms respectively.Some overlap between Uniprot: KW-0964 and GO: 0038023 is explained by the fact that transmembrane cell surface receptors are secreted.
Gene Set Enrichment Analysis was done using the tool from the Broad Institute (Subramanian et al., 2005) using the Hallmark gene sets to identify pathways.Genes were input as a ranked list by Wald statistic and pathways were considered significant with a false discover rate (FDR) of < 0.1.

Single cell analysis
CS13 dorsal aorta single cell data was processed using the Cell Ranger 2.1.0(10x Genomics) analysis pipeline by aligning reads to GRCh38 human transcriptome (Ensembl).The ScanPy pipeline (Wolf et al., 2018) was used to explore and integrate the dataset with other dorsal aorta single cell datasets spanning stages CS13 -CS15 from a publicly available resource (Zeng et al., 2019).For all datasets, cells with less than 200 genes and genes that were in less than 3 cells were filtered out.Cells with a percentage of mitochondrial genes >1 were also filtered out.Contaminating HB genes were also removed from the CS13 dorsal aorta dataset.The reads per cell were normalized and logarithmised.
The variance effects of total counts per cell, cell cycle and mitochondrial genes were regressed out.
Each dataset was then subset to include only genes with highly variable expression.The tool Scanorama (Hie et al., 2019) was used to integrate datasets.Then nearest neighbours was computed using 10 principal components and the neighbourhood graph was embedded in two dimensions in a Uniform Manifold Approximation and Projection (UMAP) (Becht et al., 2018).Clustering of subpopulations within the UMAP were made using the Leiden algorithm (Traag et al., 2019).Following cluster identification by gene signature and any known identities from prior sorting strategies, the data was then subset into two parts to include haematopoietic and endothelial populations in one subset and stromal and epithelial populations in the second.Partition-based graph abstraction (PAGA) (Wolf et al., 2019) was used to make lineage inferences from the neighbourhood graphs.
The tool NATMI (Hou et al., 2020) was used for predicting ligand-receptor interactions across spatial domains and single cell clusters and which is the foundation for the web interactive exploration tool.
described at CS16.Notably, the venous population showed the most abundant ligand-receptor Development • Accepted manuscript interactions (in both directions) with itself, arterial-like cells, HSPCs and stromal cells, suggesting an integral role of the venous compartment within the AGM niche.The precise relationship between arterial/venous endothelium within the human dorsal aorta and HSC/HSPC requires further detailed investigation.In summary, we have generated a data-rich resource, which deconvolutes the signalling landscape of the developing human HSC niche at the spatial and cell population levels over the period of HSC generation in the human AGM region.This provides a hypothesis-generating platform for functional investigation into cell lineage relationships and mechanisms of HSPC generation.An interactive application presented here enables efficient exploration of this resource.Researchers can crossreference with their preferred model organisms' datasets to facilitate identification of evolutionarily conserved and disparate mechanisms of HSC development.This resource may provide useful insights into ex vivo derivation of transplantable HSC for clinical needs.

Fig. S1 .Fig. S3 .
Fig. S1.(related to figure 1): Ai: A CS14 embryo with the regions of the mid trunk sectioned for laser capture microdissection indicated between the two dotted lines.Aii: Representative images of anatomical landmarks used for identification of the region to take for laser capture microdissection from a CS14 embryo; the caudal appearance of the midgut loop = most caudal, the caudal appearance of the liver and duodenum = most rostral.B: Transverse section of the dorsal aorta from a CS14 embryo showing CDH5+Runx1+ intra-aortic haematopoietic clusters exclusively in the ventral portion (below the dotted line).Scale bar is 50 μm.D = dorsal, V = ventral.
is enriched in HSPC Development • Accepted manuscript Cl.7 and can potentially mediate signalling from arterial Cl.3 via secreted FN1 ligand, and from venous Cl.6 via FN1, MADCAM1, VIM and FGF2 ligands