Derivation of extra-embryonic and intra-embryonic macrophage lineages from human pluripotent stem cells

Macrophages are well appreciated for their roles in pathogen responses, including phagocytosis of pathogens and removal of cellular debris, release of inflammatory cytokines, and presentation of antigen to cells of the adaptive immune system (Murray and Wynn, 2011). Over the last ten years, however, the impact of macrophages residing within tissues on organ homeostasis and repair has become increasingly understood. Studies of resident macrophages across various tissues have revealed common functions, such as promotion of angiogenesis and proper tissue patterning during development (Wynn et al., 2013). Each tissue environment also instructs macrophages to take on specific functional properties, leading to specialization of microglia in the brain, Kupffer cells in the liver, Langerhans cells in the skin and osteoclasts in the bone (Epelman et al., 2014b; Davies et al., 2013; Lavin et al., 2014).

Tissue-resident macrophages are initially derived from the hematopoietic progenitors arising in the extra-embryonic yolk sac (Hoeffel and Ginhoux, 2015; Hoeffel et al., 2015; Ginhoux et al., 2010; Yona et al., 2013). During early embryogenesis, these extra-embryonic cells migrate to developing tissues where they establish their tissue-specific identity. Over the course of fetal development and postnatal life, these tissue-resident macrophages have the capacity to self-renew and are maintained in the absence of replenishment from monocytes in many tissues, including liver, lung, brain and heart (Hashimoto et al., 2013; Epelman et al., 2014a; Ajami et al., 2007). Similarly, upon physiological stressors, such as injury or aging, monocyte-derived macrophages can migrate to these tissues and function as tissue-resident macrophages (Epelman et al., 2014a; Daemen et al., 2021). However, these ontogenically distinct macrophages generally lack the reparative functions of their extra-embryonic-derived counterparts, instead causing inflammation that can contribute to disease phenotypes (Honold and Nahrendorf, 2018; Puranik et al., 2018).

Extra-embryonic hematopoiesis in the yolk sac occurs in at least two waves, the first consisting of primitive, MYB-independent progenitors that give rise to a very limited set of hematopoietic cells (primitive erythroblasts, megakaryocytes and macrophages) and the second consisting of MYB-dependent erythromyeloid progenitors (EMPs) that give rise to erythroid, myeloid and natural killer cell lineages (Dege et al., 2020; Frame et al., 2013; Kasaai et al., 2017; McGrath et al., 2015; Palis et al., 1995; Palis et al., 1999; Palis and Yoder, 2001; Lux et al., 2008). EMPs provide the first progenitors that colonize the fetal liver, giving rise to diverse myeloid cells. In contrast, the intra-embryonic hemogenic endothelium within the aorta-gonad-mesonephros (AGM) region gives rise to definitive multipotent progenitors (MPPs) and hematopoietic stem cells (HSCs). These long-lived HSCs migrate to the fetal liver, and finally the bone marrow, to produce a full assortment of myeloid and lymphoid cells.

In order to assess the intrinsic differences between hematopoietic lineages arising from extra- and intra-embryonic hematopoiesis, we developed an in vitro human pluripotent stem cell (hPSC) differentiation model of hematopoiesis that recapitulates these anatomically distinct hematopoietic programs. By manipulating WNT signaling at an early stage of differentiation, we can generate either exclusively extra-embryonic-like (WNT-independent, WNTi) CD34⁺ hematopoietic progenitors, or exclusively intra-embryonic-like (WNT-dependent, WNTd) CD34⁺ hematopoietic progenitors (Sturgeon et al., 2014). WNTi progenitors lack the expression of HOXA genes and have restricted erythro-myeloid potential, mirroring extra-embryonic hematopoiesis, whereas WNTd progenitors express HOXA, and display multi-lineage erythro-myelo-lymphoid potential, recapitulating intra-embryonic-like MPP development (Ng et al., 2016).

Here, we present a platform to generate two ontogenically distinct myeloid populations from hPSCs. Through our ability to directly compare each population from a single isogenic source, we demonstrate that macrophages differentiated from extra-embryonic-like (WNTi) and intra-embryonic-like (WNTd) hematopoietic progenitors harbor distinct transcriptional signatures, morphologies, developmental pathways, and functions, which recapitulate the general properties observed in yolk sac-derived and intra-embryonic macrophages, respectively. These findings establish a robust in vitro system for generating functionally distinct myeloid populations for investigation of human macrophage specification, differentiation, function and regenerative potential.

Using previously described conditions to exclusively specify extra-embryonic-like and intra-embryonic-like hematopoietic progenitors from hPSCs, we sought to determine whether we could generate functional macrophages in these lineages (Fig. 1A). Consistent with previous findings, differentiation under WNTi conditions (IWP2 treatment) gave rise to CD34⁺ progenitors that rapidly transitioned to CD43 (SPN)⁺ hematopoietic progenitors by day 8 of differentiation. Within these cultures, CD34⁻CD43⁺ cells displayed predominantly erythroid potential, whereas CD34⁺CD43⁺ cells instead harbored robust myeloid potential, identifying them as a candidate population of extra-embryonic-like macrophage progenitors (Fig. S1A,B). In contrast, treatment with the GSK3 inhibitor CHIR99021, which activates WNT (and consequently Notch) signaling, and the TGFβ inhibitor SB-431542 (called WNTd conditions here for simplicity), yields CD34⁺CD43⁻CD73 (NT5E)⁻CD184 (CXCR4)⁻ hemogenic endothelial cells on day 8 of differentiation, which possess intra-embryonic-like erythroid, myeloid and lymphoid multipotent potential (Ditadi et al., 2015; Ng et al., 2016).

WNTi CD34⁺CD43⁺, WNTd CD34⁺CD43⁻CD73⁻CD184⁻ and human cord blood (CB) CD34⁺ populations were isolated by fluorescence-activated cell sorting (FACS), then cultured under serum-free conditions supplemented with M-CSF (CSF1) and additional cytokines to drive macrophage differentiation (Fig. S1A). CB CD34⁺ hematopoietic progenitors were used as a control for definitive hematopoiesis. After 14 days, the cultures were analyzed by flow cytometry for expression of the leukocyte marker CD45 (PTPRC) and common macrophage markers. Differentiation in all cultures was highly efficient, as over 90% of the cells expressed CD45 and the macrophage markers CD14 and MERTK (Fig. 1B). The WNTd culture yielded an average of 2×10⁶ macrophages per 1×10⁵ purified progenitors, whereas the WNTi culture averaged 1×10⁶ macrophages per 1×10⁵ sorted progenitors. WNTi and WNTd cultures were each generated from 5×10⁶ hPSCs. Whereas the cells within the WNTd- and CB-derived cultures had similar scatter properties, WNTi-derived cells displayed higher median forward scatter (FSC) and side scatter (SSC) (Fig. 1C, Fig. S1C). We also examined CCR2 expression in each population, as its expression on tissue-resident macrophages correlates with an intra-embryonic, but not extra-embryonic, origin (Bajpai et al., 2018; Epelman et al., 2014a). Consistent with an extra-embryonic-like origin, WNTi CD45⁺CD14⁺ cells were completely devoid of CCR2 expression, whereas the WNTd- and CB-derived cells both exhibited CCR2⁺ and CCR2⁻ subsets (Fig. 1D). Electron microscopy revealed that the cells from each culture contained morphological elements typical of macrophages, including phagosomes, lysosomes and pseudopodia (Fig. 1E). Finally, we assessed the phagocytosis efficiency of each population by culturing the cells with pHrodo Red Escherichia coli BioParticles, which will only fluoresce in low pH phagosomes. Both the WNTi and WNTd macrophages had similar percentages (comparable to CB-derived macrophages) and levels of pHrodo Red positivity, demonstrating comparable phagocytic capacity for E. coli particles (Fig. 1F, Fig. S1D). Collectively, these phenotypic analyses indicate that each CD45⁺ population comprises macrophages, with hPSC-derived WNTd macrophages displaying strong similarity to those derived from CB HSPCs, whereas WNTi macrophages are phenotypically unique.

Monocyte-derived macrophages and yolk-sac-derived tissue-resident macrophages have distinct gene expression profiles. However, the extent to which these gene expression differences are cell intrinsic or driven by their environment remains incompletely understood. As our in vitro platform yields macrophages from extra- and intra-embryonic-like progenitors, we investigated whether each population harbors gene expression differences, at single-cell resolution. From macrophage day 14 cultures, 11,113 cells were captured for single-cell RNA sequencing (scRNAseq). Both WNTd and WNTi samples had comparable numbers of genes and unique molecular identifiers (UMIs) per cell (Fig. S2A). Unsupervised clustering of the WNTi and WNTd macrophages together by uniform manifold approximation and projection (UMAP) identified 12 cell clusters, with the WNTd and WNTi cells predominantly segregating into distinct groups of clusters (Fig. 2A,B, Fig. S2B). Differentially expressed genes (DEGs) within each cluster allowed for the identification of cell fate (Fig. 2C, Table S1). Although most cells were in G0/G1 across both datasets, we also identified a separate cluster with enriched expression of S/G2/M-associated genes [TOP2A, KIAA0101 (PCLAF), MKI67, CDK1; Fig. 2B,C, Fig. S2B,C].

Macrophages were identified based on expression of SPP1, C1QA, C1QB and C1QC and made up most clusters in both the WNTi and WNTd populations (Fig. 2C). The WNTd and WNTi macrophages also expressed comparable levels of MAF, MAFB, CD68, CD163 and MRC1 (Fig. S2D,E). In contrast, a cluster of cells expressing characteristic monocyte genes (FCN1, CD52, LYZ and PLAC8), as well as CCR2, was found exclusively in the WNTd culture (Fig. 2C, Fig. S2F). We also identified a small cluster expressing mast cell genes (KIT, TPSB2, CPA3, HPGD) in the WNTi culture. At a broad level, the WNTd and WNTi macrophages each expressed a distinct gene signature that spanned all their respective clusters (Fig. 2D,E, Fig. S3). Interestingly, the WNTi macrophage signature includes LYVE1, F13A1, GAS6, FOLR2, IGFBP4 and CD36, genes that are enriched in a variety of tissue-resident macrophages (Summers et al., 2020). The WNTd macrophages, in contrast, expressed a broader repertoire of macrophage-associated genes that regulate function in disease states, such as TREM2, BTG1, ALDH2, ALOX5AP and BNIP3 (Fig. 2D,E, Fig. S3) (Corjay et al., 1998; Zhong et al., 2019; Ee et al., 2016; Bäck et al., 2007; Byrum et al., 1997; Frank et al., 2015; Zhang and Ney, 2009). In particular, TREM2 is expressed on macrophages in a variety of disease contexts, including cancer and atherosclerosis, so it is intriguing to find it present across the WNTd macrophage population (Cochain et al., 2018; Deczkowska et al., 2020; Katzenelenbogen et al., 2020). Differences in protein expression of LYVE1 and TREM2 were confirmed by immunofluorescence (Fig. S4A,B). KEGG pathway analysis revealed differences in lysosomal, inflammatory, metabolic, and protein synthetic processes between WNTi and WNTd macrophage populations (Fig. 2F).

We identified each macrophage cluster by the top DEG (Fig. 2B,G, Table S1). The WNTd culture contained five clusters of cells with distinct gene expression signatures, revealing the variety of fates these cells can take on. The TREM2^hi macrophage cluster expresses the WNTd signature described above, without an additional unique set of genes, suggesting that this may be the initial state of the WNTd macrophages. The cells in the metallothionein (MT) state express high levels of metallothionein 1 and 2 (MT1 and MT2) genes, which can promote inflammatory responses to pathogens (Subramanian Vignesh and Deepe, 2017). The amphiregulin (AREG) cluster expresses a set of genes (AREG, THBS1, RGS2, ABHD2) that have been shown to promote wound healing and angiogenesis in different tissue injury contexts (Hamidzadeh et al., 2020; Minutti et al., 2019; Zaiss et al., 2015; Kyriakides and Maclauchlan, 2009; Boelte et al., 2011; Jin et al., 2009). The WSB1 (WD repeat and SOCS box containing 1) cluster is also characterized by the expression of genes (WSB1, PNISR, VMP1, N4BP2L2) that can modulate immune responses and cellular function, including response to hypoxia and autophagy (Grasso et al., 2011; Haque et al., 2016). The MMP12 cluster is characterized almost entirely by expression of MMP12, a protease involved in macrophage migration, which is not expressed in the other clusters (Collison, 2018).

WNTi macrophages were predominantly contained in a single major cluster (LYVE1^hi) expressing the WNTi signature discussed above, with minor clusters (<5% of cells) also expressing higher levels of type 1 interferon stimulated genes (interferon state) or chemokines CCL3, CCL4 and CXCL8 (CCL3 state) (Fig. 2B,G). The WNTi macrophages also had a small cluster expressing increased levels of a subset of mitochondrially encoded genes (Fig. S4C). Analysis of variation in gene expression within the WNTd and WNTi macrophages revealed greater heterogeneity in gene expression in the WNTd macrophages, suggesting that WNTd macrophages can take on a larger number of gene expression states than the WNTi macrophages, reminiscent of the wide variety of fates recruited macrophages take on in response to tissue injury and disease (Fig. S4D) (Bajpai and Lavine, 2019; Bajpai et al., 2018). Taken together, these observations demonstrate that the WNTi and WNTd macrophages we obtain exhibit different intrinsic gene expression programs, and that these differences parallel those observed between macrophages derived from extra-embryonic and intra-embryonic hematopoietic programs in vivo.

In vivo, extra- and intra-embryonic macrophages follow distinct developmental trajectories, with only intra-embryonic myelopoiesis developing through a monocyte stage. As our scRNAseq analyses showed that WNTd, but not WNTi, macrophage cultures contained a monocytic-like precursor, we next investigated whether the hPSC myeloid populations we obtain exhibit a similar developmental progression. First, we analyzed the expression of the common monocyte and macrophage markers CD64 (FCGR1A) and CD14 on the developing macrophage cultures, from days 5 to 7, as these time points captured developmental intermediates not present in our terminally differentiated cultures. Both cultures exhibited an asynchronous maturational profile, consisting of CD64⁻CD14⁻, CD64⁺CD14⁻ and CD64⁺CD14⁺ populations (Fig. 3A). WNTd progenitors exhibited a progressive development through successive stages, first acquiring CD64 expression and then CD14, as suggested by their relative distribution over time. The distribution of these populations remained relatively static in the WNTi cultures at these time points, with a slight increase in CD64⁻CD14⁻ cells and a decrease in CD64⁺CD14⁻ cells, making it difficult to assess the relationships between these populations using these markers.

To characterize their development further, we assessed each population for the expression of IL3Rα (CD123), which is expressed on common myeloid progenitors (CMPs), and for CD45RA, which is expressed together with IL3Rα on downstream granulocyte/macrophage progenitors (GMPs; Manz et al., 2002). Consistent with our assessment of WNTd culture maturation, the CD64⁻CD14⁻ cells predominantly expressed only IL3Rα, as observed in CMPs, whereas the majority of the CD64⁺CD14⁻ cells expressed both IL3Rα and CD45RA, similar to GMPs. The fully committed CD64⁺CD14⁺ cells all expressed both IL3Rα and CD45RA (Fig. 3B). The pattern of IL3Rα and CD45RA expression differed, however, in the WNTi culture. The CD64⁻CD14⁻ population was predominantly negative for IL3Rα and CD45RA, whereas approximately half of the CD64⁺CD14⁻ cells expressed IL3Rα. Intriguingly, CD45RA was not detected in either CD14⁻ population, although the CD64⁺CD14⁺ macrophages expressed high levels of both IL3Rα and CD45RA (Fig. 3B). Exclusion of erythroid [CD235A (GYPA)⁺] and megakaryocyte (CLEC1B⁺) lineage cells increased the percentage of IL3Rα⁺ cells, but failed to increase the percentage of CD45RA⁺ progenitors (Fig. S5A). As this expression pattern suggested an absence of a canonical GMP-like stage, we also assessed an earlier time point of myeloid development. At day 3, fewer than 5% of the cells were CD64⁺CD14⁺ macrophages, indicating that the macrophage progenitors were still present in the culture. Analysis of the CD14⁻ cells showed that about 40% expressed IL3Rα and all were negative for CD45RA (Fig. S5B). These results suggest that the WNTi macrophage progenitors either do not pass through a canonical GMP progenitor state, or that they pass through this state too rapidly to be detected by flow cytometry.

To assess the differences between WNTd and WNTi progenitors further, we performed scRNAseq on each culture after 6 days in macrophage-inducing media. Based on our flow cytometry, we predicted that both cultures would contain a mixture of cells at different stages of differentiation at this time point. We obtained data on 4217 and 6994 cells for these WNTi and WNTd progenitor cultures, respectively. Both samples had comparable numbers of genes and UMIs identified per cell (Fig. S6A). Using unsupervised clustering of the WNTi and WNTd samples together, we identified 12 distinct clusters (Fig. 3C,D). We assigned cluster identities using markers characteristic of hematopoietic cell types and verified them by reference mapping our datasets to integrated human yolk sac and fetal liver scRNAseq data (Fig. 3E, Fig. S6B-D, Fig. S7A) (Bagger et al., 2019; Popescu et al., 2019). We also examined gene expression of IL3RA, FCGR1A and CD14 to correlate the scRNAseq data with our flow cytometry populations, and the proliferation markers MKI67 and TOP2A to ensure cell cycle regression was successful (Fig. S7B,C). At this early stage of differentiation, multiple cell types were present in both cultures, giving insight into the overall potential of the WNTd and WNTi progenitors. Both cultures contained clusters expressing canonical monocyte (PLAC8, LSP1, ASGR2, CORO1A) and macrophage (SPP1, VSIG4, C1QA, C1QB, C1QC) genes; the majority of the WNTd culture and a smaller fraction of the WNTi culture comprised these clusters (Fig. 3C-F, Fig. S6B,C, Table S2). None of these clusters expressed high levels of FUT4, the gene encoding the neutrophil marker CD15 (Fig. S7B). The macrophages present in the WNTi culture predominantly clustered into the Mac-1 cluster. These express higher levels of several genes (including LYVE1, SIGLEC1, LILRB5, F13A1, CD163 and DAB2) than the predominantly WNTd Mac-2 cluster, demonstrating that their gene expression signatures diverge early in macrophage maturation (Fig. 3C-F, Fig. S6C). Both cultures also had cells expressing characteristic eosinophil genes (CLC, IL5RA, EPX, PRG2, PRG3). Megakaryocyte (GP9, ITGA2B, PDLIM1 and PF4) and mast cell (KIT, TPSB2 and GATA2) clusters were present primarily in the WNTi culture (Fig. 3C-F, Fig. S6B,D). Similarly, the WNTi culture contained a large number of red blood cell (RBC) progenitors, expressing glycophorin A (GYPA), GATA1 and KLF, which were largely absent from the WNTd culture, consistent with an EPO-independent development of the primitive erythroid lineage (Malik et al., 2013). Only the WNTd culture contained a cluster expressing canonical dendritic cell genes (CD1C, FGL2, ITGB7, CLEC4A and HLA-DRA), as well as a CD34⁺ progenitor cluster that expresses characteristic intra-embryonic MPP genes (MECOM, SPINK2, MLLT3, HLF) (Figs S6D and S7D) (Kataoka et al., 2011; Zeng et al., 2019; Pina et al., 2008; Yokomizo et al., 2019). To verify that the WNTd culture represents intra-embryonic-like hematopoiesis, we reference mapped human fetal liver HSC/MPPs and yolk sac MPPs onto our WNTd progenitor dataset (Popescu et al., 2019). The fetal liver HSCs mapped strongly to the WNTd CD34⁺ progenitors, whereas the yolk sac MPPs did not (Fig. S7E). In particular, the human fetal liver HSC/MPPs and the WNTd CD34⁺ progenitors expressed HOXA9, which is absent in the human yolk sac progenitors and the WNTi cells (Fig. S7F) (Popescu et al., 2019). HOXA9 has previously been shown to be a crucial factor for generation of HSCs from endothelium (Zhou et al., 2016; Ng et al., 2016).

Interestingly, WNTi and WNTd cultures both contained cells expressing canonical GMP genes (ELANE, MPO, AZU1, PRTN3), despite the WNTi culture not having a classical CD45RA⁺ GMP population by flow cytometry (Fig. 3F, Fig. S6C). Thus, we further investigated the macrophage progenitor populations in each culture by pseudotime analysis using Palantir (Setty et al., 2019). We independently clustered the WNTd and WNTi progenitor datasets, assigned cluster identifications using the gene sets shown in Fig. S6, and superimposed pseudotime and entropy values (Fig. 4, Fig. S8A,B). We additionally plotted the probability of a cell differentiating into a macrophage or dendritic cell to identify the developmental trajectory of this lineage. Within the WNTd differentiation, CD34⁺ progenitors displayed the lowest pseudotime and greatest entropy values, consistent with these cells being the earliest population (Fig. 4A,B). Calculation of macrophage and dendritic cell potential showed the expected developmental trajectory from GMP to monocyte to macrophage/dendritic cell (Fig. 4C). Interestingly, we found two separate WNTd monocyte populations; monocyte-1 falls between the GMP and macrophage clusters, whereas monocyte-2 falls between the CD34⁺ progenitor and dendritic cell-like clusters (Fig. 4A, Fig. S8A). Monocyte-2 cells expressed several genes found in human dendritic cells, including ENHO, CD1D, ITGB7, FGL2 and CLEC10A, indicating that these monocytes are likely dendritic cell progenitors. This observation is consistent with the finding in mice that monocyte-derived dendritic cells come from a monocyte-dendritic cell progenitor (MDP) that arises directly from the CMP, rather than from the GMP (Yáñez et al., 2017). The monocyte-2 gene signature is absent from the WNTi culture, consistent with the lack of a WNTi dendritic cell-like population (Fig. S8B).

The WNTi clusters with the lowest pseudotime and highest entropy values included RBC progenitors and a population of potential myeloid progenitors that clustered between megakaryocytes and monocytes (progenitor 1, Fig. 4D,E). Calculation of macrophage potential suggested that progenitor 1 has the capacity to give rise to macrophages (Fig. 4F). This analysis also showed possible macrophage potential arising out of the eosinophil cluster, which contained cells expressing GMP genes (ELANE, MPO, AZU1, PRTN3) (progenitor 2, Fig. 4D,F,G). To determine whether these two progenitors can give rise to macrophages, we established a flow cytometry protocol to isolate these cell populations. Based on the scRNAseq data, we found that CXCR4 was expressed in both progenitor 1 and 2 (Fig. 4H). As CXCR4 is also expressed in the WNTd GMP cluster, it may be a useful marker of macrophage progenitors (Fig. S8C). The progenitor 1 cluster also expressed megakaryocyte marker CLEC1B. We isolated CD14⁻CXCR4⁺CLEC1B⁻ and CD14⁻CXCR4⁺CLEC1B⁺ cells from macrophage day 4 WNTi differentiations and cultured them in macrophage media for 7 days. The CXCR4⁺CLEC1B⁻ and CXCR4⁺CLEC1B⁺ cells both efficiently gave rise to CD14⁺ cells with similar morphology, confirming their macrophage potential (Fig. 4I, Fig. S8D). Addition of stem cell factor (SCF) to macrophage media supported mast cell differentiation from both progenitor 1 and 2 (Fig. S8E). Culture in megakaryocyte-inducing conditions gave rise to CD41⁺CD42b⁺ megakaryocytes from both progenitors, with a higher percentage of CD41⁺CD42b⁺ cells in the progenitor 1 culture (Fig. S8F). When cultured under conditions supporting granulocyte differentiation, progenitor 2 gave rise to CD32 (FCGR2)⁺CD15⁺CD16 (FCGR3)⁺ neutrophils, consistent with the expression of GMP genes in this cluster, whereas no CD32⁺ granulocytes arose from progenitor 1 (Fig. S8G,H).

To confirm that the WNTi cells expressing GMP genes were not as result of contamination from the WNTd pathway, we isolated KDR⁺CD235A⁺ cells on day 3 of a WNTi differentiation, which have previously been shown to have exclusively extra-embryonic-like hematopoietic potential (Fig. S9A) (Sturgeon et al., 2014). We also isolated KDR⁺CD235A⁻ cells from a WNTd culture, which are exclusively intra-embryonic-like, at the same time point. These cells were differentiated as before, then CD34⁺CD43⁺ WNTi cells and CD34⁺CD43⁻ WNTd cells were isolated on day 8 and cultured in macrophage media (Fig. S9B). After 6 days of macrophage differentiation, we analyzed MPO protein expression, as a marker of the GMP gene program, and found MPO⁺CD14⁻ cells in the WNTi and the WNTd cultures (Fig. S9C). Therefore, although WNTi progenitors may not pass through a canonical GMP, based on surface expression of CD45RA, the cells arising from progenitor 2 nevertheless develop through a transitional intermediate that expresses a GMP-like gene signature. Progenitor 1, in contrast, does not contain a GMP-like population, but instead has a gene expression profile comparable to the EMP and pre-macrophages (pMac) observed in mice (Fig. 4J,K, Table S3) (Mass et al., 2016).

We next sought to determine whether WNTi and WNTd macrophages exhibit functional differences. Tissue-resident macrophages express lower levels of inflammatory genes in response to stimuli, compared with monocyte-derived macrophages (Watanabe et al., 2019; Epelman et al., 2014a; Lavine et al., 2014). To determine whether WNTi and WNTd macrophages had intrinsically different inflammatory responses, we assayed levels of IL1B, TNF and IL6 gene expression following treatment with lipopolysaccharide (LPS). Upon stimulation with either 2.5 ng/ml or 25 ng/ml LPS, WNTi macrophages expressed significantly lower levels of IL1B and TNF than WNTd macrophages, whereas expression of IL6 was comparable between the two (Fig. 5A). Thus, in the context of LPS treatment, WNTi macrophages appear to be less inflammatory than WNTd macrophages.

As shown in Fig. 1, phagocytosis of E. coli particles was not different between WNTd and WNTi macrophages, but higher expression of CD36 on WNTi macrophages indicated possible increased phagocytic capacity for apoptotic cells. To test this hypothesis, we incubated WNTi and WNTd macrophages with pHrodo Red-labeled apoptotic mouse thymocytes. After 2 or 5 h of co-incubation, we assayed the number of thymocytes phagocytosed by each macrophage (Fig. 5B). At both time points, a higher percentage of WNTi macrophages had phagocytosed thymocytes, compared with the WNTd macrophages. Additionally, the mean number of thymocytes phagocytosed by each WNTi macrophage was higher, yielding a higher overall phagocytic index for apoptotic cells (Fig. 5B,C).

Production of reactive oxygen species (ROS) by macrophages is a crucial component of their role in clearance of pathogens; however, it can also modulate immune responses and promote polarization of macrophages towards a reparative phenotype (Yang et al., 2019). Thus, we wanted to determine whether there are intrinsic differences in ROS production between WNTi and WNTd macrophages. We stimulated WNTi and WNTd macrophages with either serum-opsonized zymosan (SOZ) particles or phorbol 12-myristate 13-acetate (PMA) and used a luminol-based assay to detect production of ROS. WNTi macrophages produced significantly more ROS than WNTd macrophages in response to SOZ, but there was no difference in response to PMA, suggesting a specificity for increased ROS production in WNTi macrophages in response to a particulate antigen (Fig. 5D). The NBT assay showed nearly 100% positivity in both genotypes, so was not repeated on each replicate. We also did not find differences in the expression of NADPH oxidase NOX2 (CYBB) or potential SOZ receptors Toll-like receptor 2 (TLR2), Mac-1 (ITGB2) or dectin (CLEC7A) between WNTi and WNTd macrophages (Fig. S10A). Intriguingly, it has previously been shown that mouse resident peritoneal macrophages have a sustained respiratory burst in response to zymosan that is dependent on generation of ROS by the mitochondrial respiratory chain (Davies et al., 2017). The peritoneal macrophages also had increased mitochondrial respiratory capacity, compared with bone marrow derived macrophages (BMDMs), that allowed for this sustained respiratory burst. Thus, we assayed mitochondrial function in the WNTi and WNTd macrophages by measuring oxygen consumption rate (OCR) during a mitochondrial stress test. WNTi macrophages had a higher basal OCR and showed higher spare respiratory capacity (SRC) upon electron transport chain uncoupling than WNTd macrophages (Fig. 5E,F). This phenotype was amplified upon treatment with LPS. A subset of mitochondrially encoded components of the electron transport chain (MT-ND4, MT-ND5, MT-CO2, MT-CYB) are expressed at higher levels in WNTi macrophages, compared with WNTd macrophages, which may contribute to this increased respiratory capacity (Fig. S10B). Collectively, these data indicate that WNTi and WNTd macrophages have inherent differences in metabolic and inflammatory capacity that are consistent with known functional distinctions between tissue-resident and monocyte-derived macrophages.

Tissue-resident macrophages are increasingly recognized as important contributors to embryonic development, tissue maturation, remodeling and repair. To date, few systems exist to investigate tissue resident macrophage specification, function, and applicability to regenerative medicine. This technological gap is most apparent in the study of human tissue-resident macrophages, which is largely restricted to ex vivo and transcriptomic analyses of explanted human specimens. Recent reports have demonstrated that macrophages derived from induced pluripotent stem cells through yolk-sac-like hematopoiesis can take on microglial characteristics when co-cultured with neurons, but these have not been compared with macrophages generated through intra-embryonic-like hematopoiesis (Takata et al., 2017; Haenseler et al., 2017). By leveraging known differences in the developmental origin of tissue-resident and monocyte-derived macrophages, we established an experimentally tractable in vitro system to model each of these macrophage populations. We demonstrate that macrophages derived from extra-embryonic-like and intra-embryonic-like hematopoietic progenitors have unique gene expression signatures, developmental trajectories, and functions that are consistent with in vivo characteristics of tissue-resident and monocyte-derived macrophages, respectively.

Differences between tissue-resident macrophages and monocyte-derived macrophages have been largely ascribed to the impacts of the tissue microenvironment (Lavin et al., 2014). Although the tissue environment is certainly a major component driving behavior of tissue-resident macrophages, we show intrinsic differences in transcriptional signatures and cellular function between extra- and intra-embryonic-like macrophage lineages. We revealed that macrophages derived from extra-embryonic-like hematopoiesis (WNTi) express a wide array of genes associated with general tissue-resident macrophage identity (Summers et al., 2020). These genes are expressed in multiple types of tissue-resident macrophages, as well as in human yolk-sac derived macrophages isolated prior to HSC formation, suggesting that the extra-embryonic macrophages that populate tissues during development are predisposed to this gene expression program (Bian et al., 2020; Popescu et al., 2019). Genes that are associated with a specific type of tissue-resident macrophage, such as Kupffer cell-associated CLEC4A and CD5L and Langerhans cell-associated CD207, were largely absent from both WNTi and WNTd macrophages (Summers et al., 2020).

WNTi and WNTd macrophages emerge from distinct progenitors, each with different functional capacity. The WNTd progenitors recapitulate definitive myelopoiesis, with eosinophil and megakaryocyte lineages completely separate from the monocyte and macrophage lineage, consistent with these cell types originating from distinct CMPs (Drissen et al., 2016, 2019). In contrast, the WNTi GMP-like progenitor (progenitor 2) also expressed eosinophil genes. Additionally, WNTi progenitor 1 expresses megakaryocyte genes, as has been described for yolk-sac-derived EMPs (Mass et al., 2016). This overlap between transcriptional programs in the WNTi progenitors may drive differences in gene expression and cellular function between WNTi and WNTd macrophages. The expression of MYB in both WNTi progenitors, the similarity of progenitor 1 to the mouse EMP, and the diversity of myeloid cell types present in the WNTi culture all support the conclusion that the WNTi culture recapitulates the EMP wave of yolk sac hematopoiesis that gives rise to the majority of fetal tissue-resident macrophages (Hoeffel et al., 2015).

Our results further demonstrate functional differences between WNTi and WNTd macrophages that may have relevance for the function of extra- and intra-embryonic-derived macrophages in vivo. Previous studies have shown that metabolic changes in macrophages can alter their inflammatory properties. The mitochondrial capacity we observe in WNTi macrophages recapitulates the phenotype of resident peritoneal macrophages, which are established during embryogenesis and self-renew during adulthood (Yona et al., 2013). These peritoneal macrophages have higher spare respiratory capacity than BMDMs, despite not having a greater number of mitochondria, and utilize different metabolites (Davies et al., 2017). Resident peritoneal macrophages also have higher oxygen consumption and higher ROS production in response to zymosan, and the increase in ROS was sensitive to inhibition of the mitochondrial electron transport chain (Davies et al., 2017). This study supports the possibility that metabolic differences underlie the observed increases in ROS production and spare respiratory capacity in WNTi macrophages.

In conclusion, we have established a human platform to generate distinct lineages of macrophages derived from extra-embryonic-like and intra-embryonic-like progenitors that resemble tissue-resident and monocyte-derived macrophages, respectively. This system is experimentally tractable, as it produces large numbers of cells from a renewable resource and is amenable to pharmacological and genetic manipulations. We envision that this platform will provide new opportunities to explore applications of human tissue-resident macrophages to regenerative medicine and to dissect the mechanisms that underlie tissue-resident macrophage specification, differentiation and function.

The human embryonic stem cell (hESC) line H1 (WA01; WiCell; male) was maintained on irradiated mouse embryonic fibroblasts in hESC media, as described previously (Sturgeon et al., 2014). For differentiation, hPSCs were cultured on Matrigel-coated plasticware (BD Biosciences) for 24 h, followed by embryoid body generation. Briefly, hPSCs were dissociated with trypsin-EDTA (0.05%) treatment for 1 min, and cells were detached by scraping to form small aggregates (six to ten cells). Embryoid bodies were resuspended in SFD (Sturgeon et al., 2012) supplemented with L-glutamine (2 mM), L-ascorbic acid (1 mM), monothioglycerol (MTG, 4×10⁻⁴ M), holo-transferrin (150 μg/ml) and BMP4 (10 ng/ml). After 24 h, bFGF (5 ng/ml) was added to the media. At 42 h of differentiation, activin A (1 ng/ml) and either CHIR99021 (3 μM) and SB-431452 (6 µM) (to specify WNT-dependent hematopoietic progenitors) or IWP2 (3 μM) (to specify WNT-independent hematopoietic progenitors) were added to the culture. At 72 h of differentiation, embryoid bodies were washed with IMDM (Thermo Fisher) to remove WNT modulating small molecules and placed in StemPro-34 (Thermo Fisher) supplemented with L-glutamine (2 mM), ascorbic acid (1 mM), monothioglycerol (MTG; 4×10⁻⁴ M), holo-transferrin (150 μg/ml), VEGF (15 ng/ml) and basic fibroblast growth factor (bFGF) (5 ng/ml). On day 6 of differentiation, IL6 (10 ng/ml), IL11 (5 ng/ml) and SCF (50 ng/ml) were added to the culture. Cultures were maintained in a low-oxygen, 5% CO₂/5% O₂/90% N₂ incubator. All human recombinant factors and small molecules were purchased from R&D Systems. On day 8 of differentiation, hematopoietic progenitors were isolated by FACS on a BD FACSAria. CD34-PE/Cy7 (clone 8G12, 348791), CD43-FITC (clone 1G10, 555475), CD184-APC (clone 12G5, 555976) and CD73-PE (clone AD2, 550257) antibodies were all purchased from BD Biosciences. Human CB CD34⁺ progenitors were isolated using CD34-APC (clone 8G12, 340441) from BD Biosciences.

Isolated WNT-dependent progenitors (CD34⁺CD43⁻), WNT-independent progenitors (CD34⁺CD43⁺) and umbilical CB progenitors (CD34⁺) were added to Matrigel-coated wells of a 24-well plate in StemPro-34 media (supplemented with L-glutamine, ascorbic acid, MTG and holo-transferrin as above) containing TPO (30 ng/ml), BMP4 (10 ng/ml), IL6 (10 ng/ml), IL11 (5 ng/ml), FLT3L (10 ng/ml), M-CSF (10 ng/ml), bFGF (5 ng/ml) and VEGF (5 ng/ml). These cells were cultured for 14 days to generate mature macrophages. For generation of additional myeloid cell types from WNTi P1 and P2, cells sorted on Mac Day 4 were cultured in media containing 50 ng/ml SCF (mast cells); 50 ng/ml G-CSF, 25 ng/ml IL3 and 1 ng/ml IL5 (granulocytes); or 50 ng/ml TPO and 25 ng/ml SCF (megakaryocytes). Granulocyte formation was also assessed by seeding 1×10⁴ P1 and P2 cells in 10mL MethoCult H4034 (STEMCELL Technologies) for 9 days.

WNTd and WNTi myeloid progenitors were isolated by harvesting all non-adherent cells on days 5-7 of macrophage differentiation. Cells were analyzed by flow cytometry using CD123-APC (clone 6H6, 306011), CD45RA-APC/Cy7 (clone HI100, 304127), CD64-FITC (clone 10.1, 305006) and CD14-PE (clone M5E2, 301806) antibodies from BioLegend. For characterization of macrophage progenitor potential, cells were harvested on day 4 of macrophage differentiation as above, then isolated by FACS using CXCR4-APC (clone 12G5, 306509) and CLEC1B-FITC (clone AYP1, 372007) antibodies from BioLegend. Other myeloid cell types were analyzed using KIT-BV421 (clone 104D2, BioLegend 313215), CD15-APC (clone HI98/HIM1, BD Biosciences 551376), CD16-PerCP-Cy5.5 (clone 3G8, BD Biosciences 560717), CD32-PE (clone FUN-2, BioLegend 303206), CD41a-PE (clone HIP8, BD Biosciences 555467) and CD42b-APC (clone HIP1, BioLegend 303912). For analysis of MPO expression, cells were harvested on day 6 of macrophage differentiation, fixed and permeabilized using Fix&Perm Cell Permeabilization Kit from Thermo Fisher (GAS004), then stained with MPO antibody (clone MPO455-8E6, 12-1299-42) from Thermo Fisher. Mature macrophages were characterized using CD45-PerCP/Cy5.5 (clone HI30, 304027) and CD14-APC/Cy7 (clone 63D3, 367108) from BioLegend, and MerTK-PE (clone 125518, FAB8912P) from R&D Systems. Mouse IgG1-PE isotype control for MerTK staining was 400111 from BioLegend.

All non-adherent cells were harvested on day 6 or day 14 of macrophage differentiation and prepared for scRNAseq using the 3′v2 kit from 10X Genomics, following their published protocol; 10,000 cells per sample were put into the sequencing pipeline. Samples were sequenced on a HiSeq 3000 (Illumina) and read alignment was performed using the 10X Genomics Cell Ranger pipeline (Zheng et al., 2017). Downstream analyses were performed using Seurat R software package version 4.0 (http://satijalab.org/seurat/).

The following pipeline was used for the mature macrophage (day 14) dataset: The Cell Ranger counts matrix was filtered using the following QC cutoffs: (1) genes expressed in fewer than three cells were discounted; (2) cells containing fewer than 200 or greater than 5000 genes were removed as these could be potential doublets; (3) cells with a mitochondrial content of >7.5% were removed as these are potentially dying cells. Upon filtering, normalization, variance stabilization, and scaling were performed with mitochondrial percent regression using sctransform to find 3000 highly variable genes. Principal component analysis was used to find nearest neighbors and a 2D UMAP embedding was computed for all visualization purposes. DEGs were identified using the FindAllMarkers function between clusters and WNTi and WNTd macrophages. Statistically significant genes (adjusted P<0.05) were used to annotate clusters. DEGs between WNTi and WNTd macrophages were used to perform gene set enrichment pathway analyses using WebGestalt with the KEGG functional database (http://www.webgestalt.org/).

The following pipeline was used for the progenitor (day 6) dataset: The Cell Ranger counts matrix was filtered using the following QC cutoffs: (1) genes expressed in fewer than three cells were discounted; (2) cells containing fewer than 200 or greater than 7000 genes were removed as these could be potential doublets; (3) cells with a mitochondrial content of >7.5% were removed as these are potentially dying cells. Upon filtering, the following was carried out for WNTd+WNTi merged, WNTd, and WNTi datasets: normalization, variance stabilization, and scaling was performed with mitochondrial percent/cell cycle regression using sctransform to find 3000 highly variable genes. Principal component analysis was used to find nearest neighbors and a UMAP embedding was computed for all visualization purposes. The FindAllMarkers function in Seurat was used to perform differential gene expression testing between clusters and canonical genes were used for annotation as highlighted by violin plots, heatmaps and feature plots. Additionally, GeneSet z-scores were calculated for groups of genes to highlight distinct cell states. All scripts used to analyze the data herein are available on GitHub (https://github.com/jamrute/2021_MacDiff_Bredemeyer_Lavine).

Cell types present in each cluster were determined by comparing gene expression signatures with profiles in BloodSpot (Bagger et al., 2019) and the Human Protein Atlas (https://www.proteinatlas.org/).

Palantir was used to perform all pseudotime trajectory analyses in WNTd and WNTi datasets. sctransform normalized and scaled counts from the 3000 most highly variable genes were used as an input into Palantir (https://github.com/dpeerlab/Palantir). For the WNTd dataset, the CD34⁺ progenitor cluster was used as the starting state and for the WNTi dataset, the progenitor 1 cluster was used as the starting state.

The human fetal liver and yolk sac datasets from Popescu et al. (2019) were integrated into a single data space using Harmony (Korsunsky et al., 2019). The HSC/MPP clusters from both datasets were then reference mapped onto the WNTd and WNTi progenitor datasets using a Seurat v4 pipeline.

Cells were prepared and imaged by transmission electron microscopy as described by Dege et al. (2020). Antibodies recognizing CD68 (14-0688-82), TREM2 (PA5-111856) and LYVE1 (MA5-32512) were all from Thermo Fisher. Protein expression was quantitated on a per cell basis using integrated density of sum projections of 40× confocal z-stacks. Phagocytosis by hPSC-derived macrophages was measured using pHrodo Red E. coli beads (Thermo Fisher). Macrophages were cultured with pHrodo Red beads for 1 h, then analyzed for percentage of fluorescent cells by microscopy and intensity of fluorescence by flow cytometry. The assay for phagocytosis of apoptotic thymocytes was performed as described by Miksa et al. (2009), with co-incubation times of 2 h and 5 h in 4-well chamber slides (Thermo Fisher). Slides were washed with PBS three times to remove non-attached thymocytes, fixed, and stained with anti-CD68 and anti-CD14 (PA5-13305, Thermo Fisher). Expression of inflammatory genes in hPSC-derived macrophages was measured by qRT-PCR after 4 h of treatment with either 2.5 or 25 ng/ml LPS. RNA was isolated using the PureLink RNA Mini Kit (Thermo Fisher) and cDNA was generated using the High-Capacity cDNA Reverse Transcription Kit (Thermo Fisher). qRT-PCR was performed using PowerUp SYBR Green (Thermo Fisher) and human β2 microglobulin (B2M) was used as the control gene for dCt comparisons. Primer sequences are as follows: B2M-f (ACTTTGTCACAGCCCAAGATAG); B2M-r (GCAAGCAAGCAGAATTTGGAA); IL1B-f (ATGGACAAGCTGAGGAAGATG); IL1B-r (CCCATGTGTCGAAGAAGATAGG); TNFa-f (ACTTTGGAGTGATCGGCC); TNFa-r (GCTTGAGGGTTTGCTACAAC); IL6-f (AAATTCGGTACATCCTCGACGGCA); IL6-r (AGTGCCTCTTTGCTGCTTTCACAC) (f, forward; r, reverse). Analysis of ROS production was performed on 5×10⁴ cells/well using lucigenin as detailed by Bagaitkar et al. (2017). Mitochondrial analysis was performed on a Seahorse XF96 (Agilent), using 7.5×10⁴ cells/well, and 1.5 µM oligomycin, 0.5 µM FCCP, 0.5 µM rotenone and 0.5 µM antimycin A, as indicated in the Seahorse XF Mitochondrial Stress Test Kit protocol. LPS treatment was 10 ng/ml for 18 h. Unpaired Student's t-test was used for statistical analyses, and for all experiments n=3.

All scRNAseq samples were prepared and sequenced at the McDonnell Genome Institute at the Washington University School of Medicine. Flow cytometry and FACS sorting were carried out at the Flow Cytometry and Fluorescence Activated Cell Sorting Core Facility at the Washington University School of Medicine. Electron microscopy was performed at the Washington University Center for Cellular Imaging. The authors thank the Haniffa lab for sharing their human scRNAseq datasets.

The peer review history is available online at https://journals.biologists.com/dev/article-lookup/doi/10.1242/dev.200016.