Potently cytotoxic natural killer cells initially emerge from erythro-myeloid progenitors during mammalian development

Summary

Natural Killer (NK) cells are a critical component of the innate immune system. However, their ontogenic origin has remained unclear. Here we report that NK cell potential first arises from Hoxa^neg/low Kit⁺CD41⁺CD16/32⁺ hematopoietic stem cell (HSC)-independent erythro-myeloid progenitors (EMP) present in the murine yolk sac. EMP-derived NK cells and primary fetal NK cells, unlike their adult counterparts, exhibit robust degranulation in response to stimulation. Parallel studies using human pluripotent stem cells (hPSCs) revealed that HOXA^neg/low CD34⁺ progenitors give rise to NK cells that, similar to EMP-derived NK cells, harbor a potent cytotoxic degranulation bias. In contrast, hPSC-derived HOXA⁺ CD34⁺ progenitors, as well as human cord blood CD34⁺ cells, give rise to NK cells that exhibit an attenuated degranulation response, but robustly produce inflammatory cytokines. Collectively, our studies identify an extra-embryonic origin of potently cytotoxic NK cells, suggesting that ontogenic origin is a relevant factor in designing hPSC-derived adoptive immunotherapies.

Graphical Abstract

eTOC blurb

NK cell potential is thought to arise from lymphoid progenitors; however, in parallel studies of murine embryos and human pluripotent stem cells, Dege et al. demonstrate that NK cells with a potent cytotoxic degranulation response arise from erythro-myeloid progenitors.

Introduction

Natural Killer (NK) cells are innate lymphocytes that recognize and kill virally infected cells and malignant cells, making them a critical component of innate immunity and a highly desirable cell-type for adoptive immunotherapy (Miller et al., 2005). Human pluripotent stem cell (hPSC)-derived NK cells offer the possibility of uniform activity as a renewable “off-the-shelf” allogeneic cell product (Veluchamy et al., 2017). Interestingly, hPSC-derived NK cells are potently cytolytic compared to their donor-derived counterparts (Hermanson et al., 2016; Knorr et al., 2013; Woll et al., 2009). As the differentiation of hPSCs recapitulates early embryonic development (Murry and Keller, 2008), we asked whether hPSC-derived NK cells are ontogenically distinct from their adult counterparts.

In the adult, NK cells are derived from common lymphoid progenitors (CLP), which lie downstream of hematopoietic stem cells (HSCs; (Galy et al., 1995; Kondo et al., 1997; Rosmaraki et al., 2001)). NK cells and their progenitors have been found in the mid-gestation fetal liver of both mice and humans (Boiers et al., 2013; Renoux et al., 2015; Tang et al., 2012), however, the developmental origin of NK cells remains unclear. Intriguingly, there is evidence to suggest a possible HSC-independent origin for NK cells during embryogenesis, since explants of early-stage human yolk sac can generate CD15⁺ myeloid and CD56⁺ NK cells, but not CD19⁺ B cells or CD4⁺ T cells (Tavian et al., 2001). However, as the possible maternal contribution to these lineages is unclear, we utilized the tractable murine embryo as a model of mammalian hematopoietic development. The early ontogeny of the hematopoietic system in the murine embryo is characterized by overlapping waves of progenitors that arise in the extra-embryonic yolk sac prior to the emergence of HSCs intra-embryonically at E10.5 (Medvinsky and Dzierzak, 1996). These HSC-independent progenitors include an initial “primitive” wave of primitive erythroid, megakaryocyte and macrophage lineages, followed by a second wave of “erythro-myeloid progenitors” (EMP), containing definitive erythroid, megakaryocyte and multiple myeloid lineages, but not B cell potential when cultured in vitro (McGrath et al., 2015a; Palis et al., 1999). Here, we demonstrate that the in vitro culture of EMP isolated from the yolk sac of E9.5 mouse embryos contain robust NK cell potential. In addition, in vivo lineage tracing studies support the concept that EMP generate the first NK cells in the fetal liver. We have also determined that EMP-derived and primary fetal NK cells are larger and more biased to cytotoxic degranulation than their adult counterparts.

To study human hematopoietic development, we and others have developed an in vitro hPSC differentiation model system (Ditadi et al., 2017). This approach employs stage-specific WNT signal manipulation to pattern mesoderm into distinct waves of CD34⁺ progenitors, recapitulating extra- and intra-embryonic-like hematopoiesis. Briefly, through a WNT-independent (WNTi-) process, HOXA^neg/low CD34⁺ cells are generated, which harbor restricted extra-embryonic-like hematopoietic potential. Conversely, through a WNT-dependent (WNTd-) process, HOXA⁺ CD34⁺ cells are generated, which contain intra-embryonic-like multilineage erythro-myelo-lymphoid potential (Ditadi et al., 2015; Ng et al., 2016; Sturgeon et al., 2014). As both the WNTi- and WNTd-derived CD34⁺ populations harbor robust NK cell potential (Sturgeon et al., 2014), this model system similarly supports the hypothesis that human NK cells have two developmental origins. Using both the murine and hPSC systems, we have identified the emergence of a conserved, HSC-independent NK cell population from erythro-myeloid progenitors. This fetal NK cell population is uniquely biased for cytotoxic degranulation, making it functionally distinct from adult lymphoid progenitor-derived NK cells, which are instead biased for inflammatory cytokine production.

Results

NK cell potential in the murine embryo initiates in the yolk sac within EMP

As the first blood cells in the murine conceptus originate from HSC-independent hematopoietic progenitors in the extra-embryonic yolk sac, we first asked if the early yolk sac also contains NK cell potential. Dissociated E8.5–E10.5 yolk sac cells were cultured in vitro on OP9 stroma supplemented with FLT3-L, SCF, IL-2, IL-7, and IL-15 for 14 to 20 days, using Lin⁻Kit⁺IL7R⁺ common lymphoid progenitors (CLPs) isolated from adult bone marrow as a positive control. All populations efficiently gave rise to CD3ε⁻CD19⁻CD122⁺NK1.1⁺ cells (Figure 1A). Further, these cells were negative for CD127 and positive for CD49b (DX5) surface expression, and upregulated Id2 and Eomes expression during their in vitro differentiation, consistent with their NK cell identity (Figures S1A–C).

Figure 1: NK cell potential first arises from erythro-myeloid progenitors (EMP) in the yolk sac.

(A) Representative flow cytometric analyses of the immunophenotypic identification of NK cells cultured from adult mouse bone marrow common lymphoid progenitors (Adult CLP; Lin⁻Kit⁺CD127⁺) or from E9.5 yolk sac (E9.5 YS; see S1A) on OP9 stromal cells. (B) NK cell potential arises from cultured E8.5, E9.5, and E10.5 yolk sac tissues. YS-whole yolk sac, Prim-primitive hematopoietic progenitors, EMP (E9.5 YS, Kit⁺CD16/32⁺CD41⁺), Not-EMP-Kit+ yolk sac cells excluding CD16/32⁺CD41⁺ cells. (C) Limiting dilution assay of NK cell potential from sorted E9.5 EMPs. (D) NK cells are present in fetal liver as early as E13.5 and increase in numbers over the next two days (NK1.1⁺CD49b⁺Ter119⁻CD117⁻CD3ε⁻CD19⁻CD127⁻, see S1C), n=3. (E) Representative flow cytometric analyses of higher scatter properties of primary fetal liver and EMP-derived NK cells (purple) compared to their adult counterparts (black). Images of Wright Giemsa-stained NK cells derived from E9.5 EMP (left) or adult CLP (right; scale bar 10 μm, 100X magnification). (F) Quantification of side scatter (SSC) of NK cells isolated from E15.5 fetal liver or from adult spleen (in vivo), and derived from the in vitro culture of E9.5 EMP or of adult CLP, n=3. (G) Representative flow cytometric analyses of degranulation, measured by CD107a expression, of PMA-activated primary murine NK cells isolated from E15.5 fetal liver or from adult spleen (in vivo), and derived from E9.5 EMP or Adult-CLP (in vitro). (H) Quantification of NK-cells that express CD107a after PMA/ionomycin-activation. n≥3. (I) Median fluorescence intensity (MFI) of CD107a on CD107a⁺ PMA-activated NK cells isolated from E15.5 fetal liver (in vivo) or derived from E9.5 EMP (in vitro), relative to MFI of CD107a on CD107a⁺ PMA/ionomycin-activated NK cells isolated from adult spleen or derived from adult CLP (black line=1), n=3. Mean ± SEM. *p <0.01, **p<0.001, ***p<0.0001. Paired student’s t-test. See also Figure S1.

We next asked whether this yolk sac-derived NK cell potential was associated with the primitive or EMP waves of hematopoiesis. Progenitors of the primitive and EMP programs can be distinguished at E8.5 by their differential expression of kit and CD41 (McGrath et al., 2015a). In vitro culture of either E8.5 kit^lowCD41^lowCD16/32⁻ primitive hematopoietic progenitors, or kit⁺CD41⁺ EMP, indicated that NK cell potential is found within EMP (Figure 1B). We then asked whether this NK cell potential at E9.5 was restricted to kit⁺CD41⁺CD16/32⁺ EMP, or also found in the remaining “non-EMP” fraction of the E9.5 and E10.5 yolk sac, which specifically contains rare B-cell potential (McGrath et al., 2015a). Surprisingly, ex vivo culture of these cell fractions revealed that NK cell potential was restricted to EMP (Figure 1B). Furthermore, limiting dilution analyses indicated that 1/41 EMP harbor NK cell potential, which suggests that approximately 30 NK cell progenitors exist within EMP in the yolk sac of the E9.5 mouse conceptus (Figure 1C; (McGrath et al., 2015a)).

EMP seed the fetal liver and differentiate into definitive erythroid, megakaryocyte and multiple myeloid lineages (Gomez Perdiguero et al., 2015; Lux et al., 2008; McGrath et al., 2015a). We therefore examined mouse embryos for emerging NK cells and detected a population of CD3ε⁻CD19⁻CD117⁻CD127⁻CD49b⁺NK1.1⁺ cells as early as E13.5 in the fetal liver. This population progressively increased in numbers in the liver over the ensuing 48 hours of gestation (Figure 1D, S1D). Consistent with their NK cell identity, these cells express NK cell transcripts for Eomes and Prf1 (Perforin), and contain the granule protein Granzyme B, albeit at lower levels than adult NK cells (Figure S1E,F). Phenotypic analyses revealed that primary E15.5 fetal liver NK cells, as well as E9.5 EMP-derived NK cells, were larger and more granular than adult splenic NK cells or adult CLP-derived NK cells (Figure 1E,F).

When NK cells are stimulated, they rapidly undergo cytotoxic degranulation, expressing CD107a on their cell surface (Alter et al., 2004). We assessed the function of primary E15.5 fetal NK cells, as well as E9.5 EMP-derived NK cells, comparing their degranulation response to primary adult splenic NK cells and adult marrow CLP-derived NK cells, respectively (Figure 1G). Importantly, both the primary fetal NK and EMP-derived NK cell populations exhibited a robust degranulation response, with significantly more of these cells undergoing degranulation than their adult counterparts (Figure 1G,H). Further, the degree of degranulation per cell was far greater in the fetal NK cells, as evidenced by greater CD107a signal intensity (Figure 1I).

The morphological and functional similarities between E9.5 yolk sac EMP-derived NK cells and primary E13.5 – E15.5 fetal liver NK cells suggested that EMP give rise to fetal NK cells. To test this hypothesis, we utilized the tamoxifen-inducible Csf1r^MerCreMerRosa26^YFP mouse model, which has been used to lineage trace the macrophage potential of EMP (Gomez Perdiguero et al., 2015; Hoeffel et al., 2015). Tamoxifen was injected at E9.5 to label EMP, and multiple hematopoietic populations were subsequently analyzed in the fetal liver of E15.5 mouse embryos. YFP⁺ (CD3ε⁻CD19⁻CD127⁻CD122⁺NK1.1⁺CD49b⁺) NK cells were detected at similar levels as Ly6G⁺ granulocytes (Figures 2A–C, S2A,B). These YFP⁺ cell populations were also detected in the liver of E14.5 embryos (Figure 2D,E). To determine if this lineage tracing strategy was labeling not only EMP, but also pre-HSCs, we assayed LT-HSCs (Lin-Kit⁺Sca1⁺CD150⁺CD48⁺) and found no YFP⁺ LT-HSC in over 4,000 LT-HSCs analyzed in the livers of 11 fetuses (Figures 2B, S2B). Together, these data indicate that EMP in the yolk sac serve as an initial source of fetal NK cells, which are phenotypically and functionally distinct from their adult counterparts.

Figure 2: Lineage tracing supports an HSC-independent source of fetal NK cells.

(A) Experimental design for lineage tracing of hematopoietic cell populations in fetuses of Csf1r^MerCreMer X Rosa26^YFP mice following a single dose of tamoxifen at E9.5. (B) Gating strategy for detection of YFP+ NK cells (CD3ε⁻CD19⁻CD127⁻CD122⁺NK1.1⁺CD49b⁺) in the livers of E15.5 fetuses. Examples of fluorescence minus one (FMO) controls for NK1.1 and CD49b are also shown. (C) Quantification of lineage labeling of NK cells (as in B), as well as long-term hematopoietic stem cells (LT-HSCs-lineage⁻Kit⁺Sca1⁺CD150⁺CD48⁻), erythroid (Ter119+) cells, granulocytes (Ly6G+), monocytes (Ly6C+), and macrophages (F4/80+) (as in S2A,B). Statistical differences between %YFP in experimental (n=5) compared to littermate controls lacking Csf1r^MerCreMer n=3. Mean ± SEM. (D) Gating strategy detecting YFP+ NK cells (CD3ε⁻CD19⁻CD127⁻CD122⁺NKp46⁺) in the livers of E14.5 fetuses. (E) Quantification of lineage labeling of NK cells (as in D), and other hematopoietic lineage cells (as in S2 A,B) in the E14.5 liver. Statistical difference between %YFP in experimental (n=6) compared to littermate controls (n=4). Mean ± SEM. Unpaired two tail student’s t-test, *p<0.05, **p<0.01, ***p<0.001. See also Figure S2.

Stage-specific differentiation of hPSCs recapitulates 2 distinct ontogenic origins of NK cells

hPSC-derived NK cells have been described as functionally distinct, with a greater cytotoxic degranulation response when compared to donor-derived NK cells (Woll et al., 2009). As this was reminiscent of our observed differences between murine fetal EMP-derived and adult CLP-derived NK cells, we asked whether NK cell emergence from hPSCs recapitulates the ontogenic differences found in the murine system. Indeed, by manipulating WNT signaling solely during the development of mesendoderm, and not thereafter on hematopoietic progeny (Figure S3A), we previously described two ontogenically distinct origins for hPSC-derived NK cells (Sturgeon et al., 2014). Since the phenotypic and functional properties of these two hPSC-derived NK cell populations remained unclear, we sought to better characterize each hPSC-derived population (Figure S3A,B). As expected, HOXA1-9 expression was enriched in WNTd CD34⁺ multipotent progenitors (Figure S3C), recapitulating the preferential expression of medial HOXA cluster genes in intra-embryonic-like (WNTd) versus extra-embryonic-like (WNTi) hematopoietic progenitors (Dou et al., 2016; Gao et al., 2018; Ng et al., 2016). Consistent with this paradigm, we found an analogous lack of Hoxa gene expression in primary E9.5 murine EMP, when compared to adult marrow CLP (Figure S3D).

WNTi HOXA^neg CD34⁻CD43⁺ cells (P4) harbored predominantly primitive erythroid potential (Figure S3E) and lacked detectable CD45⁺ lymphoid potential (NK or T cell potential; Figure S3F,G), consistent with extra-embryonic-like primitive hematopoiesis. Similarly, the WNTi HOXA^neg/low CD34⁺CD43^+/− populations (P1–3) harbored erythro-myeloid potential, albeit with granulocyte potential (Figure S3E), which is not consistent with primitive hematopoiesis in the murine system (McGrath et al., 2015b). P1–3 also harbored NK cell, but not T lymphocyte, potential (Figure S3F,G). Curiously, this WNTi-derived NK cell (WNTi-NK) potential was always coincident with robust CD15⁺ granulocyte potential (Figure 3A), suggesting that similar to murine EMP-derived NK cells, WNTi-derived NK cells emerge from an erythro-myeloid biased progenitor. In contrast, WNTd HOXA⁺ CD34⁺ cells, as well as primary cord blood CD34⁺ cells, could efficiently give rise both to NK cells and to T lymphocytes (Figure S3F,G). Interestingly, as found in the murine system, extra-embryonic-like WNTi-derived NK cells were more granular than intra-embryonic-like WNTd-derived NK (WNTd-NK) or cord blood-derived NK cells (CB-NK; Figure 3B–C).

Figure 3: Human WNT-independent hematopoietic progenitors have NK cell potential.

(A) Representative flow cytometric analyses of CD56⁺ NK cell and CD15⁺ granulocyte potential of WNTi CD34⁺43⁺, WNTd CD34⁺ and cord blood CD34⁺ progenitors, n=12. (B) Quantification of geometric mean fluorescence intensity (MFI) of SSC from CD56⁺ WNTi-derived NK, WNTd-derived NK, and CB-DERIVED NK cells. n=55 (WNTi-NK), n=56 (WNTd-NK) and n=52 (CB-NK). Mean ± SEM. (C) Morphology of WNTi-, WNTd-, and CB-NK CD56⁺ and WNTi-CD15⁺ cells, obtained as in Supplemental Figure 3A, under transmission electron microscopy (scale bar=1 μm, 5,000X magnification). (D) Representative flow cytometric analyses of CD56 and CD3ε expression on Week 4 of OP9-DL4 co-cultures, gated on Live (DAPI⁻) singlets. (E) Representative flow cytometric analyses of CD94 and CD16 expression of CD56⁺ hPSC-derived WNTi-NK, WNTd-NK and CB-NK cells, over time. (F) Quantification of CD94 and CD16 expression on each CD56⁺ cell population, as shown in E. Week 2, n=5, Week 3, n=11, Week 4 n=58 (WNTi-derived NK and WNTd-derived NK) and n=54 (CB-NK), Week 5, n=4. Mean ± SEM. Gated on Live (DAPI⁻) singlet, CD56⁺ cells. (G) Representative histogram of CD16 expression on WNTi-derived NK (purple), WNTd-derived NK (blue), and CB-derived NK (black). Gated on CD56⁺ cells. Mean ± SEM. Paired and unpaired student’s t-test. *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001. See also Figure S3,S4.

Each in vitro-derived NK cell population was CD56⁺CD3ε⁻ (Figure 3D), and exhibited no significant differences in the expression of common NK cell surface markers (Figure S4A). In vitro-derived CD56⁺ populations emerged from all sources at comparable rates (Figures 3D–F). However, WNTi-derived NK cells were consistently enriched for CD16 expression in comparison to WNTd-derived NK and to CB-derived NK cells (Figures 3E–G). Whole-transcriptome analyses revealed that all in vitro CD56⁺ populations, as well as CD34⁺ progenitors, expressed multiple transcription factors associated with NK cell specification and differentiation, and displayed significant similarities in gene expression, but with some ontogeny-specific differences (Figure 4A–C). All CD56⁺ populations expressed genes for NK cell-associated surface antigens and cytolytic proteins, as well as 13 “core signature” human NK cell genes (Figure 4B; (Crinier et al., 2018)). While each population displayed low levels of KIR cell surface expression (Figure S4B), there was significant KIR transcript expression, which varied across ontogenic origin (Figure 4B). Finally, each population displayed robust transcript and protein expression of GZMA, GZMB (Granzymes A/B) and PRF1 (Perforin; Figure 4B,C, S4C). However, both hPSC-derived populations did harbor enriched GZMK (Granzyme K) expression in comparison to cord blood-derived NK cells (Figure 4C, S4C). Collectively, these analyses strongly suggest that these in vitro-derived populations consist of bona fide NK cells.

Figure 4: Whole transcriptome sequencing of WNTd-derived NK, WNTi-derived NK, and CB-DERIVED NK cells.

(A-C) Heat maps showing gene expression within WNTi and WNTd CD34⁺ progenitors, and CD56⁺ WNTi-derived NK, WNTd-derived NK, and CB-derived NK cells, of (A) Heat map of genes (RPKM >1 in each group) enriched within WNTi-derived NK cells (279), WNTd-derived NK (100), common to hPSC-NK cells (274), CB-derived NK genes (236), genes shared by either WNTd -and CB-NK (100) or WNTi- and CB-NK (70), and all NK cell populations (1043). Dendrogram was constructed using one minus Pearson correlation with averaged linkage. Color scale is row dependent, with highest RPKM set to red and lowest RPKM set to blue. (B) Heat map of transcription factors associated with NK cell specification and development, NK cell “signature” genes, and Killer-cell immunoglobulin-like receptors (KIRs). (C) Heat map of common NK cell surface receptors and cytolytic enzymes and proteins required for NK cell-mediated effector functions (upper) and differential gene expression (lower) between NK cell populations. Each square represents the average RPKM of each gene per sample group, n=3. (D-E) Gene Set Enrichment Analysis (GSEA) of WNTi-derived NK cells compared to WNTd-derived NK cells. (D) Quantification of False Discover Rate (FDR) q-value correlate transcriptional and functional differences observed in WNTi-derived NK and WNTd-derived NK cells. (E) Select enrichment plots from gene sets significantly enriched in WNTi-derived NK (purple border) or WNTd-derived NK cells (blue border). See also Figure S4.

Gene Set Enrichment Analysis (GSEA) revealed that WNTi-derived NK cells were enriched in genes associated with myeloid differentiation and innate immune responses (Figure 4D,E), whereas WNTd-derived NK cells were enriched in hallmark pathways associated with inflammatory responses. Concordant with the GSEA, WNTd-derived NK cells exhibited higher expression of inflammatory response and CD56^bright-associated genes (IFNG, IL23A, IL2RA and GPR183; Figure 4C; (Allan et al., 2017)), suggesting that WNTd-derived NK and WNTi-derived NK cell populations may differ in their functional properties. Collectively, these observations indicate that hPSCs give rise to two ontogenically distinct NK cell populations, each of which share phenotypic characteristics conserved across species.

Ontogeny dictates the function of hPSC-derived NK cells

Given the striking cytotoxic degranulation capacity of murine primary fetal NK cells and EMP-derived NK cells (Figure 1G–I), we asked whether hPSC-derived WNTi-NK or WNTd-derived NK cells are functionally distinct. We stimulated each population with either K562 cell co-culture, antibody-dependent cellular cytotoxicity (ADCC) mechanism, or PMA/ionomycin (Figure 5A–F, S5A–C). Interestingly, WNTi-derived NK, WNTd-derived NK and CB-derived NK cells exhibited differential effector responses (Figure S5A–C). WNTd-derived NK cells exhibited robust IFN-γ production when stimulated by K562 tumor cells or PMA/ionomycin, while WNTi-derived NK cells exhibited a robust degranulation response across all assays. Indeed, this degranulation response was 3-fold stronger for a degranulation response over IFN-γ production. In contrast, WNTd-derived NK cells were consistently biased for IFN-γ production over degranulation (Figure 5A–F). Consistent with a bias for cytotoxic degranulation, WNTi-derived NK cells were superior at killing Raji tumor targets across a spectrum of effector to target ratios (E:T), in comparison both to WNTd-derived NK and to CB-derived NK cells (Figure 5G). These functional differences were not due to the acquisition of a “memory-like” phenotype, as all in vitro-derived NK cells expressed low levels of CD57, KLRG1, NKp80, and NKG2C (Figure S4D; (Foley et al., 2012; Lopez-Verges et al., 2010; Lopez-Verges et al., 2011; Romee et al., 2012)).

Figure 5: WNTi-derived NK and WNTd-derived NK cells are functionally distinct.

CD56⁺ NK cells from each ontogenic origin were isolated by FACS and assessed for IFN-[γ] and CD107a by flow cytometry. (A, B) Response of CD56⁺ cells to K562 tumor targets (effector:target (E:T) ratio 1.25:1). (A) Representative flow cytometric analyses of CD107a and IFN-[γ] expression within CD56⁺ NK cells. B) Quantification of CD107a⁺IFN-γ⁻ and CD107a⁻IFN-[γ]⁺ populations, as in A. n≥7. Mean ± SEM. (C, D) Response of CD56⁺ cells to ADCC. (C) Representative flow cytometric analyses of CD107a and IFN-[γ] expression within CD56⁺ NK cells. (D) Quantification of CD107a⁺IFN-[γ]⁻ and CD107a⁻IFN-[γ]⁺ populations, as in C. n=6. (E, F) Response of CD56⁺ cells to PMA/ionomycin stimulation. (E) Representative flow cytometric analyses of CD107a and IFN-[γ] expression within CD56⁺ NK cells. (F) Quantification of CD107a⁺IFN-[γ]⁻ and CD107a⁻IFN-[γ]⁺ populations, as in E. n=4. Mean ± SEM. (G). Quantification of specific killing by 7-AAD uptake of Rituximab-coated Raji cells. n>4. Mean ± SEM. Two-way ANOVA with multiple comparisons. ****p<0.0001. (H, I) Response of CD56⁺CD16⁺ WNTi-derived NK, WNTd-derived NK, CB-DERIVED NK, and PB-NK cells to K562 tumor targets. E:T ratio of 1.25:1. (H) Representative flow cytometric analyses s of CD107a and IFN-[γ] within CD56⁺CD16⁺ NK cells. (I) Quantification of CD107a⁺IFN-[γ]⁻ and CD107a⁻IFN-[γ]⁺ populations, as in H. n = 4. Mean ± SEM. Paired and unpaired student’s t-test. *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001. See also Figure S5.

In adults under homeostatic conditions, the expression of CD16 on NK cells is generally associated with a bias upon stimulation for cytotoxic degranulation as opposed to inflammatory cytokine production (Caligiuri, 2008). Given the high CD16 expression on WNTi-derived NK cells, we asked whether their key effector functional bias was a reflection of CD16 expression, or whether it was an intrinsic bias. We isolated adult donor peripheral blood CD56^dimCD16⁺ cells (PB-NK), and compared their effector functional response to the CD56⁺CD16⁺ population present within the WNTi-derived NK, WNTd-derived NK and CB-derived NK cell populations. All in vitro-derived CD16⁺ NK cell populations exhibited superior TNF and IFN-γ production in comparison to PB-NK cells (Figure S5D). However, the inflammatory response was greatest in CD16⁺ WNTd-derived NK and CB-derived NK cells (Figure 5H,I). In contrast, CD16⁺ WNTi-derived NK cells exhibited a robust degranulation response, greater than all other NK cell populations, including CD56^dimCD16⁺ PB-NK cells (Figure 5H,I). Collectively, these analyses indicate that, regardless of ontogenic origin, in vitro-derived NK cells are functionally distinct from CD56^dimCD16⁺ PB-NK cells. However, hPSC-derived WNTi-NK and WNTd-derived NK cells, originating from ontogenically separate progenitors, are functionally distinct from each other, with each being intrinsically biased for different NK cell effector functions.

Discussion

EMP were originally defined by their definitive erythroid and myeloid potential, and subsequent lineage tracing experiments have shown that EMP contribute to multiple fetal and adult tissue-resident macrophage populations in mice independently of HSCs (Bertrand et al., 2005; Bertrand et al., 2007; McGrath et al., 2015a). We have previously determined that EMP lack B-cell potential, which is present in the E9.5 yolk sac outside of EMP (McGrath et al., 2015a; Yoshimoto et al., 2011). Here, we find that erythro-myeloid-biased progenitors in the yolk sac possess robust NK cell potential. Importantly, our lineage tracing studies indicate that fetal NK cells initially arise from an HSC-independent source that also contributes to erythroid and myeloid cells in the midgestation fetal liver. Taken together, these studies support the concept that yolk sac EMP contain not only erythroid, megakaryocyte, and myeloid, but also innate lymphoid progenitor activity. Future studies will be directed at understanding whether EMP also give rise to other innate lymphoid cell populations in the fetus. EMP may not be the only source of HSC-independent innate lymphoid cells in the mammalian embryo. NK cell lineage potential has been identified in a lympho-myeloid-restricted progenitor present in the fetal liver of E11.5 murine embryos (Boiers et al., 2013). Understanding the potential lineage relationships of HSC-independent EMP, immune-restricted hematopoietic progenitors, and emerging HSCs await future studies.

Our previous studies of WNT-mediated hematopoietic specification of hPSCs originally revealed two ontogenically distinct NK cell progenitor populations (Sturgeon et al., 2014). However, it was not known whether the NK cell potential in both populations was merely in vitro potential or representative of a physiologically-relevant in vivo process. While in vitro studies revealed NK cell potential in the human yolk sac (Tavian et al., 2001), the functional properties of these NK cells were not determined. Here, we find that primary murine fetal NK cells, murine EMP-derived NK cells and human WNTi-derived NK cells are all highly granular and exhibit a potent degranulation response following stimuli, in contrast to adult murine splenic NK cells, and human WNTd- and CB-NK cells, which have a bias for inflammatory cytokine production. These findings indicate that uniquely functional NK cells first emerge during mammalian extra-embryonic development from an evolutionarily conserved, HSC-independent hematopoietic program. Our findings also provide a developmental basis for the previously observed differences in cytotoxic potential between hPSC-derived and cord blood-derived NK cells (Woll et al., 2009).

Our observations raise questions regarding the role of EMP-derived NK cells during embryonic development. While fetal NK cells may play a conserved but unappreciated role in embryonic development, immunodeficient mice lacking NK cell potential do not exhibit gross developmental defects (Kennedy et al., 2000; Kim et al., 2000; Satoskar et al., 1999). However, fetal NK cells may serve as a surveillance system to protect against fetal malignancies or maternally-derived viral exposure, which are not normally present in standard immunodeficient mouse husbandry practices. It will be of interest to investigate the potential in vivo role(s) of fetal NK cells.

One challenge with current NK cell-based immunotherapies is their variable success in clinical trials (Mehta et al., 2018). While the response of malignant cells to therapeutics is undoubtedly complex, this variable response may be due in part to the intrinsic heterogeneity of NK cell populations found across donors (Horowitz et al., 2013). Indeed, even within our in vitro derived populations, we observed key effector function heterogeneity. As this is a reproducible, scalable platform, the hPSC system may be of use in future studies to dissect the mechanism(s) governing cytotoxic versus inflammatory responses in NK cells. The use of hPSC-derived NK cells is an alternative that would lead to the generation of scalable, reproducible NK cells populations that produce homogenous, functionally distinct NK cells. Our studies demonstrated that both murine and human NK cell populations have different functional properties, and that ontogeny must be taken into consideration when interpreting these studies. While these observations should be taken in light of the use of IL-15 in their derivation, the concentration used here is much lower than that used to “prime” adult NK cells into a potent cytotoxic response (Wagner et al., 2017). Further, each NK cell population was derived under identical conditions, strongly suggesting that the observed differences in cytotoxic degranulation response are intrinsic, i.e., due to ontogenic origin. Given our observations, it is reasonable to speculate that hPSC-derived WNTi-NK cells, or more broadly, myeloid progenitor-derived NK cells (Chen et al., 2015; Grzywacz et al., 2011), may represent an ideal population for pre-clinical assessment of adoptive immunotherapeutic potential.

STAR METHODS

CONTACT FOR REAGENT AND RESOURCE SHARING

Further information and requests for resources should be directed to and will be fulfilled by the corresponding authors, Christopher Sturgeon (csturgeon@wustl.edu) and James Palis (James_Palis@URMC.Rochester.edu).

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Animals

Mice were used in accordance with the guidelines of the University of Rochester School of Medicine and Dentistry Institutional Animal Care and Use Committee. C57BL/6 were obtained from Charles River and for lineage tracing experiments, Csf1r^MerCreMer ;FVB-Tg(Csf1r-cre/Esr1*)1Jwp/J) and Rosa26^YFP (B6.129X1-Gt(ROSA)26Sortm1(EYFP)Cos/J) reporter mice were purchased from The Jackson Laboratory. Both female and male embryos were analyzed in these experiments. All mice were maintained on 12-hour dark and light cycles.

Cell culture

The hESC line, H1, (WA01; WiCell; male) was maintained on irradiated mouse embryonic fibroblasts in hESC media, as described previously (Kennedy et al., 2012; Sturgeon et al., 2014). K562 cell line (ATCC, Manassas, VA), leukemia cell line established from a patient with chronic myelogenous leukemia. K562 cells were maintained in Iscove’s Modified Dulbecco’s media, supplemented with 10% fetal calf serum (FCS), L-glutamine (2 mM), and Penicillin/Streptomycin (1%) in a 5% CO₂ incubator. Raji cell line (ATCC, Manassas, VA), lymphoma cell line from a patient with Burkitt’s lymphoma. Raji cells were maintained in Roswell Park Memorial Institute (RPMI) media, supplemented with 10% FCS, L-glutamine (2 mM), 1M HEPES (1%), Non-essential amino acids (1%), Penicillin/Streptomycin (1%), Sodium Pyruvate (1 mM) in a 5% CO₂ incubator. OP9-DL4 cells were a gift from J.C. Zuniga-Pflucker and were maintained in α-MEM supplemented with 20% FBS, L-glutamine (2 mM) and Penicillin/Streptomycin (1%) in a 5% CO₂ incubator. Anonymous human platelet apheresis donor PBMCs were obtained by Ficoll centrifugation. NK cells were purified using RosetteSep (StemCell Technologies; ≥95% CD56⁺CD3⁻) or by flow cytometric cell sorting (BD FACSAria II).

METHOD DETAILS

Mice, Tissue Collection and Processing

Mice were mated overnight and the morning of vaginal plug detection was considered embryonic day (E) 0.3. Mice were sacrificed by CO₂ narcosis. Embryonic tissues were isolated from staged embryos and dissociated with Type I Collagenase as previously described (Stem Cell Technologies, diluted 1:3 with PB2, (McGrath et al., 2011). Adult minced spleens or flushed femoral marrow were dissociated by trituration. For lineage tracing, 75 mg per gm body weight of 4-hydroxytamoxifen (Sigma) and 37.5 mg per gm of body weight progesterone (Sigma) were delivered as a single i.p. injection to pregnant Csf1r^MerCreMer x Rosa26^YFP females, as reported (Gomez Perdiguero et al., 2015). Hematopoietic cells in the fetal liver of resulting embryos were analyzed by flow cytometry at days 14.5 or 15.5 of gestation.

NK cell derivation from murine yolk sacs

For murine NK cell cultures sorted fetal EMP (kit⁺CD16/32⁺CD41⁺) or adult CLP (Lineage⁻kit⁺CD127⁺) were cultured at 500 to 2000 sorted cells/well in a 24-well plate containing 80% confluent OP9 cells and cultured for 10–21 days in opti-MEM+GlutaMax supplemented with 10% FBS (HyClone), 1% Penicillin/Streptomycin, βME (1.8 ug/mL), mFlt3-L (25 ng/mL), mSCF (25 ng/mL), IL-2 (50 ng/mL), IL-7 (20 ng/mL), and IL-15 (25 ng/mL) at 37°C, 5% CO 2. For limiting dilutions 300, 100, 30, 10 or one EMP were sorted directly into a 96 well plate containing 100 μL co-culture media and OP9 cells as above.

Murine NK cell functional assays

For in vivo NK cells, adult splenic and embryonic day 15.5 fetal liver cells were depleted for Ter119, CD19, and CD3ε using Miltenyi cell isolation (LS) columns and QuadroMACS separation unit (Miltenyi). Cells were plated in a 24-well plate at 1.5 × 10⁶ cells/well in 1mL Opti-MEM+Glutamax supplemented with 10% FBS (Hyclone). In vitro NK cells were obtained from EMP cells or adult CLP cells co-cultures on OP9 as described above for progenitor assays, except plated on 3-cm dishes with 4mL co-culture media. At day 19–21 of co-culture, all cells were filtered through 20 μm filter, washed with PBS. In either case, anti-CD107a (BD Biosciences) was added to each well, 1:100 dilution, prior to stimulation. Cells were stimulated with 4 μg/mL Ionomycin + 500 ng/mL PMA and incubated in a 37°C, 5% CO 2 incubator for 30 minutes. Cells were placed on ice for 10 minutes then using ice cold buffers, analyzed by flow cytometry for surface NK cell markers and CD107a expression (Supplemental Table 1).

Differentiation of hPSCs

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 (6–10 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), bFGF (5 ng/mL) and either CHIR99021 (3 μM) to specify WNT-dependent hematopoietic progenitors or IWP2 (3 μM) to specify WNT-independent hematopoietic progenitors. Following 72 hrs of differentiation, embryoid bodies were placed in StemPro-34 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), IL-6 10 ng/mL), IGF-1 (25 ng/mL), EPO (2U/mL final), IL-11 (5 ng/mL), bFGF (5 ng/mL) and SCF (50 ng/mL). Cultures were maintained in a low oxygen, 5% CO₂/5% O₂/90% N₂ incubator. All human recombinant factors were purchased from R&D Systems, except EPO (Peprotech) (Creamer et al., 2017; Dege and Sturgeon, 2017; Ditadi and Sturgeon, 2016; Ditadi et al., 2015; Kennedy et al., 2012; Sturgeon et al., 2014).

T cell derivation from hPSCs

OP9 cells expressing Delta-like 4 (OP9-DL4) were generated and described previously (La Motte-Mohs et al., 2005). Isolated WNT-dependent progenitors (CD34⁺CD43⁻CD73⁻CD184⁻) and WNT-independent progenitors (CD34⁺CD43⁻CD73⁻CD184⁻, CD34⁺CD43⁺, or CD34⁻CD43⁺), and umbilical cord blood progenitors (CD34⁺) were added to individual wells of a 24-well plate containing confluent OP9-DL4 cells, and cultured for 21–28 days, as described previously (Kennedy et al., 2012). Briefly, co-cultures were cultured in α-MEM supplemented with 20% FBS (HyClone), SCF (30 ng/mL, first 5 days only), IL-7 (5 ng/mL), FLT3-L (5 ng/mL). Every 4–5 days, co-cultures were triturated and passed through a 40 μm cell strainer and passaged onto fresh OP9-DL4 cells in a 6-well plate.

NK cell derivation from hPSCs

Isolated WNT-dependent progenitors (CD34⁺CD43⁻), WNT-independent progenitors (CD34⁺43⁻, CD34⁺CD43⁺, or CD34⁻CD43⁺), and umbilical cord blood progenitors (CD34⁺) were added to individual wells of a 24-well plate containing confluent OP9-DL4 cells, as described above, and cultured for 14–35 days. Co-cultures were cultured in α-MEM supplemented with 20% FBS (HyClone), IL-3 (5 ng/mL, first 5 days only), 30 ng/mL SCF, IL-7 (5 ng/mL), FLT3-L (5 ng/mL), and IL-15 (10 ng/mL). Every 5 days, half media changes were performed in the 24-well plate.

Human NK cell functional assays

CD56⁺ or CD56⁺CD16⁺ NK cells were isolated by FACS (BD FACS AriaII) and cultured in R10 media supplemented with 1 ng/mL IL-15 and 6 μM SB431542, for 48 hours. For cytokine production and degranulation functional assays, CD56⁺ NK cells were co-incubated with the human K562 erythroleukemia tumor cells (E:T ratio of 1.25:1), co-incubated with plate-bound anti-FcγRIIIa (α-CD16; 10 μg/mL; Biolegend) or isotype control (10 μg/mL; Biolegend), or with phorbol 12-myristate-13-acetate (PMA; 50 ng/mL) and ionomycin (1 μg/mL) for 6 hours, with GolgiPlug and GolgiStop (BD Biosciences) for the final 5 hours. Cells were washed with PBS then stained with Live/Dead Fixable Aqua (Invitrogen). Cells were then subsequently washed and surface stained for NK cell markers, fixed/permeabilized, and intracellularly stained for IFN-γ and TNF (Cytofix/Cytoperm; BD Biosciences). For degranulation assays, anti-CD107a mAb was added at the start of the functional assay. Cells were analyzed via flow cytometry and data was acquired on a Fortessa X20 flow cytometer (BD Biosciences). For flow based killing assay, CD56⁺ NK cells were incubated for 4 hours with carboxyfluorescein succinimidyl ester (CFSE)-labeled human Raji Burkitt’s lymphoma targets or rituximab-coated Raji lymphoma targets at various effector to target (E:T) ratios. After 4 hour incubation, cells were stained with 7-aminoactinomycin (7-AAD) and analyzed immediately using a Fortessa X20 flow cytometer (BD Biosciences). Specific killing was calculated by subtracting spontaneous Raji death (no effector control), and was less than 5% in all cases.

Methylcellulose assays

Hematopoietic progenitors were isolated by FACS and plated directly into MethoCult H4034 (StemCell Technologies), and were scored after 10 days of culture, as per the manufacturer’s instructions.

RNA isolation and qRT-PCR analysis

For human samples, qRT-PCR was run as described previously (Sturgeon et al., 2014). Briefly, total RNA was isolated for human samples with the RNAqueous RNA Isolation Kit (Invitrogen), followed immediately by transcription into cDNA utilizing random hexamers and Oligo (dT) with Superscript III Reverse Transcriptase (Invitrogen). For murine samples, RNA was isolated using the Qiagen RNeasy Kit (Qiagen) and transcribed into cDNA using the iScript cDNA synthesis kit (BioRad). Real-time quantitative PCR was performed on a StepOnePlus thermocycler (Applied Biosystems), using Power Green SYBR mix (Invitrogen). Murine samples and human HOXA3 qRT-PCR was performed with TaqMan Probes using TaqMan Gene Expression master mix II (Applied Biosystems). Primer sequences are found in Key Resources Table. Gene expression was evaluated as DeltaCt relative to control (ACTB for human, 18S for murine samples).

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Anti-Mouse PE-NK1.1	eBioscience	Cat# 12–5941-82
Anti-Mouse PE-NKp46	eBioscience	Cat#12–3351-80
Anti-Mouse PECy7-CD117	eBioscience	Cat# 25–1171-82
Anti-Mouse FITC-CD127	eBioscience	Cat# 11–1271-82
Anti-Mouse APC-CD127	eBioscience	Cat#17–1271-82
Anti-Mouse PE-CD127	eBioscience	Cat# 12–1271-82
Anti-Mouse PerCPCy5.5-CD19	eBioscience	Cat# 45–0193-82
Anti-Mouse APC-CD3ε	eBioscience	Cat# 17–0031-82
Anti-Mouse APC-eFluor780 Ter119	eBioscience	Cat# 47–5921-82
Anti-Mouse PE-Ter119	BD Biosciences	Cat#553673
Anti-Mouse PECF594-CD122	BD Biosciences	Cat#564763
Anti-Mouse FITC-CD122	eBioscience	Cat# 11–1222-82
Anti-Mouse PECy7-CD122	eBiosciences	Cat#25–1222-82
Anti-Mouse PE-CD127	eBioscience	Cat# 12–1271-82
Anti-Mouse PECF594-CD49b	BD Biosciences	Cat# 562453
Anti-Mouse PEcy7-CD49b	eBioscience	Cat#25–5971-82
Anti-Mouse PE- CD48	eBioscience	Cat#12–0481-81
Anti-Mouse APC-CD107a	BD Biosciences	Cat# 560756
Anti-Mouse APCeFluor780-GR-1	eBioscience	Cat# 47–5931-82
Anti-Mouse APCeFluor780-CD11b	eBioscience	Cat#47–0112-82
Anti-Mouse PE-CD16/32	eBioscience	Cat# 12–0161-82
Anti-Mouse FITC-CD41	BD Biosciences	Cat# 553848
Anti-Mouse APC CD150	BioLegend	Cat#115910
Anti-mouse PerCPCy5.5-Sca1	eBioscience	Cat# 45–5981-82
Anti-mouse APCeFluor780-CD3ε	eBioscience	Cat# 47–0032-82
Anti-mouse APCeFluor780-CD19	eBioscience	Cat# 47–0193-82
Anti-mouse APC-Fire750 Ly6G	BioLegend	Cat#127652
Anti-mouse PEcy7 Ly6C	eBioscience	4Cat#5–5932-82
Anti-Mouse F4/80 APC	eBioscience	Cat# 17–4801-82
Anti-Mouse PE GranzymeB	BioLegend	Cat#372207
Anti-Human FITC-CD3ε	BD Biosciences	Cat# 555916
Anti-Human FITC-CD43	BD Biosciences	Cat# 555475
Anti-Human FITC-CD94	BD Biosciences	Cat# 555888
Anti-Human FITC-Perforin	BD Biosciences	Cat# 556577
Anti-Mouse FITC IgG2b, κ isotype	BD Biosciences	Cat# 556577
Anti-Human APC-CD15	BD Biosciences	Cat# 551376
Anti-Human APC-CD34	BD Biosciences	Cat# 340441
Anti-Human APC-CD56	BD Biosciences	Cat# 555518
Anti-Human APC-CD57	BD Biosciences	Cat# 560845
Anti-Human APC-CD107a	BD Biosciences	Cat# 560664
Anti-Human APC-CD184	BD Biosciences	Cat# 555976
Anti-Human APC-Granzyme B	Thermo Fisher	Cat# GRB05
Anti-Mouse APC IgG1, κ isotype	BD Biosciences	Cat# 555751
Anti-Human APC-NKG2A	Beckman Coulter	Cat# A60797
Anti-Human PE-CD8	BD Biosciences	Cat# 561950
Anti-Human PE-CD43	BD Biosciences	Cat# 560199
Anti-Human PE-CD73	BD Biosciences	Cat# 550257
Anti-Human PE-IFN-γ	BD Biosciences	Cat# 554701
Anti-Human PE-CD117	BD Biosciences	Cat# 340529
Anti-Human PE-NKp30	BD Biosciences	Cat# 558407
Anti-Human PE-NKp44	BD Biosciences	Cat# 558563
Anti-Human PE-NKp46	BD Biosciences	Cat# 557991
Anti-Human PE-NKp80	R&D Systems	Cat# FAB1900P-025
Anti-Human PE-CD226 (DNAM1)	BD Biosciences	Cat# 559789
Anti-Human PE-NKG2C	R&D Systems	Cat# FAB138P-100
Anti-Human PE-NKG2D	BD Biosciences	Cat# 561815
Anti-Human PE-CD158a,h	Beckman Coulter	Cat# A09778
Anti-Human PE-CD158b	BD Biosciences	Cat# 559785
Anti-Human PE-CD158e1	BD Biosciences	Cat# 555967
Anti-Human PE-CD158i	Beckman Coulter	Cat# IM3337
Anti-Human PE-Granzyme A	BD Biosciences	Cat# 558904
Anti-Mouse PE IgG1, κ isotype	BD Biosciences	Cat# 558904
Anti-Human PE Cy7-CD56	BD Biosciences	Cat# 557747
Anti-Human PE Cy7-CD34	BD Biosciences	Cat# 348791
Anti-Human PE Cy7-KLRG1	Biolegend	Cat# 368614
Anti-Human Krome Orange CD45	Beckman Coulter	Cat# A96416
Anti-Human APC Cy7 CD45	BD Biosciences	Cat# 348795
Anti-Human V450 CD45	BD Biosciences	Cat# 642275
Anti-Human PerCPCy5.5 CD4	BD Biosciences	Cat# 560650
Anti-Human PerCPCy5.5 CD16	BD Biosciences	Cat# 560717
Anti-Human PerCPCy5.5 Granzyme K	Biolegend	Cat# 370513
Anti-Mouse PerCPCy5.5 IgG1, κ isotype	BD Biosciences	Cat# 552834
Anti-Human eF450 TNF	eBioscience	Cat# 50–164-06
Anti-Human LEAF purified CD16	Biolegend	Cat# 302014
Anti-Human LEAF purified isotype IgG1,κ	Biolegend	Cat# 400124
Anti-Human Rituxan	Genentech	N/A
Biological Samples
Human umbilical cord blood	St. Louis Cord Bank SSM Health Cardinal Glennon Children’s Hospital	N/A
Human peripheral blood mononuclear cells	Mississippi Valley Regional Blood Center, Davenport, IA	N/A
Chemicals, Peptides, and Recombinant Proteins
α-MEM	Life Technologies	Cat# 12000–022
Fetal Bovine Serum (FBS)	HyClone	Cat# SH30396.03
Iscove's Modified Dulbecco's Medium (IMD)	Life Technologies	Cat# 12200069
Ham’s F12	Corning	Cat# 10–080
N2 Supplement	Life Technologies	Cat# 1752048
B27 Supplement, no vitamin A	Life Technologies	Cat# 12587010
Stempro-34	Life Technologies	Cat# 10639011
Human Serum transferrin	Sigma-Aldrich	Cat# 10652202001
L-absorbic acid	Sigma-Aldrich	Cat# A4403
Monothioglycerol	Sigma-Aldrich	Cat#M6145
L-glutamine, 200 mM solution	Life Technologies	Cat# 25030–081
Penicillin-streptomycin	Life Technologies	Cat# 15070–063
Knockout Serum replacement	Life Technologies	Cat# 10828028
Non-essential amino acids	Life Technologies	Cat# 11140050
β-mercaptoethanol, 55 mM solution	Life Technologies	Cat# 21985023
Opti-MEM™ Reduced Serum Medium, GlutaMAX™ Supplement	ThermoFisher	Cat# 51985034
GlutaMAX™	ThermoFisher	Cat# 35–050-061
Recombinant Murine FLT3L	Peprotech	Cat#250–31L
Recombinant Murine SCF	Peprotech	Cat# 250–03
Recombinant Murine IL-2	Peprotech	Cat# 212–12
Recombinant Murine IL-17	Peprotech	Cat#210–17F
Recombinant Murine IL-15	Peprotech	Cat#210–15
Recombinant human bFGF	R&D Systems	Cat# 233-FB
Recombinant human BMP4	R&D Systems	Cat# 314-BP
Recombinant human Activin A	R&D Systems	Cat# 338-AC
Recombinant human VEGF	R&D Systems	Cat# 293-VE
Recombinant human SCF	R&D Systems	Cat# 255-SC
Recombinant human IGF-1	R&D Systems	Cat# 291-G1
Recombinant human IL-3	R&D Systems	Cat# 203-IL
Recombinant human IL-6	R&D Systems	Cat# 206-IL
Recombinant human IL-7	R&D Systems	Cat# 207-IL
Recombinant human IL-11	R&D Systems	Cat# 218-IL
Recombinant human IL-15	R&D Systems	Cat# 247-IL
Recombinant human EPO	PeproTech	Cat# 100–64
Recombinant human FLT3L	R&D Systems	Cat# 308-FK
CHIR99021	Tocris	Cat# 4423
IWP2	Tocris	Cat# 3533
SB431542	Tocris	Cat# 1614
MethoCult H4034	StemCell Technologies	Cat# 4034
GolgiPlug	BD Biosciences	Cat# 555029
CFSE	BD Biosciences	Cat# 565082
DAPI	Thermo Scientific	Cat# EN62248
Phorbol 12-myristate 13-acetate (PMA)	SigmaAldrich	Cat# P8319
Ionomycin	SigmaAldrich	Cat# I0634
SuperScript III Reverse Transcriptase	Life Technologies	Cat# 18080044
Growth factor reduced Matrigel	Corning	Cat# 354230
Live/Dead Fixable Aqua	Invitrogen	Cat# L34957
7-AAD	BD Biosciences	Cat# 559925
Critical Commercial Assays
Fixation/Permeablization Solution Kit with GolgiStop	BD Biosciences	Cat# 554715
FoxP3/Transcription Factor Intracellular Staining Kit	eBiosciences/Thermo Fisher	Cat# 00–5523-00
RosetteSep Human NK cell enrichment kit	StemCell Technologies	Cat# 15025
RNAqueous RNA Isolation Kit	Life Technologies	Cat# AM1912
PowerSybr Green Master Mix	Life Technologies	Cat# 4367659
TaqMan Gene Expression Master Mix	Life Technologies	Cat# 4369016
QIAzol Lysis Reagent	Qiagen	Cat# 79306
QIAshredder	Qiagen	Cat# 79654
RNeasy Micro Kit	Qiagen	Cat# 74004
LS columns	Miltenyi Biotech	Cat# 130–042-401
Deposited Data
RNA-sequencing data	This paper	SRA Accession# PRJNA525404
Experimental Models: Cell Lines
Human: H1 hESC cells	WiCell	Cat# WA01
Human: K-562 cells	ATCC Luis Batista Lab	Cat# CCL-243
Human: Raji Cells	ATCC Todd Fehniger Lab	Cat# CCL-86
Mouse: OP9-DL4 cells	Juan-Carlos Zuniga-Pflucker
Mouse: OP9 cells	Juan-Carlos Zuniga-Pflucker
Experimental Models: Organisms/Strains
C57BL/6 wildtype Inbred mice	Charles River	C57BL/6NCrl
CSF1R-ERcre	Jackson Labs	CSF1Rmer-cre-mer ;FVB-Tg(Csf1r-cre/Esr1*)1Jwp/J
Rosa26YFP	Jackson Labs	B6.129X1-Gt(ROSA)26Sortm1(EYFP)Cos/J
Oligonucleotides
ACTB F	AAACTGGAACGGTGAAGGTGACAG	N/A
ACTB R	CAATGTGCAATCAAAGTCCTCGGC	N/A
HOXA1 F	ATCTTCTCCAGCGCAGACTT	N/A
HOXA1 R	CAGGTAGCCGTACTCTCCAA	N/A
HOXA5 F	TGGAACTCCTTCTCCAGCTC	N/A
HOXA5 R	AGATCTACCCCTGGATGCG	N/A
HOXA7 F	AGGACTGTGGAGATGCTTCC	N/A
HOXA7 R	AGGAAACATCAGGGCGTACA	N/A
HOXA9 F	CGAGAGGCAGGTCAAGATCT	N/A
HOXA9 R	TGGCATCACTCGTCTTTTGC	N/A
HOXA10 F	AGCACACCACAATTCTCCCT	N/A
HOXA10 R	AGCACACCACAATTCTCCCT	N/A
ACTB FAM TaqMan Probe HS99999903_m1	Life Technologies	Cat# 4331182
HOXA3 FAM TaqMan Probe HS00601076_m1	Life Technologies	Cat# 4331182
Mouse Eomes TaqMan Probe	Thermo Fisher Scientific	Mm01351984_m1
Mouse Id2 TaqMan Probe	Thermo Fisher Scientific	Mm00711781_m1
Mouse Prf1 TaqMan Probe	Thermo Fisher Scientific	Mm00812512_m1
Mouse Hoxa3 TaqMan Probe	Thermo Fisher Scientific	Mm01326402_m1
Mouse Hoxa5 TaqMan Probe	Thermo Fisher Scientific	Mm00439362_m1
Mouse Hoxa7 TaqMan Probe	Thermo Fisher Scientific	Mm00657963_m1
Mouse Hoxa9 TaqMan Probe	Thermo Fisher Scientific	Mm00439364_m1
Human/Mouse 18S TaqMan Probe	Thermo Fisher Scientific	Hs99999901_s1
Software and Algorithms
GraphPad Prism v7.0	GraphPad Software, Inc	https://www.graphpad.com/scientific-software/prism/
FlowJo v10.3	FlowJo, LLC	https://www.flowjo.com/solutions/flowjo
ImageJ		https://imagej.nih.gov/ij/
STAR		http://code.google.com/p/rna-star/
SubRead	Walter + Eliza Hall Institute of Medical Research	http://subread.sourceforge.net
DESeq2		http://www.bioconductor.org/packages/release/bioc/html/DESeq2.html
Morpheus	Broad Institute	https://software.broadinstitute.org/morpheus/
Gene Set Enrichment Analysis	Broad Institute	http://software.broadinstitute.org/gsea/index.jsp
Other

Transmission Electron Microscopy

CD56⁺ NK cells or CD15⁺ granulocytes were isolated by FACS (BD FACS AriaII) and cultured in R10 media supplemented with 1 ng/mL IL-15 for 48 hours. Cells were then gently pelleted in a 1.7mL Eppendorf tube and media was aspirated and replaced with 2.5% glutaraldehyde with 2% paraformaldehyde in 0.15M cacodylate buffer with 2mM CaCl₂ that was warmed to 37˚C and allowed to fix for 1 hour. After fixation, the cells were gently pelleted and washed in 0.15M cacodylate buffer with 2mM CaCl₂. This was repeated two additional times after which the cells were resuspended and pelleted in 12% gelatin in PBS and placed on ice for 20 minutes to gel. After gelling, the tip of the tube was cut off and the cells in gelatin were removed and diced into 2mm cubes. These were then rinsed 3 times in 0.15M cacodylate buffer with 2mM CaCl₂ and post-fixed in 1% osmium tetroxide with 1.5% potassium ferrocyanide in the same buffer for 1 hour in the dark. Cells were then rinsed in ultrapure water 3 times for 10 minutes each, and stained in an aqueous solution of 2% uranyl acetate for one hour. The samples were then again washed in ultrapure water 3 times for 10 minutes each and dehydrated in a graded ethanol series (50%, 70%, 90%, 100% x2) for 10 minutes in each step. The cells were then infiltrated with microwave assistance (Pelco BioWave Pro, Redding, CA) into LX112 resin, placed into BEEM capsules and cured in an oven at 60˚C for 48 hours. Blocks were then sectioned and counterstained with 2% uranyl acetate for 10 minutes and Reynolds lead citrate for 2 minutes. Grids were imaged on a TEM (JEOL JEM-1400, Tokyo, Japan) at 120 KeV.

RNA-seq

CD34⁺ hematopoietic progenitors were isolated by FACS (BD FACS AriaII). CD56⁺ NK cells were isolated by FACS (BD FACS AriaII) and cultured in R10 media supplemented with 1 ng/mL IL-15 for 48 hours before RNA isolation. Total RNA was prepared for whole-transcriptome sequencing using the Clontech SMARTer kit and was sequenced with an Illumina HiSeq 3000 with 1×50 single reads. Bulk reads were processed using several software packages; alignment was completed using STAR, reads assigned with SUBREAD, and differential expression analysis utilized DESeq2. Heat maps were generated using Cluster 3.0 and Morpheus. After the RNAseq reads were aligned and quantified in each sample, ~10,000–13,000 genes were found to be expressed (mean RPKM > 1) in the WNTi-derived NK, WNTd-derived NK, CB-DERIVED NK, and WNTi- and WNTd-CD34⁺ populations. Reads were uploaded to the Sequence Read Archive (SRA) and are available under accession number PRJNA525404. Gene set enrichment analysis (GSEA) was performed comparing WNTi-derived and WNTd-derived NK cells. The following parameters were chosen for the analysis: c2 (curated) and H (hallmark) gene sets from the Molecular Signatures Database (MSigDB), 1000 permutations of gene set labels, signal-to-noise ratio metric, and weighted scoring. Gene sets with a false discovery rate (FDR) <0.25 were considered for the analysis.

Flow cytometry and cell sorting

All antibody combinations used in this study are listed in Key Resources table and Supplemental Table 1. Cells were sorted with a FACSAria™II (BD) cell sorter, and analyzed on a LSRii or LSR Fortessa X20 (BD) cytometer. Data were analyzed using FlowJo v10.3 (TreeStar).

QUANTITATION AND STATISTICAL ANALYSIS

Quantitation and statistics

The unpaired student’s t test was used to test for the significance. Mean values ± SEM are shown. Differences between groups were compared with unpaired student’s t-test using Prism 7.0 (GraphPad Software, San Diego, CA). A two tailed paired student’s t-test and a Two Way ANOVA with multiple comparisons were used where applicable. No statistical method was used to predetermine sample size. All experiments are representative of at least three independent experiments, unless otherwise noted.

DATA AND SOFTWARE AVAILABILITY

The accession number for the RNA sequencing data is available under NCBI SRA: PRJNA525404.

Supplementary Material

3

Supplemental Table 1: Flow cytometry stains by figure panel, Related to STAR Methods.

Highlights.

• NK cell potential arises from erythro-myeloid progenitors (EMP) in the yolk sac

• EMP-derived NK cells, like fetal NK cells, have a potent degranulation response

• hPSC differentiation yields 2 distinct CD34⁺ populations, each with NK cell potential

• hPSC-derived EMP-like NK cells are more potently cytotoxic than adult CD16⁺ NK cells