CD56 expression marks human group 2 innate lymphoid cell divergence from a shared NK cell and group 3 innate lymphoid cell developmental pathway

SUMMARY:

According to the established model of murine innate lymphoid cell (ILC) development, helper ILCs develop separately from natural killer (NK) cells. However, it is unclear how helper ILCs and NK cells develop in humans. Here we elucidated key steps of NK cell, ILC2, and ILC3 development within human tonsils using ex vivo molecular and functional profiling and lineage differentiation assays. We demonstrated that while tonsillar NK cells, ILC2s, and ILC3s originated from a common CD34⁻CD117⁺ ILC precursor pool, final steps of ILC2 development deviated independently and became mutually exclusive from those of NK cells and ILC3s, whose developmental pathways overlapped. Moreover, we identified a CD34⁻CD117⁺ ILC precursor population that expressed CD56 and gave rise to NK cells and ILC3s but not to ILC2s. These data support a model of human ILC development distinct from the mouse, whereby human NK cells and ILC3s share a common developmental pathway separate from ILC2s.

eTOC BLURB:

Human innate lymphoid cells (ILCs) develop from tissue-resident ILC precursors (ILCPs) in secondary lymphoid tissues. Here, Chen and colleagues describe a model of human ILC development in tonsils in which NK cells and group 3 ILCs derive from a CD56⁺ subset of ILCPs that cannot differentiate into group 2 ILCs.

INTRODUCTION

Innate lymphoid cells (ILCs) are “lineage antigen” (Lin) negative cells that are enriched in lymphoid tissues and play multiple roles in health and disease (Artis and Spits, 2015). ILCs include cytotoxic natural killer (NK) cells and three types of non-cytotoxic “helper” ILCs designated as ILC1s, ILC2s, and ILC3s. All ILCs lack rearranged antigen-specific receptors, yet they share many features with adaptive T cell subsets. NK cells express the transcription factors (TFs) T-BET and EOMES, produce interferon-γ (IFN-γ), and kill tumor and virally-infected cells (Sun and Lanier, 2011). ILC1s are T-BET⁺EOMES⁻ and produce IFN-γ, and while rare in the healthy gut, ILC1s are expanded in Crohn’s disease (Bernink et al., 2013; Fuchs et al., 2013; Klose et al., 2014). ILC2s express the TFs GATA3 and RORa and produce type 2 cytokines in response to helminth infections (Mjosberg et al., 2011; Moro et al., 2010; Neill et al., 2010). ILC3s express the TFs AHR and RORγt and secrete IL-17 and/or IL-22 to regulate mucosal inflammation (Cella et al., 2009; Cupedo et al., 2009; Hughes et al., 2009).

Human ILCs can be identified by flow cytometry among Lin⁻ lymphocytes using subset associated markers. NK cells express CD16, CD56, CD94, NKG2A, NKG2C, NKp80, and/or killer immunoglobulin-like receptors (KIRs) (Scoville et al., 2017); ILC2s express CD294 and/or KLRG1 (Huang et al., 2015; Mjosberg et al., 2011); ILC3s express CD117 and NKp44 (Cella et al., 2009; Cupedo et al., 2009; Hughes et al., 2009); and ILC1s can be identified as Lin⁻ lymphocytes that lack NK cell, ILC2, and ILC3 markers yet express the pan-ILC marker, CD127, the alpha chain of the interleukin 7 (IL-7) receptor (Bernink et al., 2013). Of note, some of the above markers are shared by ILC subsets. For instance, CD56 and NKp44 can be expressed by both ILC3s and NK cells (Cella et al., 2009; Cupedo et al., 2009).

In mice ILCs originate from bone marrow-derived common lymphoid progenitors (CLPs) (Serafini et al., 2015). CLPs in turn yield common ILC progenitors (CILPs) that give rise to all ILCs but lack non-ILC differentiation potential (Seehus et al., 2015; Yang et al., 2015; Yu et al., 2014). CILPs then give rise to common helper ILC progenitors (CHILPs) that can generate helper ILCs but not NK cells (Constantinides et al., 2014; Klose et al., 2014), as well as to lineage-committed NK cell progenitors (NKPs) that can only differentiate into NK cells (Carotta et al., 2011; Fathman et al., 2011; Rosmaraki et al., 2001). The existence of CHILPs and lineage-committed NKPs supports a divergent model for terminal murine helper ILC and NK cell development, respectively, and this model is reinforced by genetic fate-mapping studies tracing the mRNA expression of key TFs including Rorγt and Plzf (Constantinides et al., 2014; Klose et al., 2014; Satoh-Takayama et al., 2010; Vonarbourg et al., 2010).

In humans, multipotent CD34⁺CD117⁻ CLP-like early tonsil progenitor cells (ETPs) as well as CD34⁺CD117⁺ CILPs have been described in secondary lymphoid tissues (SLTs) such as lymph nodes and tonsils (Freud et al., 2006; Scoville et al., 2016). In addition, systemic pan-ILC precursors (ILCPs) that are functionally similar to CILPs but that are CD34⁻CD117⁺ were recently characterized (Lim et al., 2017), suggesting some similarities between murine and human ILC development. However, a human progenitor population analogous to the murine CHILP, which can differentiate into helper ILCs but not NK cells, has not been described. Furthermore, the distinct pathway by which each human ILC subset develops in vivo remains unclear (Scoville et al., 2017).

Here we provide evidence that human NK cell, ILC2, and ILC3 development occurs sequentially within tonsils from CD34⁺CD117⁻ ETPs through CD34⁺CD117⁺ CILP and CD34⁻CD117⁺ ILCP intermediates that in turn give rise to terminally differentiated ILCs. Whereas we did not identify a human CHILP analogous to that described in mice, we did identify a CD56⁺ subset of CD34⁻CD117⁺ ILCPs that exclusively gave rise to NK cells and ILC3s but not to ILC2s. Our findings support a model in which human NK cells and ILC3s can share a common developmental pathway separate from that of ILC2s.

RESULTS

Putative NK cell, ILC2, and ILC3 developmental pathways exist in human tonsils

Based on previous discoveries of lymphoid progenitor and precursor populations in human SLTs (Freud et al., 2005; Lim et al., 2017; Montaldo et al., 2014; Renoux et al., 2015; Scoville et al., 2016), we hypothesized that human ILCs share a common developmental pathway in SLTs and that at some point each ILC subset terminally differentiates along its own developmental trajectory. To test this hypothesis, we first used mass cytometry (Behbehani et al., 2015) together with visual Stochastic Network Embedding (viSNE) analyses (Amir el et al., 2013). Excluding Lin⁺ cells via biaxial gating, 25 ILC associated markers were configured into two dimensions, t-SNE1 and t-SNE2 [Table S1, antibodies labeled with “(v)”]. CD34, CD117, CD94, KLRG1, and NKp44 were particularly useful in identifying distinct clusters of CD34⁺CD117⁻ ETPs and CD34⁺CD117⁺ CILPs, CD34⁻CD117⁺ ILCPs, CD94⁺ NK cells, KLRG1⁺ ILC2s, and NKp44⁺CD94⁻ ILC3s in the two dimensional viSNE plots (Figure 1). In contrast, we did not detect distinct clusters of ILC1s (data not shown), likely due to the rarity of ILC1s in healthy donor human tonsils (Bernink et al., 2013; Simoni et al., 2017; Vely et al., 2016). The above cell cluster designations were further supported by viSNE plots showing expression data of other ILC subset associated markers (Figures 1A and S1), and suggested hierarchical relationships stemming from the CD34⁺ ETP and CILP cluster (Figure 1B, left box). Additionally, the expression patterns of NKp44, CD161, and CD56 suggested that some developing NK cells can differentiate from an ILC3 phenotype (Figure 1B, red dashed arrows) as has been recently proposed (Freud et al., 2016; Hughes et al., 2014). These data suggested that NK cells, ILC2s, and ILC3s can develop in tonsils from CD34+ ETPs and CILPs, through an intermediate CD34⁻ CD117⁺ ILCP population, and finally along distinct terminal differentiation pathways (Figure 1B, dashed arrows).

Figure 1.

Putative ILC developmental pathways exist in human tonsils. (A) Representative viSNE cluster plots showing cell localizations and expression patterns of ILC associated markers among freshly enriched human tonsil lymphoid cells. The relative expression of each antigen in the plots spans from low (purple) to intermediate (blue) to high (green). Following the exclusion of Lin⁺ cells (Lin = CD3, CD4, CD5, CD14, CD19, and CD20) and inclusion of live CD45⁺ cells via biaxial gating, 25 separate markers were configured into 2 dimensions, t-SNE1 and t-SNE2. (B) Cluster designations (left plot) based on the data in (A) suggested putative pathways of NK cell, ILC2, and ILC3 development stemming from CD34⁺ progenitor cells in the human tonsils. Curved red, blue, and green dashed arrows represent NK cell, ILC2, and ILC3 developmental pathways, respectively. ETP = Early Tonsil Progenitor. CILP = Common ILC Progenitor. ILCP = ILC Precursor. The data shown are representative of n = 3 fresh human tonsils from three independent experiments. See also Table S1 and Figure S1.

Developing tonsillar ILCs emerge directly from a pool of CD34⁻CD117⁺ ILCPs

To further test the hypothesis that each developing ILC emerges directly from a CD34⁻CD117⁺ ILCP population, we performed flow cytometric analyses of freshly enriched human tonsil-derived cells. NK cells (“NK” in Figure 2A, red events) were identified as Lin⁻ cells (i.e. lacking CD3, CD4, CD5, CD14, CD19, CD20, CD123, and FcεR1α) that expressed one or more of the following: CD16, CD94, NKG2A, NKG2C, NKp80, KIR2D, and/or KIR3DL1-2; ILC2s (“ILC2” in Figure 2B, blue events) were identified as Lin⁻NK⁻ cells that expressed CD294 and/or KLRG1; and ILC3s (“ILC3” in Figure 2C, green events) were identified as Lin⁻NK⁻ILC2⁻ cells expressing NKp44. Using this approach, Lin⁻CD34⁻NK⁻ILC2⁻ILC3⁻CD117⁻CD127⁺ ILC1s were again scarcely detected (Figure 2D, purple box and data not shown, n = 11), consistent with other reports (Bernink et al., 2013; Simoni et al., 2017; Vely et al., 2016). Therefore we focused on investigating NK cell, ILC2, and ILC3 development in the tonsils.

Figure 2.

Human tonsil ILC subsets emerge directly from CD34⁻CD117⁺ ILCPs. Representative flow cytometry analyses of freshly enriched human tonsil ILCs after exclusion of lineage antigens (CD3, CD4, CD5, CD14, CD19, CD20, CD123, and FcxεR1α), henceforth referred to as Lin⁻. (A) Depiction of the human NK cell developmental pathway, in which the “NK” axis label = CD16, CD94, NKG2A, NKG2C, NKp80, KIR2D, and KIR3DL1-2. Antibodies targeting all seven NK cell markers were evaluated together in the same channel by flow cytometry. NK cells were painted red in the plots, which were gated on Lin⁻ lymphocytes. (B) Depiction of the putative human ILC2 developmental pathway, in which the “ILC2” axis label = CD294 and KLRG1. Antibodies targeting both ILC2 markers were evaluated together in the same channel by flow cytometry. ILC2s were painted blue in the plots, which were gated on Lin⁻ NK⁻ lymphocytes. (C) Depiction of the putative human ILC3 developmental pathway, in which the “ILC3” axis label = NKp44. ILC3s were painted green in the plots, which were gated on Lin⁻ NK⁻ILC2⁻ lymphocytes. CD117⁻ ILC3s are located within the green box in the right plot. (D) Depiction of CD127 expression by Lin⁻CD34⁻NK⁻ILC2⁻ILC3⁻CD117⁺ ILCPs. The location of where Lin⁻CD34⁻NK⁻ILC2⁻ILC3⁻CD117⁻CD127⁺ ILC1s would fall in the plot is designated by the purple box, purple arrow, and “ILC1” label. (A-C) Dashed red, blue, and green arrows represent the putative NK cell, ILC2, and ILC3 developmental pathways, respectively. CD34⁺CD117⁻ ETPs and CD34⁺CD117⁺ CILPs are designated by the black arrows and black labels. The data shown are representative of n = 10 fresh human tonsils from four independent experiments.

In accordance with prior findings (Freud et al., 2006), CD34⁺ cells (including CD34⁺CD117⁻ ETPs and CD34⁺CD117⁺ CILPs) were mutually exclusive from NK cells (Figure 2A, left plot). This pattern suggested that for tonsil-resident CD34⁺ cells to develop into mature NK cells the former must traverse CD34⁻NK⁻ developmental stage(s) (Figure 2A, left plot, red dashed arrow). Because CILPs express both CD34 and CD117 (Scoville et al., 2016), the putative NK cell developmental pathway ultimately stemming from CD34⁺CD117⁻ ETPs should run as illustrated by the red arrow in the middle plot of Figure 2A. Moreover, since immature NK cells express CD117 before they acquire CD16 or KIRs (Matos et al., 1993), we charted the NK cell developmental pathway as traversing through Lin⁻CD34⁻CD117⁺NK⁺ intermediates juxtaposed between Lin⁻CD34⁻CD117⁺NK⁻ and Lin⁻CD34⁻CD117⁻NK⁺ populations (Figure 2A, right plot). Thus, CD117 represented a useful bridging marker for the putative NK cell developmental pathway stemming from CD34⁺ progenitors in the tonsil.

Excluding NK cells from the analyses (Figure 2A, right plot, blue brackets), we next observed that CD34⁺ cells and ILC2s were also mutually exclusive, again defining a likely developmental pathway (Figure 2B, left plot, blue arrow). Moreover, as CD34⁺CD117⁺ CILPs serve as progenitors for ILC2s (Scoville et al., 2016) we could define the early ILC2 developmental pathway as differentiating from CD34⁺CD117⁻ ETPs to CD34⁺CD117⁺ CILPs as illustrated by the blue arrow in the middle plot of Figure 2B. We hypothesized that fully mature ILC2s do not express CD117 and in fact we could detect Lin⁻CD34⁻NK⁻CD117⁺ILC2⁺ putative intermediate cells poised between Lin⁻CD34⁻NK⁻CD117⁺ILC2⁻ and Lin⁻CD34⁻NK⁻CD117⁻ILC2⁺ populations (Figure 2B, right plot).

Likewise, among Lin⁻NK⁻ILC2⁻ cells in human tonsils (Figure 2B right plot, green brackets), CD34⁺ cells and NKp44⁺ ILC3s were mutually exclusive (Figure 2C, left plot) yet ILC3s are derived from CD34⁺CD117⁺ CILPs (Scoville et al., 2016) and co-express both CD117 and NKp44 when functionally mature (Cella et al., 2009; Cupedo et al., 2009). Thus, we could illustrate the putative ILC3 developmental pathway as also traversing a CD117⁺ intermediate population losing CD34 and lacking NKp44 (Figure 2C, middle and right plots), and then ultimately acquiring NKp44 (Figure 2C, green arrow in right plot). Small numbers of Lin⁻CD34⁻NK⁻ILC2⁻CD117⁻NKp44⁺ cells were also detected in the tonsils (Figure 2C, green box in right plot) consistent with a previous report (Bernink et al., 2013), but the ex vivo immunophenotypic and functional characteristics of this minor population have not been extensively evaluated. In considering that Lin⁻CD34⁻NK⁻ILC2⁻CD117⁻NKp44⁺ cells represent a minor subset of ILC3s we included this population in our analyses and referred to it as CD117⁻ ILC3s.

Finally, as shown in the left plot in Figure 2C (purple brackets) and in Figure 2D, the tonsil-resident Lin⁻CD34⁻NK⁻ILC2⁻ILC3⁻CD117⁺ population expressed CD127 consistent with CD34⁻CD117⁺ ILCPs (Lim et al., 2017). Therefore, these combined ex vivo mass and flow cytometry data supported a model whereby NK cells, ILC2s, and ILC3s can each develop in tonsils from Lin⁻CD34⁺CD117⁻ ETPs sequentially through Lin⁻CD34⁺CD117⁺ CILP and Lin⁻CD34⁻ CD117⁺ ILCP intermediates (arrows in Figure 1B and2A-C).

NK cells and ILC3s, but not ILC2s, emerge from a CD56⁺ subset of ILCPs

The mutual exclusivity of CD34⁺ progenitors and maturing ILCs within human tonsils (Figure 2A, 2B, and 2C, left plots) supported the hypothesis that the latter emerge directly from CD34⁻CD117⁺ ILCPs rather than directly from CD34⁺CD117⁺ CILPs. This notion was further supported by 3D dot plot analyses depicting the mutual exclusivity between CD34⁺ cells and mature ILC subsets (Figure 3A). Similarly, the 2D dot plots in Figure 3B show that while mature ILC markers were not detected when gating on total Lin⁻CD34⁺CD117⁺ CILPs (top row), such markers were detected when gating on total Lin⁻CD34⁻CD117⁺ cells, which included CD34⁻ CD117⁺ ILCPs (bottom row).

Figure 3.

The ILC2 pathway is mutually exclusive from those of NK cells and ILC3s, which overlap and emerge from a CD56⁺ ILCP subset. (A) Representative 3D flow cytometry dot plots depicting the mutual exclusivity of CD34⁺ cells from ILCs, the mutual exclusivity of ILC2s from NK cells and ILC3s, and the overlapping pattern of NK cells and ILC3s. (B) Similar findings as in (A) are depicted in 2D plots gated on either Lin⁻CD34⁺CD117⁺ cells (top row) or Lin⁻CD34⁻CD117⁺ cells (bottom row). (C) The left plot depicts a representative flow cytometric analysis of tonsillar Lin⁻CD34⁻ cells showing two CD117⁺ subsets according to CD56 expression. The six dot plots to the right show representative flow cytometry analyses of Lin⁻CD34⁻ CD117⁺CD56⁻ cells (top row) and Lin⁻CD34⁻CD117⁺CD56⁻ (bottom row) from fresh tonsils. ILC2s are detected among the Lin⁻CD34⁻CD117⁺CD56⁻ fraction but not among the Lin⁻CD34⁻CD117⁺CD56⁺ fraction. The data in (A-C) are representative of n = 10 fresh human tonsils from four independent experiments.

In addition to providing support for the direct emergence of mature ILCs from CD34⁻CD117⁺ ILCPs, these ex vivo flow cytometry data revealed that CD117⁺ ILC2s were mutually exclusive from CD117⁺ NK cells and CD117⁺ ILC3s. In contrast, some Lin⁻CD34⁻ CD117⁺ cells expressed both ILC3 and NK cell markers (Figure 3A and3B), suggesting that human NK cells and ILC3s could share a developmental pathway in tonsils that is mutually exclusive from the terminal steps of ILC2 development. Given that CD56 can be expressed by both NK cells and ILC3s but not by ILC2s (Cupedo et al., 2009; Spits et al., 2013), we hypothesized that CD56 expression would mark the divergence of ILC2s from a shared NK cell and ILC3 developmental pathway. In support of this, we identified a distinct CD56⁺ subset of Lin⁻ CD34⁻CD117⁺ ILCPs in the tonsils and from which NK cells and ILC3s, but not ILC2s, appeared to directly emerge (Figure 3C, left plot). Although mature markers from all three ILC subsets were detected when gating on Lin⁻CD34⁻CD117⁺CD56⁻ cells (Figure 3C, top row), only NK cell and ILC3 associated markers were detected when gating on Lin⁻CD34⁻CD117⁺CD56⁺ cells (Figure 3C, bottom row). These data and those described in Figures 1 and 2 suggested that the human tonsillar ILC2 developmental pathway diverges from a shared NK cell and ILC3 pathway at the CD34⁻CD117⁺ ILCP stage and in relation to the expression of CD56.

Human tonsil ILC developmental intermediates have distinct transcriptional profiles

To gain additional insight into the putative pathways of human ILC development, we sought to characterize ex vivo the following ten human tonsil populations that appeared to represent early, intermediate, and late stages of NK cell, ILC2, and ILC3 development: CD34⁺CD117⁻ ETPs; CD34⁺CD117⁺ CILPs; CD34⁻CD117⁺CD56⁻ and CD34⁻CD117⁺CD56⁺ ILCPs; CD117⁺ and CD117⁻ NK cells; CD117⁺ and CD117⁻ ILC2s; and CD117⁺ and CD117⁻ ILC3s. These populations are defined in Table S2 and were freshly purified as shown in Figure S2.

First, we performed total RNA sequencing analysis of each of the ten populations freshly purified from each of three donors’ tonsil specimens. Figure 4A shows a heatmap and unsupervised hierarchical clustering of the 30 sorted samples according to the relative expression of the 431 most differentially expressed genes. The ten ILC developmental intermediate populations segregated naturally based on distinct transcriptional profiles, although there was partial overlap among the three donors’ CD117⁺ and CD117⁻ ILC2s and among the three donors’ CD34⁻CD117⁺CD56⁻ and CD34⁻CD117⁺CD56⁺ ILCPs (Figure 4A). Figure 4B depicts the lists of genes whose average expression values were highest among the 431 genes that clustered the ten tonsil populations. Despite largely distinct gene expression profiles, some of the top genes were shared between CD34⁺CD117⁻ ETPs and CD34⁺CD117⁺ CILPs (MYB); between CD34⁻CD117⁺CD56⁻ and CD34⁻CD117⁺CD56⁺ ILCPs (LINC00861, LINC01011); between CD117⁺ and CD117⁻ NK cells (KLRF1, KLRD1, GZMK, GZMA, PYHIN1); between CD117⁺ and CD117⁻ ILC2s (HPGDS, C10orf128); and between CD117⁺ and CD117⁻ ILC3s (NELL2, ZNF319) (Figure 4B).

Figure 4.

Genome wide transcriptomic analyses reveal key differences among tonsil-derived ILC developmental intermediates. (A) Heatmap depicting the clustering of 431 differentially expressed genes among the ten indicated tonsil-derived populations as detected by RNA-seq. The relative expression of each gene is color coded with a scale based on z-score distribution from −4 (purple) to 4 (yellow). (B) Top ten expressed genes among the 431 total genes that differentiate the ten tonsil-derived populations. (C) PCA of the ten populations based on the average expression of the variable genes. Each population was sorted from n = 3 tonsils from one experiment. See also Figure S2 and Table S2.

Principle component analyses (PCA) also revealed distinct transcriptional profiles of the ten populations freshly isolated from the three human tonsils. For example, the PC 2 versus PC 3 plot shown in Figure 4C demonstrated that each of the ten populations from each donor’s tonsil clustered with the respective population from each of the other two donors’ tonsils (e.g. all three donors’ ETPs clustered together, etc.). In addition, the PCA shown in Figure 4C revealed that the CD34⁺CD117⁺ CILPs were positioned in between the CD34⁺CD117⁻ ETP and CD34⁻ CD117⁺ ILCP clusters on PC 2, further supporting our hypothesized model of progressive development from ETPs to CILPs to ILCPs. Moreover, the PCA showed that while the CD34⁻ CD117⁺CD56^−/+ ILCPs, CD117^+/− NK cells, and CD117⁺ ILC3s clustered in close approximation to one another, the CD117⁺ and CD117” ILC2s clustered separately and appeared to be “bridged” by the cluster of CD34⁻CD117⁺CD56⁻ ILCPs. Collectively, these data indicated that the ten putative ILC developmental intermediates identified through mass and flow cytometry (Figures 1-3) were indeed transcriptionally distinct cell populations. Moreover, these ex vivo gene expression data provided additional evidence to suggest that ILC2s can have a developmental pathway that is independent of NK cell and ILC3 developmental pathways in the human tonsil.

CD56⁻ and CD56⁺ ILCPs have similar functional and phenotypic profiles

We next sought to further characterize the ten tonsil-derived, putative ILC developmental intermediate populations functionally and for their expression of known ILC subset-associated markers. First, we stimulated enriched tonsil ILCs and progenitor cells with either ILC subset-specific cytokines for 24 hours or with phorbol 12-myristate 13-acetate (PMA), ionomycin, and IL-2 (PMA-I-2) for 6 hours and then assessed for intracellular cytokine production via flow cytometry. As predicted, the NK cells, ILC2s, and ILC3s produced IFN-γ, IL-13, and IL-22, respectively, whereas ETPs, CILPs, and CD56^−/+ ILCPs did not produce detectable amounts of these cytokines (Figure S3). We also observed that the proportions of cytokine⁺ cells were significantly higher among CD117” ILCs compared with their respective CD117⁺ ILCs in the ILC-specific stimulation conditions for NK cells and ILC3s and also in the PMA-I-2 condition for NK cells and ILC2s (Figure 5A-C). These data indicated that the capacities for mature ILC subset associated cytokine production were first acquired downstream of the ILCP stages of development. In addition, the data also indicated that while very rare in the tonsils, CD117⁻ ILC3s were qualitatively functionally similar to CD117⁺ ILC3s.

Figure 5.

CD56⁻ and CD56⁺ subsets of CD34⁻CD117⁺ ILCPs have similar ex vivo profiles and are distinct from CD34⁺CD117⁺ CILPs and CD117^+/− ILCs. (A-C) Comparisons of the average percentages of IFN-γ⁺ (A), IL-13⁺ (B), and IL-22⁺ cells (C) produced by CD117⁺ (black bars) vs CD117⁻ (gray bars) subsets of NK cells, ILC2s, and ILC3s, respectively (n = 6 from 3 independent experiments). Intracellular production of the aforementioned cytokines was measured via flow cytometry analysis following stimulation with either ILC-specific cytokines (NK: IL-12, IL-15, and IL-18; ILC2: IL-2, IL-25, and IL-33; ILC3: IL-2, IL-1β, and IL-23) for 24 hours or with P-I-2 for 6 hours. (D) Representative ex vivo flow cytometric analyses of the ten indicated tonsil-derived populations (n = 6 from 3 independent experiments for each of the indicated markers). Histogram data were generated by gating on Lin⁻ ILC populations as defined in Table S2. Isotype control staining of total Lin⁻ cells is shown at the top of each row. Black dashed rectangles highlight surface marker and TF expression of CD56⁻ and CD56⁺ ILCPs. Error bars in A-C indicate SEM. *p < 0.05, **p < 0.01,****p < 0.0001, ns = not significant. See also Figure S3 and Tables S2 and S4.

Next we characterized the ex vivo expression patterns of known ILC-associated surface markers and transcription factors using flow cytometry of freshly enriched Lin⁻ tonsil cells or realtime quantitative PCR analyses of freshly purified tonsil populations (Figures 5D and S3C-D). These data revealed distinct patterns of antigen acquisition and loss during the putative transitions from CD34⁺CD117⁻ ETPs to CD34⁺CD117⁺ CILPs to CD34⁻CD117⁺ ILCP stages and also at the putative late stages of terminal ILC differentiation associated with the downregulation of CD117. We noted that ex vivo tonsil-derived CD34⁻CD117⁺CD56⁻ and CD34⁻CD117⁺CD56⁺ ILCPs showed nearly identical expression patterns of these ILC-associated markers (Figures 5D and S3C-D, black dashed rectangles) despite showing quite distinct phenotypic profiles compared to their putative immediate precursors (CD34⁺CD117⁺ CILPs) and progeny (CD117⁺ ILCs). Therefore, in conjunction with the ex vivo flow cytometry patterns described earlier (Figure 3C), the extensive phenotypic data shown in Figure 5 indicated that CD56 was unique among all of the markers tested in denoting a possible point of ILC2 divergence from NK cells and ILC3s at the level of CD34⁻CD117⁺ ILCPs.

CD56⁺ ILCPs can produce NK cells and ILC3s but not ILC2s

Given our ex vivo phenotypic data suggesting that CD56 marks a point of divergence of ILC2s from NK cells and ILC3s (Figure 3C), we tested the hypothesis that freshly purified tonsillar CD56⁻ and CD56⁺ ILCPs have distinct lineage differentiation potentials. Each CD34⁻CD117⁺ ILCP subset was freshly sorted by CD56 expression from tonsils to ≥98% purity (Figure S2) and then cultured in vitro for 28 days with murine OP9-DL1 stroma plus recombinant human Flt3 ligand (FL) and IL-7. These in vitro conditions were used because they support the simultaneous development of NK cells, ILC2s, and ILC3s from CD34⁺CD117⁻ ETPs and CD34⁺CD117⁺ CILPs (Scoville et al., 2016). In addition, the ILCs produced under these in vitro conditions were similar to those freshly purified from tonsils with regards to the specific associations of ILC subset associated surface markers with their respective cytokines and RORγt expression (Figure S4). Of note, only minute populations of putative non-NK ILC1s (defined functionally) were detected (<1% of total ILCs, n = 10, Figure S4A and S4B, black boxes in left plots), similar to what we previously observed when evaluating CD34⁺CD117⁺ CILP-derived ILCs generated under the same conditions (Scoville et al., 2016).

As shown in Figures 6A and S5 (black dashed boxes) tonsil-derived CD34⁻ CD117⁺CD56⁻ ILCPs generated mutually distinct populations of NK cells, ILC2s, and ILC3s, whereas CD34⁻CD117⁺CD56⁺ ILCPs that were cultured in parallel and under identical conditions produced NK cells and ILC3s but essentially no ILC2s. As previously performed, cell tracer dye labeling experiments confirmed that these results were not due to rare contaminating mature ILCs present in the sorted ILCP populations (data not shown) (Freud et al., 2016). In addition, these bulk culture findings were further substantiated by data obtained from clonal culture assays initiated with individually sorted CD56⁻ ILCPs or CD56⁺ ILCPs and supplemented with stem cell factor (SCF) and IL-2 during the first two weeks of culture to facilitate proliferation (Yang et al., 2015). Notably, NK cells, ILC2s, and/or ILC3s were derived from CD56⁻ ILCP clones (94 total clones from n = 4 tonsils; 18% cloning efficiency), whereas only NK cells and/or ILC3s were derived from CD56⁺ ILCP clones (83 total clones from n = 4 tonsils; 15.5% cloning efficiency) (Figures 6B and6C). No clones derived from CD56⁻ ILCPs exclusively generated combinations of ILC2s and ILC3s or of ILC2s and NK cells despite the fact that some clones from both CD56⁻ and CD56⁺ ILCPs exclusively generated combinations of NK cells and ILC3s (Figure 6B and S6). In addition, any clones containing NK cells also contained ILC3s, further supporting the notion that NK cells and ILC3s are closely developmentally related. Thus, our bulk and clonal culture assays revealed that, under the conditions tested, the expression of CD56 on tonsil-derived CD34⁻CD117⁺ ILCPs was associated with a loss of potential for ILC2 lineage differentiation.

Figure 6.

CD34⁻CD117⁺CD56⁺ ILCPs can produce NK cells and ILC3s but not ILC2s. (A, D, E, F). Representative surface flow cytometry analyses of ILCs generated in vitro following a 28-day culture of the indicated freshly purified tonsil-derived populations (see labels in left column) with OP9-DL1 feeder cells and recombinant human IL-7 and FL. Dot plots were gated on total viable Lin⁻CD45⁺ lymphocytes, in which Lin = CD3, CD5, and CD14. NK cells were defined as Lin⁻CD45⁺CD94⁺CD294⁻NKp44^+/− cells (red boxes); ILC2s were defined as Lin⁻CD45⁺CD94⁻CD294⁺NKp44⁻ cells (blue boxes); and ILC3s were defined as Lin⁻CD45⁺CD94⁻CD294⁻NKp44⁺ cells (green boxes). Data shown are representative of n = 20 tonsils from ten independent experiments. (B) Clonal analysis of freshly purified tonsil-derived CD56⁻ and CD56⁺ ILCPs individually sorted into wells containing OP9-DL1 feeder cells supplemented with recombinant IL-2 (first two weeks only), SCF (first two weeks only), IL-7, and FL for 28 days. Clones were analyzed for NK cells, ILC2s, and ILC3s (n = 4 tonsil donors, two independent experiments; 94 CD56⁻ clones and 83 CD56⁺ clones were analyzed using this panel). Cloning efficiency was 18% for CD56⁻ clones and 15.5% for CD56⁺ clones. (C) Average percent frequency of ILCs produced from CD56⁻ ILCPs (black bars) versus CD56⁺ ILCPs (gray bars). Error bars indicate SEM. ***p < 0.001, ****p < 0.0001, ns = not significant. See also Figures S2 and S4-S7.

CD117⁺ ILCs retain lineage differentiation plasticity

We also compared the lineage differentiation potentials of tonsil-derived CD117⁺ ILCs with those of CD117⁻ ILCs. While tonsil-derived CD117⁻ NK cells gave rise to virtually pure populations of NK cells in culture, some tonsil-derived CD117⁺ NK cells lost their apparent NK cell differentiation status and gave rise to distinct populations of ILC3s that were capable of producing IL-22 (Figures 6D and S5B). Only rare ILC2s were produced from the CD117⁺ NK cells. Likewise, CD117⁺ ILC2-initiated cultures also contained some NK cells and ILC3s, whereas CD117” ILC2-initiated cultures yielded essentially pure populations of ILC2s (Figures 6E and S5C). Under the same conditions, both CD117⁺ and CD117⁻ ILC3-initiated bulk cultures yielded ILC3s as well as some NK cells; however, negligible ILC2s were detected (Figures 6F and S5D). Combined analyses of the percentages of NK cells, ILC2s, and ILC3s made from each population are shown in Figures S5E-G The differentiation potentials as determined through the use of surface immunophenotypic data (Figure 6) were consistent with the functional cytokine production data (Figures S5E-G). These data indicated that while lineage differentiation plasticity was at least partially retained among CD117⁺ ILCs purified ex vivo, CD117⁻ NK cells and CD117⁻ ILC2s were essentially completely committed to their respective lineages under the conditions tested. These data also further supported our hypothesis that human NK cells and ILC3s are more closely developmentally related as compared to ILC2s, given that at least some CD117⁺ NK cells and CD117⁺ and CD117⁻ ILC3s could give rise to each other but showed negligible ILC2 differentiation potential under the conditions tested.

The observation that in vitro lineage differentiation plasticity was retained by the population of tonsillar CD117⁺ NK cells (Figures 6D and S5) argues against the notion of early-stage NK cell lineage commitment, which has been proposed to occur at the level of a CD34⁺ NKP (Renoux et al., 2015). Indeed tonsil-derived human CD34⁻CD117⁺ ILCPs gave rise to both NK cells and non-NK ILCs in vivo following transplantation into busulfan-pretreated nonobese diabetic (NOD)-scid IL2Rgammanull (NSG) immunodeficient mice that were then injected twice weekly for four weeks with recombinant human FL, SCF, IL-1β, IL-2, IL-7, IL-15, and IL-23 (Figure S7A), consistent with findings by the Di Santo laboratory (Lim et al., 2017). Moreover, we observed that fresh tonsil-derived CD34⁺CD38⁺CD123⁻CD45RA⁺CD7⁺CD10⁺CD127− putative human NKPs (Renoux et al., 2015) that were cultured with OP9-DL1 stroma and recombinant IL-7 and FL for 28 days generated T cells, NK cells, ILC2s, and ILC3s, with T cells representing the most abundant Lin⁻CD45⁺ subset produced in each experiment (n = 13) (Figure S7B-E). Collectively, these data support the notion that human NK cell lineage differentiation commitment occurs at a late stage of development in the tonsil.

CILPs and CD56⁻ ILCPs can differentiate into CD56⁺ ILCPs

The combined ex vivo cytometric, transcriptional, phenotypic, and functional data described above supported a model whereby human NK cells, ILC2s, and ILC3s can develop in tonsils from tissue-resident CD34⁺CD117⁻ ETPs that sequentially differentiate through CD34⁺CD117⁺ CILP and CD34⁻CD117⁺ ILCP intermediates. Furthermore, they suggested that CD56 expression at the ILCP stage marks the divergence of ILC2s from a shared NK cell and ILC3 developmental pathway. A corollary to this hypothesis is that CD56⁺ ILCPs are the direct progeny of CILPs and/or CD56⁻ ILCPs. To determine if CD34⁺CD117⁺ CILPs and/or CD34⁻ CD117⁺CD56⁻ ILCPs could generate CD34⁻CD117⁺CD56⁺ ILCPs, we isolated CILPs and CD56⁻ ILCPs from fresh tonsils and then cultured these populations for seven days with OP9-DL1 stroma and human IL-7 and FL. At this time point, tonsil-derived CILPs (Figure 7A) and tonsil-derived CD56⁻ ILCPs (Figure 7B) gave rise to both Lin’CD45⁺CD34’CD117⁺CD56’ and Lin⁻ CD45⁺CD34⁻CD117⁺CD56⁺ cells that lacked NK cell (CD94), ILC2 (CD294), and ILC3 (NKp44) markers. Furthermore, following isolation of the day 7 Lin⁻CD45⁺CD34⁻CD117⁺CD56⁻ and Lin⁻ CD45⁺CD34⁻CD117⁺CD56⁺ subsets and subsequent re-culture for an additional 21 days, we observed that the Lin⁻CD45⁺CD34⁻CD117⁺CD56⁻ cells gave rise to NK cells, ILC2s, and ILC3s, whereas the Lin⁻CD45⁺CD34⁻CD117⁺CD56⁺ cells gave rise to NK cells and ILC3s but not ILC2s (Figure 7C). These data reflected what we observed when culturing fresh tonsil-derived CD56⁻ and CD56⁺ ILCPs (Figures 6 and S5) and thus supported the conclusion that both CILPs and CD56⁻ ILCPs gave rise to CD56⁺ ILCPs. These precursor-progeny relationships are depicted in a comprehensive model of human tonsil ILC development shown in Figure 7D.

Figure 7.

CD34⁺CD117⁺ CILPs and CD34⁻CD117⁺CD56⁻ ILCPs can differentiate into CD34⁻ CD117⁺CD56⁺ ILCPs. (A-B) Representative surface flow cytometry analyses of cells derived from 7-day in vitro cultures of fresh tonsil-derived CILPs (A) or CD56⁻ ILCPs (B) with OP9-DL1 feeder cells and recombinant human IL-7 and FL. Dot plots were gated on total viable Lin⁻CD45⁺ lymphocytes, in which Lin = CD3, CD4, CD5, CD14, CD19, CD20, CD123, and FcεR1α (n = 6 tonsils, three independent experiments). (C) Representative surface flow cytometry analyses of ILCs generated following purification of day 7 in vitro-derived CD56⁻ and CD56⁺ ILCPs (as shown in B) that were re-sorted and then cultured again for an additional 21 days with OP9-DL1 stroma, IL-7, and FL (n = 6 tonsils, three independent experiments). NK cells were defined as Lin⁻CD45⁺CD94⁺CD294⁻NKp44^+/− cells (red boxes); ILC2s were defined as Lin⁻CD45⁺CD94⁻ CD294⁺NKp44⁻ cells (blue boxes); and ILC3s were defined as Lin⁻CD45⁺CD94⁻CD294⁻NKp44⁺ cells (green boxes). (D) Model of human ILC development in tonsils. Solid black lines represent developmental transitions between the indicated cells supported by ex vivo immunophenotypic and molecular data as well as lineage differentiation assays. The thin dashed black line represents transdifferentiation between CD117⁺ NK cells and CD117⁺ ILC3s.

DISCUSSION

In the murine system, multiple elegant adoptive transfer and genetic fate-mapping studies have facilitated the identification of key steps and cellular intermediates in the development of each ILC subset (Constantinides et al., 2014; Klose et al., 2014; Satoh-Takayama et al., 2010; Vonarbourg et al., 2010). These data have in turn led to the generation and wide acceptance of a comprehensive model of mouse ILC development, providing a fundamental framework for ongoing investigations into the molecular regulation of innate lymphopoiesis (Zook and Kee, 2016). Recent studies have also identified human progenitor populations that appear to be analogous to those described in mice (Freud et al., 2006; Lim et al., 2017; Montaldo et al., 2014; Renoux et al., 2015; Scoville et al., 2016). However, the complete picture of human ILC development has remained less well defined than its murine counterpart (Scoville et al., 2017).

In this study, we performed an extensive ex vivo phenotypic, molecular, and functional analysis of fresh, human, tonsil-derived, Lin⁻ lymphoid cells in order to formulate a comprehensive model of human NK cell, ILC2, and ILC3 development in mucosal SLTs. Working under the premises that human ILCs ultimately derive from CD34⁺ progenitor cells and can develop in these tissues, we provide evidence for the coexistence and partial overlap of NK cell, ILC2, and ILC3 developmental continuums in human tonsils. Specifically, our data suggest that each of these ILC subsets can develop along a shared pathway starting from CD34⁺CD117⁻ ETPs and progressing through CD34⁺CD117⁺ CILP and CD34⁻CD117⁺ ILCP intermediates. The latter then give rise to CD117⁺ and CD117⁻ ILCs that are capable of producing subset associated cytokines. As our ex vivo immunophenotypic and molecular profiling data suggest, ILCs appeared to directly emerge from CD34⁻CD117⁺ ILCPs rather than directly from CD34⁺CD117⁺ CILPs, implicating the ILCP stage as a likely requisite step during ILC development within the tonsil.

Our ex vivo analyses also indicated that as ILC2s emerged from the pool of ILCPs they were mutually exclusive from developing NK cells and ILC3s. In contrast, the NK cell and ILC3 pathways appeared to partially overlap suggestive of a close developmental relationship between NK cells and ILC3s. The latter was further supported by the identification of two distinct ILCP subsets in the tonsils, with the CD34⁻CD117⁺CD56⁻ ILCP subset demonstrating NK cell, ILC2, and ILC3 lineage differentiation potential and the CD34⁻CD117⁺CD56⁺ ILCP subset demonstrating NK cell and ILC3 potential but lacking ILC2 potential under the conditions tested. The human CD34⁻CD117⁺CD56⁺ ILCP population appears to represent a pool of restricted NK cell and ILC3 precursors, which to the best of our knowledge has not been described in mice or humans. Notably, our ex vivo immunophenotypic and single-cell clonal culture data also failed to provide any concrete evidence for the existence of a human tonsil-resident CHILP population analogous to the CHILP described in mice (Constantinides et al., 2014; Klose et al., 2014). Therefore, our studies highlight potential major distinctions between human and mouse ILC development, with human ILC2s diverging away from a shared NK and ILC3 developmental pathway. Nonetheless we note that while we used tonsils as the source of human ILC developmental intermediates in this study, many studies of mouse ILC development were performed using fetal liver- and/or bone marrow-derived progenitor cells (Constantinides et al., 2014; Klose et al., 2014; Serafini et al., 2014). Therefore, some of these differences between human and mouse ILC development may be related to tissue-specific effects. Further investigation will be necessary to determine to what extent proposed models of human and mouse ILC development are applicable across a broad range of tissues including non-lymphoid tissues where ILCs may be resident and develop locally (Gasteiger et al., 2015; Lim et al., 2017; Moro et al., 2016).

In light of the association of CD56 with the NK cell and ILC3 developmental pathways, we note that CD56 expression was bimodally expressed on the tonsil ILC3 populations. Therefore, while it is anticipated that CD34⁻CD117⁺CD56⁺ ILCPs represent a physiologic pool of ILC3 and NK cell precursors, ILC3s may also derive directly from CD34⁻CD117⁺CD56⁻ ILCPs. Our polyclonal and clonal in vitro culture data support this notion and imply that CD56 is not required for ILC3 development. In contrast, most mature NK cells in human tissues constitutively express CD56, and a recent study by Mace et al demonstrated that CD56 mediates the formation of an important developmental synapse between immature human NK cell precursors and stromal cells that can support NK cell functional maturation (Mace et al., 2016). Therefore, CD56 expression may be required for human NK cell development in vivo, and we hypothesize that the developmental pathway from the CD34⁻CD117⁺CD56⁺ ILCP to the mature NK cell is functionally operative and potentially dominant in vivo.

One unexpected finding in our study was the observation that only CD34⁻CD117⁻ NK cells and CD34⁻CD117⁻ ILC2s demonstrated complete lineage commitment under the conditions tested, whereas at least some CD34⁻CD117⁺ cells expressing ILC subset-associated markers demonstrated lineage differentiation plasticity in vitro. These data challenge the principle of early lineage commitment that are based on mouse data as well as on previous reports of putative lineage-committed CD34⁺ human progenitor populations such as ILC3 lineage-specified progenitors (Montaldo et al., 2014) and NKPs (Renoux et al., 2015). Indeed it was previously shown that cells designated as ILC3 lineage-specified progenitor cells, which express CD34 and CD117 and hence phenotypically overlap with CD34⁺CD117⁺ CILPs, can also give rise to non-ILC3 cells in vitro and in vivo (Freud et al., 2006; Scoville et al., 2016). Likewise, we demonstrated in this study that fresh tonsil-derived putative human NKPs, which express CD34 and CD10 and hence phenotypically overlap with CD34⁺CD117⁻ ETPs, gave rise to heterogeneous mixtures of T cells and ILCs following in vitro culture under the conditions used in this study. Therefore, the principle of early lineage commitment may not be entirely accurate, at least with regards to human ILC development in the tonsil. Continued investigation of this process is warranted.

In light of the above, our findings do not exclude the possibility that other ILC differentiation pathways exist in other tissues. Indeed CD34⁺ progenitor cells similar to those present in human SLTs also exist in the liver, intestine, and uterus (Yu et al., 2013). Likewise, human fetal liver, umbilical cord blood, adult lungs, and adult peripheral blood harbor CD34⁻CD117⁺ ILCPs (Lim et al., 2017), which are phenotypically and functionally similar to CD56⁻ ILCPs in tonsils. Additionally, mesenchymal cells were recently shown to support late-stage ILC differentiation in peripheral tissues (Koga et al., 2018). Therefore, it is likely that ILCs can differentiate in multiple tissues throughout the body. Moreover, we cannot exclude the existence of a human CHILP or other functionally distinct early lineage-committed ILC progenitors in other issues. Finally, we cannot exclude the possibility that under certain conditions in vivo, ILC2s could be produced from CD34⁻CD117⁺CD56⁺ ILCPs. More investigation is warranted to evaluate ILC development in other human tissues and also to determine how ILC1s that are expanded in some inflammatory conditions might be generated from progenitor and precursor populations described here and by other groups.

In summary, we have elucidated and characterized putative pathways of NK cell, ILC2, and ILC3 development in human tonsils. We have provided extensive ex vivo phenotypic, transcriptional, and in vitro clonal culture data supporting the notion that the ILC2 developmental pathway diverges from overlapping NK cell and ILC3 developmental pathways, which emerge from the CD34⁻CD117⁺ ILCP stage and correlate with the acquisition of CD56. We also identified a distinct CD56⁺ subset of human tonsillar CD34⁻CD117⁺ ILCPs whose developmental potential was restricted to NK cells and ILC3s. These findings support a comprehensive model for human ILC development in SLTs that is distinct from the established murine model of ILC development, advancing our understanding of basic human ILC biology and of the innate immune system.

CONTACT FOR REAGENT AND RESOURCE SHARING

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Aharon G. Freud (aharon.freud@osumc.edu).

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Human tissue samples

Human tissues used in this study were collected in accordance with protocols approved by The Ohio State University Institutional Review Board (OSU IRB), and donor consent was acquired when appropriate under the approved OSU IRB protocol. Human pediatric tonsils were obtained fresh via the NCI-approved Cooperative Human Tissue Network (CHTN) from Nationwide Children’s Hospital (Columbus, OH) as previously executed (Hughes et al., 2014).

METHODS DETAILS

Cell isolation, selection, and sorting

Human ILCs were enriched from fresh tonsil specimens as previously described (Scoville et al., 2015). When appropriate, the following selections were performed from enriched ILC populations: 1) CD34⁺ selection using the Indirect CD34 MicroBead Kit (Miltenyi); 2) Biotin⁺ selection of CD34⁻ cells labeled with biotinylated NK-specific surface antibodies (CD16, CD94, NKG2A, NKG2C, NKp80, KIR2D, and KIR3DL1-2) using the Anti-Biotin MicroBead Kit (Miltenyi); and 3) PE⁺ selection of CD34⁻biotin⁻ cells labeled with PE CD117 using the Anti-PE MicroBead Kit (Miltenyi). All selections were carried out according to the manufacturer’s instructions with LS columns (Miltenyi). Each population was sorted to ≥98% purity with a FACSAriaII sorter (BD Biosciences) (Figure S2).

Mass cytometry

Five million enriched human ILCs were stained for mass cytometry analyses as previously described (Bendall et al., 2011; Horowitz et al., 2013). Antibodies were purchased or labeled using Maxpar-X8 labeling reagent kits based on the manufacturer’s instructions (Fluidigm). Cells were fixed with Smart tube buffer for 10 min at room temperature (RT), washed with cell staining media (CSM) and incubated with surface antibodies for 50 min at RT. Cells were then washed three times with CSM, fixed with 1.5% paraformaldehyde, and permeabilized with ice cold methanol at −20°C for 15 min. Following incubation, cells were washed thrice with CSM and stained with intracellular antibodies for 50 min at room temperature. Cells were then washed twice with CSM and incubated with iridium intercalator pentamethylcyclopentadienyl-Ir(III)-dipyridophenazine (Fluidigm) Intercalator solution with 1.5% paraformaldehyde (1:4000 dilution) at 4°C until further analysis. Excess intercalator was removed by washing cells once with CSM and twice with deionized water. Cells were re-suspended at a concentration of 1 million/ml of pure water mixed with 4 Elemental Equilibration beads (Fluidigm). Cells were acquired on a third-generation Helios mass cytometer at an event rate of 200-400 events per second. Noise reduction was utilized, sigma=3, and the event duration and lower convolution threshold were 8-150 and 600, respectively. Data were normalized using the method of Finck et al. (Finck et al., 2013). A compensation matrix was used to account for expected small (1-2%) spillovers from isotopic contamination in the metal isotopes. Data were evaluated using viSNE analysis (Amir el et al., 2013). All mass cytometry antibodies used are listed in Supplemental Experimental Procedures (Table S1).

RNA sequencing, library construction, and analysis

Extracted RNA samples underwent quality control assessment using the RNA Pico 6000 chip on Bioanalyzer 2100 (Agilent) and were quantified with Qubit Fluorometer (Thermo Fisher). The RNA libraries were prepared and sequenced at University of Houston Seq-N-Edit Core per protocols. Total libraries were prepared with Ovation® RNA-Seq System V2 (NuGen) and Ovation® Ultralow Library System V2 (NuGen) using 500 pg-2 ng input RNA. The size selection for libraries were performed using SPRIselect beads (Beckman Coulter) and purities of the libraries were analyzed using the High Sensitivity DNA chip on Bioanalyzer 2100 (Agilent). The prepared libraries were pooled and sequenced using NextSeq 500 (Illumina), generating ~20 million 2×76 bp paired-end reads per samples. RNA sequencing data were aligned to Human Genome Reference Consortium GRCh38 (https://www.ncbi.nlm.nih.gov/assembly/GCF000001405.38) using the STAR alignment tool (Dobin et al., 2013). Normalization and differential gene expression analysis was performed using the DESeq2 tool (Love et al., 2014). Visualizations for Differential Expression Heat map and Principle Component analysis were generated using Trinity Differential Expression Analysis Tools (Haas et al., 2013). All bioinformatics programming was performed using bash commands in the Linux/ Ubuntu system.

In vitro cultures

OP9-DL1 feeder cells were maintained in MEM-α + Glutamax (Thermo Fisher Scientific) with 20% fetal bovine serum and 1% antibiotic and antimycotic. Differentiation assays used media containing DMEM and F12 (2:1), 1% antibiotic and antimycotic (Thermo Fisher Scientific), 20 mg/mL ascorbic acid, 24 mM 2-mercaptoethanol, 0.05 mg/mL sodium selenite (Sigma), and 10% heat-inactivated human AB serum (Valley Biomedical) (Cichocki and Miller, 2010). One night before culture, 1 × 10⁴ stromal cells were pre-seeded in 24-well tissue culture plates for bulk cultures. Sorted tonsil-derived populations (500 – 1000 cells/well) were plated on non-irradiated OP9-DL1 stroma in media supplemented with 20ng/mL human IL-7 and 10ng/mL human FL (Miltenyi) (Cichocki and Miller, 2010). Fresh media and cytokines were replenished every 3 days and the OP9-DL1 stromal layer was replaced every 7 days. For single cell clonal assays, CD56⁻ or CD56⁺ ILCPs were sorted directly into 96-well plates pre-seeded with 2 × 10³ stromal cells after an initial round of sorting. Every 3 days, clones were supplemented with recombinant human 20ng/mL human IL-7 and 10ng/mL human FL with the following additional cytokines for the first two weeks of culture: human SCF and IL-2 (20ng/ml each, Miltenyi). Op9-DL1 stromal cells were not replenished weekly for the clonal experiments. Clones were identified as wells containing visible clusters of cells not detectable in wells seeded only with OP9-DL1 stroma. Overall cloning efficiency was 18% for CD56⁻ ILCPs (n = 4 tonsils, 94 total clones analyzed) and 15.5% for CD56⁺ ILCPs (n = 4 tonsils, 83 total clones analyzed). Cultures were grown for a total of 28 days and harvested for surface or intracellular cytokine profiling via flow cytometric analysis.

Flow cytometric analysis of surface and intracellular markers and cytokine production

Ex vivo and in vitro derived ILC populations were first stained with Flexible Viability Dye eFluor 506 (eBiosciences) for 10 min, followed by a 15 min surface stain with the appropriate antibodies on ice. For intracellular cytokine analysis, cells were either stimulated for 1) 24 hr with combinations of the following ILC-specific cytokines: IL-2 (1 nM) (Peprotech), IL-12 (10ng/ml), IL-15 (10ng/ml), IL-18 (10ng/ml), IL-25 (10ng/ml), IL-33 (10ng/ml), IL-1β (10ng/ml), and/or IL-23 (10ng/ml) (Miltenyi) (Bernink et al., 2015; Freud et al., 2016; Mjosberg et al., 2011); or 2) 6 hr with IL-2, PMA (81 nM), and ionomycin (1.34 mM) (eBioscience). Brefeldin A (BD Biosciences) was added 4 hr prior to collection. Intracellular staining was performed using the Cytofix and Cytoperm Fixation and Permeabilization Solution Kit (BD Biosciences) for cytokine analysis or the Foxp3 Transcription Factor Staining Buffer Set (eBiosciences) for transcription factor analysis. Cells were acquired on an LSRII cytometer (BD Biosciences) and analyzed using FlowJ10 (TreeStar) as previously described (Scoville et al., 2016). To determine nonspecific staining, the appropriate use of unstimulated controls and isotype controls was performed. All antibodies utilized for flow cytometry immunophenotyping and post culture analyses are listed in Supplemental Experimental Procedures (Table S3).

In vivo transplantation of ILC precursors in NSG mice

Experiments involving NSG mice (purchased from The Jackson Laboratory) were performed in accordance with a protocol approved by the OSU Institutional Animal Care and Use Committee. One day prior to injection of human ILCPs, 25 mg/kg busulfan was injected into 6-10 week old female NSG mice to promote engraftment. The following day, 40,000-50,000 tonsil-derived human CD34⁻CD117⁺ ILCPs were sorted to >98% purity and injected intravenously into NSG mice. At day zero and twice weekly thereafter for four weeks, the NSG mice were given intraperitoneal injections of the following recombinant human cytokines: FL, SCF, IL-1β, IL-2, IL- 7, IL-15, and IL-23 (0.5 μg/mouse). After four weeks, the mice were sacrificed, and their blood, spleens, bone marrow, lungs, and small intestinal lamina propria were harvested and analyzed using flow cytometry. Murine leukocytes were excluded using an anti-mouse CD45.1 antibody and human leukocytes were positively identified using an anti-human CD45 antibody.

RNA isolation and qPCR

Each population was sorted and mRNA isolated using the Total RNA Purification Plus Kit (Norgen Biotek). Reverse transcription was performed utilizing the Superscript VILO master mix (Life Technologies). Gene expression was normalized to 18S mRNA, and relative quantification was performed using the ΔΔCt method. The y-axes portray fold differences in mRNA expression quantified based on the population in which expression was lowest (arbitrarily normalized to 1). All primers used are listed in Table S4 of Supplemental Experimental Procedures.

QUANTIFICATION AND STATISTICAL ANALYSIS

Data are represented as mean ± SEM unless specified. Sample sizes for each experiment and the replicate number of experiments are included in the figure legends. Linear mixed effects models were used for analysis to take into account the correlation of multiple observations from the same donor. Holm’s procedure was used to control for multiple comparisons. Adjusted p-values < 0.05 are considered as significant. All analyses were performed using SAS9.4 (SAS Institute, Inc., NC).

DATA AND SOFTWARE AVAILABILITY

The RNA-seq raw data files (accession numbers SAMN09767432, SAMN09767433, SAMN09767434, SAMN09767435, SAMN09767436, SAMN09767437, SAMN09767438, SAMN09767439, SAMN09767440, SAMN09767441, SAMN09767442, SAMN09767443, SAMN09767444, SAMN09767445, SAMN09767446, SAMN09767447, SAMN09767448, SAMN09767449, SAMN09767450, SAMN09767451, SAMN09767452, SAMN09767453, SAMN09767454, SAMN09767455, SAMN09767456, SAMN09767457, SAMN09767458, SAMN09767459, SAMN09767460, SAMN09767461) are available at the NCBI Gene Expression Omnibus (BioSample) archive.

HIGHLIGHTS:

• Human NK cells, ILC2s, and ILC3s develop from tonsil CD34⁻CD117⁺ ILC precursors

• A CD56⁺ subset of CD34⁻CD117⁺ precursors generates NK cells and ILC3s but not ILC2s

• Human ILC development is distinct from the established murine model