Human innate lymphoid cell precursors express CD48 that modulates ILC differentiation through 2B4 signaling

See allHide authors and affiliations

Science Immunology  20 Nov 2020:
Vol. 5, Issue 53, eaay4218
DOI: 10.1126/sciimmunol.aay4218

Differentiation distinctions

Murine innate lymphoid cell (ILC) differentiation has been studied in detail, but human ILC development is less well understood because of differences in surface markers between species. Tufa et al. show that human common innate lymphoid progenitors (CILPs), characterized as CD34+CD117+α4β7+Lin, express CD48 and CD52 and can give rise to NK cell progenitors (NKPs) and ILC progenitors. CD34+CD117+α4β7+LinCD48CD52+ and CD34+CD117+α4β7+LinCD48+CD52+ NKP differentiated into two different NK cell subsets under in vitro differentiation conditions. ILCPs within the CD34+CD117+α4β7+LinCD48+CD52+ subset could differentiate into ILC1s, ILC2s, and NCR+ ILC3s, but CD34+CD117+α4β7+LinCD48+CD52 ILCPs gave rise to a different ILC3 subset with lymphoid tissue inducer–like properties. Ligation of the 2B4 receptor by CD48 was needed for ILC2 development, and these results define a role for CD48 in human ILC differentiation.


Innate lymphoid cells (ILCs) develop from common lymphoid progenitors (CLPs), which further differentiate into the common ILC progenitor (CILP) that can give rise to both ILCs and natural killer (NK) cells. Murine ILC intermediates have recently been characterized, but the human counterparts and their developmental trajectories have not yet been identified, largely due to the lack of homologous surface receptors in both organisms. Here, we show that human CILPs (CD34+CD117+α4β7+Lin) acquire CD48 and CD52, which define NK progenitors (NKPs) and ILC precursors (ILCPs). Two distinct NK cell subsets were generated in vitro from CD34+CD117+α4β7+LinCD48CD52+ and CD34+CD117+α4β7+LinCD48+CD52+ NKPs, respectively. Independent of NKPs, ILCPs exist in the CD34+CD117+α4β7+LinCD48+CD52+ subset and give rise to ILC1s, ILC2s, and NCR+ ILC3s, whereas CD34+CD117+α4β7+LinCD48+CD52 ILCPs give rise to a distinct subset of ILC3s that have lymphoid tissue inducer (LTi)–like properties. In addition, CD48-expressing CD34+CD117+α4β7+Lin precursors give rise to tissue-associated ILCs in vivo. We also observed that the interaction of 2B4 with CD48 induced differentiation of ILC2s, and together, these findings show that expression of CD48 by human ILCPs modulates ILC differentiation.


Innate lymphoid cells (ILCs) have been classified into three major groups based on the expression of transcription factors that drive cell function (1). Group 1 ILCs (ILC1) express Tbet and produce interferon-γ (IFN-γ) upon activation (2, 3). They are functionally similar to the traditional natural killer (NK) cells but lack granule-based cytotoxicity (2, 3). Group 2 ILCs (ILC2) express GATA3 and produce interleukin-5 (IL-5) and IL-13 upon activation (4, 5). Group 3 ILCs (ILC3) express RORγt (retinoic acid–related orphan nuclear receptor γ) and, following stimulation, express IL-22 (6, 7).

Each ILC group has multiple subpopulations, characterized by distinct surface receptor expression, gene profiles, and functionality. Expression of CD103, NKp44, and CD127 defines two ILC1 subsets that initiate IFN-γ responses against pathogens (2, 8). ILC2 subpopulations vary in their activation-induced cytokine production [IL-5/IL-13 or transforming growth factor–β (TGF-β)/IL-10] (911), with the former being GATA3 dependent and having a role in maintaining homeostatic type 2 responses (4, 5), whereas the latter are ID3 (inhibitor of DNA binding 3) dependent and regulate intestinal inflammation by attenuating other ILC populations (9). At least two human ILC3 subpopulations have been described, including the lymphoid tissue inducer (LTi)–like cells and the natural cytotoxicity receptor (NCR)–expressing ILC3 cells (7, 12). Both ILC3 subsets are known to maintain intestinal tissue homeostasis, but evidence suggests that these two subsets also have distinct functions (12). These data highlight the previously underappreciated diversity within the ILC compartment and raise questions about whether these differences are alternative states of activation and ILC plasticity, or conversely, whether they are different cell types that arise through distinct developmental trajectories (1315).

In mice, ILCs develop from common lymphoid progenitors (CLPs), which further differentiate into the common ILC progenitor (CILP). The CILPs, in turn, develop into either an NK progenitor (NKP) or the common “helper” ILC progenitor (CHILP), the latter of which has the capacity to differentiate into all helper ILC subtypes including ILC1, ILC2, and ILC3 (1, 16). Murine CLP (LinThy-1Sca1lowc-kitlowFlt3+IL-7Ra+), CILP (LinIL-7Ra+α4β7+CD25CXCR6+), and CHILP (Id2+IL-7Ra+α4β7+CD25PLZF+/−) are now well characterized [reviewed in (1, 17)]. In contrast, the developmental intermediates of human ILCs are less clearly defined. This gap in knowledge is due, in part, from the study of murine ILC progenitors (ILCPs) using transcription factor reporter mice and the use of receptors (Thy-1 and Sca1) not commonly used to characterize human hematopoietic stem cells (HSCs) and developmental intermediates (16). We demonstrated previously that functionally mature ILC3s could develop in vitro from CD34+ cells in the presence of instructive cytokines including IL-7, IL-15, SCF (stem cell factor), and FLT3L (FMS-like tyrosine kinase 3 ligand) (18). More recently, tonsillar-derived α4β7+CD34+ cells were shown to coexpress c-kit (CD117) and RORγt and gave rise to ILC3 cells (19). Others have observed that CD117 expression on Lin progenitors marks cells that potentially develop into all ILC subpopulations (18, 20). Using a similar approach, Renoux et al. (21) identified a LinCD34+CD38+CD123CD45RA+CD7+CD10+CD127 progenitor that specifically differentiates into the NK lineage and lacks ILC potential.

In addition to the commonly recognized surface antigens used to characterize human HSCs (CD34, CD38, CD117, and CD45), these cells and their downstream progenitors express signaling lymphocyte activation molecule (SLAM) family receptors including CD150, CD48, and 2B4 (CD244) (22). These receptors are located on chromosome 1 (23) and are present in a stage-specific manner such that long-term repopulating HSCs are CD150+ but lack CD48 or CD244. In contrast, multipotent hematopoietic progenitors lack CD150 and CD48, but express CD244. Last, more restricted progenitors express varying combinations of CD244 and CD48 but are CD150 (22). SLAM family member expression has been used to characterize HSCs and progenitors phenotypically, but their function in this context is largely unstudied. Here, we use single-cell RNA sequencing (scRNA-seq) and lineage differentiation assays to define the human CILP, NKP, and ILCP in vitro and in vivo. We provide evidence that NK and ILC progenitors can be better defined using CD48 and CD52 staining, and that 2B4 ligation by CD48 is required for the development of ILC2 lineage cells, at the cost of NK cell development.


Human CILPs are contained in CD34+α4β7+CD117+ HSCs

We previously demonstrated that umbilical cord blood (UCB)–derived CD34+ cells cultured in the presence of IL-3, IL-7, IL-15, SCF, and FLT3L give rise to NK cells and ILC3s over 3 to 4 weeks (18). On the basis of this, we reasoned that CD34+ cells would transit through intermediates that could develop into all ILC subtypes. Using a panel of CD56, CD94, CD117, CD161, CD294, and CD336 and a lineage cocktail (containing CD1a, CD3, CD4, CD5, CD11c, CD14, CD19, CD34, TCRαβ, TCRγδ, FcεRI, CD123, and CD303), we identified Lin ILCs, including NK cells, ILC1, ILC2, and ILC3 (fig. S1, A and B), in these cultures. Consistent with previous data (5, 11, 20, 24, 25), NK cells expressed CD56 and CD94, and ILC2s expressed CD117 and CD294 (fig. S1, A and B). On the basis of previous studies (19, 20), we characterized the ILC3 compartment using CD56, CD117, CD161, and CD336 (fig. S1, A and B). Last, ILC1 cells were defined as lacking CD56, CD94, CD294, and CD336, but partially expressing CD117 and CD161 (fig. S1, A and B). This surface expression pattern was validated using intracellular transcription factor staining, demonstrating Tbet in NK cells, GATA3 in ILC2s, and RORγt in ILC3s (fig. S1C). Further confirmation of cell identity was established by cytokine production, showing IFN-γ by NK cells and ILC1s, IL-13 by ILC2s, and IL-22 by ILC3s (fig. S1D).

We investigated whether CILPs were present in UCB CD34+ and integrin α4β7+ compartments (1, 26, 27). Nearly all UCB-derived CD34+ cells were CD117+, whereas only ~3% of these were α4β7+ (fig. S2, A and B). However, after 5 days of culture in medium containing FLT3L, LDL (low-density lipoprotein), SCF, and TPO (thrombopoietin), which is a cytokine cocktail used to expand HSCs, a higher fraction of purified CD34+ cells acquired α4β7 (fig. S2, B and C) (28). These 5-day expanded CD34+ HSCs were fluorescence-activated cell sorting (FACS)–sorted into CD34+α4β7+ and CD34+α4β7 subsets, followed by culture in cytokines that support ILC and NK development (fig. S2D). Compared with the CD34+α4β7 HSC subset, CD34+α4β7+ cells were more likely to differentiate into CD117+ ILCPs (fig. S2, D and E) (20, 21, 25). The CD34+α4β7+ progenitors were also more likely to develop into CD94+ NK cells as well as CD294+ ILC2s (fig. S2, F and G). Similarly, cells differentiating from the CD34+α4β7+ population expressed RORγt, an ILC3-associated transcription factor (fig. S2H). Thus, the human CILP capable of developing into all ILC subtypes is contained within the CD34+α4β7+ cell fraction.

Expression of SLAM family receptor CD48 on CILPs marks the ILCPs

We aimed to further characterize the ILCPs and identified a subpopulation of CD34+α4β7+Lin cells that coexpress CD48 (Fig. 1A). When compared with CD34+α4β7+LinCD48 cells, the CD34+α4β7+LinCD48+ cells expressed higher mRNA transcripts associated with ILCs including GATA3, ID2, RORγt, and Tbet (Fig. 1, B to E). These two populations were FACS-purified and differentiated with ILC- and NK cell–supporting cytokines. The CD34+α4β7+LinCD48+ cells gave rise to NK cells and ILCs, whereas CD34+α4β7+LinCD48 cells developed into NK cells, but not ILCs (Fig. 1, F to H). Thus, this finding indicates that the CD34+α4β7+LinCD48+ progenitors contain ILCPs.

Fig. 1 Expression of CD48 by CD34+α4β7+ HSC progenitors.

UCB-derived CD34+ cells were expanded for 5 days and FACS-sorted into CD34+α4β7+CD48 and CD34+α4β7+CD48+ subsets, and gene expression analysis was performed using qPCR. Cells were also cultured for 21 days under conditions that favor ILC differentiation. (A) Sorting gating strategy (representative of n = 4 experiments). (B to E) GATA3, ID2, RORγt, and Tbet mRNA expression analysis by qPCR in the CD34+α4β7+CD48+ subset relative to the CD34+α4β7+CD48 subset [n = 3 (B) or 4 (C to E) per group]. (F) Differentiating cells were stained for ILC surface markers at day 21 of culture, and the percentages of CD94+ NK cells, CD294+ ILC2, CD56CD336+ ILC3, and CD56+CD117+ ILC3 derived from CD34+α4β7+CD48+ and CD34+α4β7+CD48 are shown (n = 4 per group). (G and H) Differentiating CD34+α4β7+CD48 and CD34+α4β7+CD48+ subsets were stained for ILC surface markers at day 21 of culture, and the percentages of CD94+ NK cells, CD294+ ILC2, CD56CD336+ ILC3, and CD56+CD117+ ILC3 are shown; values represent the percentages of ILCs in the culture (representative of n = 4 experiments). (B to F) Data are shown as means ± SD, paired t tests, and P values are depicted.

CD34+α4β7+LinCD48+ ILCPs give rise to tissue-associated ILCs in vivo

The development of CD34+α4β7+LinCD48 and CD34+α4β7+LinCD48+ cells was next tested in vivo using humanized NSG (nonobese diabetic scid gamma) mice. Mice engrafted with CD34+α4β7+LinCD48 cells contained human CD45+ cells only in their bone marrow, whereas mice engrafted with CD34+α4β7+LinCD48+ cells showed human CD45+ cells in their bone marrow, spleen, liver, and lung (Fig. 2A). The reconstituting CD45+ cells from both progenitors lacked T cells, B cells, and monocytes (fig. S2I). The bone marrow from both groups of mice contained NK cells, but lacked ILC2 and ILC3 cells (Fig. 2B). The spleen, liver, and lung mainly showed human NK cells that were Tbet+ and IFN-γ+ (Fig. 2, B to D). Mice that received CD34+α4β7+LinCD48+ cells also had GATA3- and IL-13–expressing ILC2s, as well as RORγt- and IL-22–expressing ILC3s within these tissues (Fig. 2, B to D). Together, these results demonstrate that in vivo CD34+α4β7+LinCD48 cells give rise to NK cells, but not ILCs, whereas CD34+α4β7+LinCD48+ cells give rise to tissue-associated NK cells, ILC2s, and ILC3s.

Fig. 2 CD34+α4β7+CD48+ precursors give rise to tissue-associated ILCs in vivo.

UCB-derived CD34+ HSCs were expanded for 5 days, FACS-sorted into CD34+α4β7+CD48+ and CD34+α4β7+CD48 cells, and infused into immunodeficient NSG mice. After 4 weeks, mice were sacrificed and lymphocytes were isolated and analyzed by flow cytometry. (A) Representative dot plot showing Lin human CD45+ cells in murine tissues based on progenitor population (n = 5 per group). (B) Representative plots depicting human CD56 expression within the Lin human CD45+ cells (top row), human CD117+CD94 ILC3s and CD94+ NK cells within the Lin human CD45+CD56+ cells (middle row), and human CD294+ ILC2s within the Lin human CD45+CD56 cells (bottom row) in murine tissues (n = 5 per group). (C and D) Human CD94+ NK cells, CD294+ ILC2s, and CD117+CD94 ILC3s from murine lung were shown for IFN-γ, IL-13, and IL-22 (C) or Tbet, GATA3, and RORγt (D) (n = 5 per group). Values in (A) and (B) represent the percentage of positives.

Expression of CD52 in UCB-derived progenitors distinguishes ILCPs

To identify an additional antigen to further resolve the CD34+α4β7+LinCD48 and CD34+α4β7+LinCD48+ cell populations into various ILCPs, scRNA-seq was performed on differentiating CD34+ cells, and from this, we identified CD52 as a candidate gene differentially expressed by the ILCP fraction. CD52, also known as CAMPATH-1 antigen, is a surface glycoprotein expressed on mature lymphocytes, macrophages, and monocytes, but not HSCs (29, 30). Accordingly, freshly isolated UCB-derived CD34+ HSCs lacked CD52; however, after 5 days of culture, some CD34+α4β7+ cells expressed CD52, concurrent with CD48 acquisition (Fig. 3A). Similar to freshly isolated UCB-derived HSCs, nonmobilized peripheral blood CD34+ HSCs lacked α4β7, CD48, and CD52 expression. In contrast, CD34+α4β7+ ILCPs in secondary lymphoid tissues (tonsil) expressed CD48 and CD52 (fig. S3).

Fig. 3 Expression of CD52 in CILPs identifies NKPs and ILCPs.

UCB-derived CD34+ HSCs were expanded for 5 days, and CD34+α4β7+ progenitors were FACS-sorted into four (CD48CD52, CD48CD52+, CD48+CD52, and CD48+CD52+) hematopoietic progenitor populations and studied. Differentiating CD34+ cells were also used as control data. (A) Gating strategy to sort CD34+α4β7+ into four subsets is shown and is representative of n = 5 independent experiments. (B to E) qPCR and quantitative expression of GATA3 (B), RORγt (C), Tbet (D), and IL-23R (E) in hematopoietic progenitors is shown relative to the CD48CD52 subset [n = 5 (B to D) or 3 (E) per group]. (F) UMAP of differentiating and terminally differentiated ILC data, and values are cluster number. (G) UMAP of CD48CD52 (left) and CD48CD52+ (right) samples after preprocessing and integration to differentiating ILC data. (H) UMAP of CD48+CD52 (left) and CD48+CD52+ (right) samples after preprocessing and integration to differentiating ILC data. (B to E) Data are shown as means ± SD, one-way ANOVA, and P values are shown.

Gating on cultured UCB CD34+α4β7+Lin cells showed four populations based on CD52 and CD48 expression (Fig. 3A). scRNA-seq data from these FACS-sorted subpopulations corroborated the flow cytometry data, showing differential expression of CD52 and CD48 in these four UCB-derived CD34+ HSCs subsets (fig. S4A), and global transcriptome analysis demonstrated distinct expression patterns between these subpopulations (fig. S4B). Furthermore, relative to the CD34+α4β7+CD48CD52 population, both the CD34+α4β7+CD48CD52+ and CD34+α4β7+CD48+CD52+ subsets expressed Tbet and GATA3, while the CD34+α4β7+CD48+CD52 and CD34+α4β7+CD48+CD52+ cells also expressed RORγt (Fig. 3, B to D). Given that RORγt was expressed by both the CD34+α4β7+CD48+CD52 and CD34+α4β7+CD48+CD52+ populations, we tested for IL-23R, as it functionally distinguishes NCR+ and LTi-like ILC3 cell development (12, 31). As shown in Fig. 3E, there was considerably higher expression of IL-23R in the CD34+α4β7+CD48+CD52+ subset relative to the CD34+α4β7+CD48+CD52 cells.

Using scRNA-seq, we were unable to further cluster the gene signatures from the four CD34+α4β7+ cell populations (fig. S4C). Instead, we examined the terminally differentiated progeny of the bulk culture and identified 13 distinct cell clusters (Fig. 3F), and this analysis was used to determine which clusters were present in the CD34+α4β7+ progenitors (Fig. 3, G and H). Assessment of the top 10 up-regulated genes in each cluster from the differentiated sample (Fig. 3F) showed that most expressed genes were not clearly related to NK or ILC cells (table S1). Clusters 3 and 4 showed higher expression of NK-associated transcripts (including PRF1, GZMK, XCL1, XCL2, CCL3, and CCL4), whereas clusters 5 and 6 expressed ILC-related genes (such as KLRB1, KRT81, and TNFRSF18) (fig. S5A), suggesting that clusters 3 and 4 are likely to be NK cells and clusters 5 and 6 are likely to be ILCs. Further analysis of clusters 3 to 6 in the CD34+α4β7+ progenitor subsets (Fig. 4A) showed that cluster 3 is most abundant in the CD34+α4β7+CD48CD52+ population. Conversely, cluster 4 was dominant in the CD34+α4β7+CD48+CD52+ population. Clusters 5 and 6 were equally distributed across all four subsets (Fig. 4B). Comparison of differentially expressed genes within these four clusters also identified distinct genes (table S2). Despite assignment to the same cluster, differences were noted in gene expression across the CD34+α4β7+CD48−/+CD52−/+ subsets (Fig. 4, C to F). Together, these data suggest that clusters 3 and 4 of the CD34+α4β7+ subsets correspond to the NKPs and that clusters 5 and 6 represent ILCPs (9, 24).

Fig. 4 Differential gene expression by CD34+α4β7+ NKPs and ILCPs.

UCB-derived CD34+ HSCs were expanded for 5 days, and CD34+α4β7+ hematopoietic progenitors were FACS-sorted into four (CD48CD52, CD48CD52+, CD48+CD52, and CD48+CD52+) hematopoietic progenitor populations and studied. (A) UMAP of CD48CD52, CD48CD52+, CD48+CD52, and CD48+CD52+ samples showing clusters 3 to 6 after preprocessing and integration to terminally differentiated ILC data. (B) Bar graph shows the percentage of cells in clusters 3 to 6 for each CD48CD52, CD48CD52+, CD48+CD52, and CD48+CD52+ sample. (C to F) Heatmap shows the top 10 differentially expressed genes for CD48CD52, CD48CD52+, CD48+CD52, and CD48+CD52+ samples within clusters 3 (C), 4 (D), 5 (E), and 6 (F). Color bar shows expression on a natural log scale. Genes are ordered by average normalized expression within each sample.

Purified ILCPs differentiate to give rise to distinct and functionally active ILCs

We purified the four CD34+α4β7+ populations based on CD48 and CD52 expression and cultured them under conditions that favored NK and ILC development to validate the gene expression findings. Gene expression data suggested the presence of clusters corresponding with NKPs (clusters 3 and 4) and ILCPs (clusters 5 and 6) in the CD34+α4β7+CD48CD52 subset (Fig. 4A), but these cells differentiated into non-ILC cells (Fig. 5, A and B). Likewise, the CD34+α4β7+CD48CD52+ precursors contained clusters 5 and 6 cells but did not give rise to ILCs and instead differentiated into NK cells and non-ILC cells (Fig. 5, A and B). In contrast, the presence of clusters 3 and 4 within the CD34+α4β7+CD48+CD52 population gave rise to ILC3 cells, but not NK cells (Fig. 5, A and B). Last, as above, the CD34+α4β7+CD48+CD52+ cells gave rise to NK cells and multiple ILC types, including CD56CD117+/− ILC1, CD294+CD117+ ILC2, and CD56+CD94CD117+ ILC3 (Fig. 5, A and B). Single-cell culture of CD34+α4β7+CD48CD52 and CD34+α4β7+CD48+CD52 progenitors on irradiated OP9 stromal cells resulted in lack of cell proliferation. Cells proliferated in ~10% (39 of 384) and ~36% (140 of 384) of the wells seeded with single CD34+α4β7+CD48CD52+ and CD34+α4β7+CD48+CD52+ progenitors, respectively. The CD34+α4β7+CD48CD52+ single-cell culture predominantly (30 of 39) generated Lin+CD117+CD294+ cells, whereas the remaining wells (9 of 39) exclusively generated NK cells (fig. S5B). Previously, CD117+CD294+FcεRI+ mast cells were shown to develop from LinCD34hiα4β7+CD117+ progenitors (32, 33). The CD34+α4β7+CD48+CD52+ progenitors generated either NK cells alone (11.6%), ILC1s alone (13.9%), ILC2s alone (11.6%), ILC3s alone (44.2%), or a combination of these subsets (18.6%) (fig. S5C).

Fig. 5 CD34+α4β7+CD48+/− ILCPs differentiate into functionally mature ILCs.

UCB-derived CD34+ HSCs were expanded for 5 days, and CD34+α4β7+ hematopoietic progenitors were sorted using FACS (as shown in Fig. 3A) after 21 days of culture under conditions that favor ILC differentiation. (A and B) Output of the sorted populations (CD48CD52, CD48CD52+, CD48+CD52, and CD48+CD52+). Cells were stained for NK cell or ILC surface markers: NK cells, ILC1, ILC2, ILC3, and NKp44+ ILC3 are shown in dot plots (A) or bar graphs (B). Results are representative of n = 4 experiments. (C to E) ILCs were sorted and stimulated for 16 hours to assess IFN-γ (C), IL-13 (D), and IL-22 (E) production using qPCR. (F to I) ILCs were sorted to assess Tbet (F), GATA3 (G), RORγt (H), and EOMES (I) expression using qPCR. (C to I) The expression for NK cells and ILCs was relative to that of non-ILCs (n = 3 per group). (B to I) Data are shown as means ± SD, one-way ANOVA (*P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001).

We further tested the progeny of the above populations for functional differences. Activation of NK cells and ILC1 cells shows that both up-regulate IFN-γ mRNA (Fig. 5C). Stimulation of ILC2s and ILC3s led to induction of mRNAs for IL-13 and IL-22, respectively (Fig. 5, D and E). FACS-sorted NK cells and ILCs were assessed for transcription factors and showed preferential expression of Tbet, GATA3, and RORγt by NK cells/ILC1s, ILC2s, and ILC3s, respectively (Fig. 5, F to H). Last, we found that ILC1s up-regulated EOMES (Fig. 5I), corroborating previous reports (34, 35). In summary, the expression pattern of CD48 and CD52 by CD34+α4β7+Lin cells defines individual progenitor populations that give rise to distinct human NK, ILC1, ILC2, and ILC3 cells.

CD34+α4β7+CD48CD52+-derived NK cells are distinct from their CD34+α4β7+CD48+CD52+-derived counterparts

Our findings indicated that both CD34+α4β7+CD48CD52+ and CD34+α4β7+CD48+CD52+ cells can give rise to NK cells, so both subsets were purified and differentiated in vitro for further investigation. There were similarities in CD56 and CD94 expression, but CD34+α4β7+CD48+CD52+-derived NK cells expressed higher levels of CD336 and lower levels of CD244, CD335, and CD337 compared with CD34+α4β7+CD48CD52+-derived NK cells (Fig. 6A). Moreover, CD34+α4β7+CD48CD52+-derived NK cells showed more CD107a-mediated degranulation against K562 target cells (Fig. 6B). Transcriptome analysis showed similar expression of the NK-associated genes including TBX21 (Tbet), NCAM1 (CD56), KLRD1 (CD94), and KLRB1 (CD161) in these subsets (fig. S6A). Despite these similarities, scRNA-seq showed distinct patterns between the two NK subsets (Fig. 6C), and CD34+α4β7+CD48CD52+-derived NK cells expressed higher levels of CCL5, FCGR3A (CD16a), TNFSF14, and TYROBP (DAP12) (Fig. 6, C and D). Conversely, CD34+α4β7+CD48+CD52+-derived NK cells expressed higher CD7, GZMB, human leukocyte antigen (HLA) class I, and KLRC1 (NKG2A) (Fig. 6, C and D). Ingenuity Pathway Analysis (IPA) showed distinct signaling pathway activation between the NK subsets (fig. S6, B and C). Collectively, these results show that the CD34+α4β7+CD48CD52+- and CD34+α4β7+CD48+CD52+-derived NK cells are distinct and may represent CD56dim and CD56bright NK cells, respectively (3638).

Fig. 6 α4β7+CD48+- and α4β7+CD48-derived NK cells are distinct cell types.

UCB-derived CD34+ HSCs were expanded for 5 days, FACS-sorted into CD34+α4β7+CD48+ and CD34+α4β7+CD48 cells, and cultured for 21 days off-stroma. (A) Differentiating cells were stained for CD56, CD94, CD244, CD335, CD336, and CD337 (representative of n = 4 experiments). (B) NK cells were cultured with K562 in 5:1 ratio for CD107a assessment, and the percentage of positives is shown (n = 3 per group). (C) scRNA-seq and heatmap showing the top 50 differentially expressed genes between CD34+α4β7+CD48+- and CD34+α4β7+CD48-derived NK cells. Color bar shows expression on a natural log scale. Genes are ordered by average normalized expression with respect to the higher expressing sample. (D) Violin plots derived from scRNA-seq data showing the distributions of CD7, CCL5, FCGR3A, GZMB, HLA-A, KLRC1, TNFSF14, and TYROBP in CD34+α4β7+CD48+- and CD34+α4β7+CD48-derived NK cells. (B) Data are shown as means ± SD, paired t tests, and P values are depicted.

Separate ILC2 and ILC3 subsets differentiate from CD34+α4β7+CD48+ ILCPs

We also sought to identify phenotypically and functionally distinct ILC2 and ILC3 cell populations that can differentiate from CD34+α4β7+CD48+ ILCPs. We used CD11a, CD117, and CD294 (11) to identify two subsets (CD11alowCD117high and CD11ahighCD117low) of CD34+α4β7+CD48+CD52+-derived CD294+ ILC2s (Fig. 7A). Stimulation of the CD11ahighCD117lowCD294+ ILC2 population was associated with IL-13 and IL-10 expression (Fig. 7, B and C). Compared with other ILCs or NK cells, both ILC2 subsets up-regulated TGFB1 after stimulation (Fig. 7D). TGF-β induced more IL-10 expression in the CD11alowCD117highCD294+ ILC2s (Fig. 7E).

Fig. 7 Peculiar ILC2 and ILC3 subsets generated in vitro.

UCB-derived CD34+α4β7+ progenitors were FACS-sorted into four subsets (Fig. 3A) and differentiated for 21 days off-stroma, and ILC2s or ILC3s were analyzed. (A) CD294+ ILC2s were displayed for CD11a and CD117 (representative of n = 4 experiments). Values are the percentage of positives. (B to D) CD294+CD117high and CD294+CD117low ILC2s were stimulated with IL-25 + IL-33, and IL-13 (B), IL-10 (C), and TGFB1 (D) were analyzed using qPCR (n = 3 per group). ns, not significant. (E) CD294+CD117high and CD294+CD117low ILC2s were stimulated with TGF-β, and IL-10 mRNA was depicted as fold change compared with unstimulated cells (n = 3 per group). (F to J) Surface staining for CD336 (F) and NRP1 (J) as well as intracellular staining for IL-22 (G), RORγt (H), and LTA (I) in stimulated CD34+α4β7+CD48+CD52+- and CD34+α4β7+CD48+CD52-derived ILC3. Representative histograms and mean fluorescence intensity (MFI) in bar graphs (n = 4 per group). (K) scRNA-seq and heatmap showing the top 50 differentially expressed genes in cluster 5 versus cluster 6 (Fig. 3F) of ILC3s. Genes are ordered by average normalized expression with respect to the higher expressing sample. (L and M) Violin plots showing the distributions of PRF1, GZMB, CCL5, LTA, LTB, and RORC in cluster 5 versus cluster 6 (Fig. 3F) of ILC3s. P values are displayed. (B to J) Means ± SD, paired t tests (**P < 0.01; ***P < 0.001; ****P < 0.0001).

CD34+α4β7+CD48+CD52-derived ILC3s showed less CD336 expression compared with CD34+α4β7+CD48+CD52+-derived ILC3s (Fig. 7F). Previous reports indicate that LTi-like ILC3 cells express higher IL-22, RORγt, LTA, and NRP1 compared with the NCR+ ILC3 (8, 12) and these were higher on CD34+α4β7+CD48+CD52-derived ILC3s (Fig. 7, G to J). As above, transcriptome analysis of the progeny of CD34+-differentiated cells (as in Fig. 3F) resulted in two distinct ILC3 subsets (clusters 5 and 6). Comparison of the top 50 up-regulated genes within these two mature ILC3 subsets showed that cluster 5 expressed relatively more NK-associated genes including granzymes, PRF1, CCL3, and CCL5 (Fig. 7, K and L). Concomitantly, previous reports demonstrate that NCR+ ILC3 cells express NK cell–associated receptors in addition to ILC3-restricted genes (8, 12). In contrast, cluster 6 up-regulated most of the ILC3-associated genes including KIT, KLRB1, LTA, LTB, and RORC (Figs. 7M and 7K). Moreover, IPA indicated activation of distinct signaling pathways between cluster 5 and 6 ILC3s (fig. S7, A and B). Together, these data provide evidence that the CD34+α4β7+CD48+CD52-derived ILC3s are LTi-like ILC3s, whereas the CD34+α4β7+CD48+CD52+-derived ILC3s represent NCR+ ILC3s.

2B4 receptor signaling enhances the development of ILC2 from ILCPs

The 2B4 (CD244) receptor modulates NK and T cell activation through CD48 binding (39, 40). Consistent with previous reports (41), nearly all freshly isolated UCB-derived CD34+ HSCs express CD244, whereas only a proportion show expression of CD48 (Fig. 8A). CD244 surface density on freshly isolated CD34+CD48 HSCs is similar to CD34+CD48+ HSCs, but expression was up-regulated on the latter progenitors after 3 days of in vitro culture (Fig. 8B). We hypothesized that CD244 signaling may influence ILC differentiation from CD34+α4β7+CD48+ cells. SLAM-associated protein (SAP), known as an adapter protein for CD244 signaling (23), was up-regulated within CD34+α4β7+CD48+ cell population, suggesting activation of CD244 signaling (Fig. 8C). During development, the addition of anti-CD244 or anti-CD48 blocking antibody abrogated ILC2 differentiation, whereas the proportion and absolute number of NK cells were increased (Fig. 8, D and E, and fig. S8, A and B). Conversely, activation of CD244 signaling using an agonist antibody increased ILC2 differentiation at the expense of NK cells (Fig. 8, F and G, and fig. S8, A and B). Moreover, culturing CD34+α4β7+CD48+ progenitors on a layer of irradiated CD48-expressing OP9 cells (versus control OP9 stromal cells) showed an enhanced ILC2 generation (fig. S8C). We used CRISPR-Cas9 to knock out CD244 in CD34+α4β7+CD48+ progenitors followed by single-cell culture on irradiated OP9 feeders to confirm this observation. CD244 expression was lost in 92% of the single-cell cultures, whereas 94% of the single-cell cultures containing control guide RNA (gRNA) expressed CD244 (Fig. 8H and fig. S8D). Notably, ILC2 development was completely lost in the absence of CD244 (Fig. 8, I and J). In contrast, NK cells and ILC3s differentiated from CD244 knockout progenitors (fig. S8, E and F). We identified two separate ILC2 populations once again based on CD11a and CD117 expression. Both ILC2 subsets increased proportionally with CD244 activation, suggesting that CD244 signaling acts upstream of ILC2 subtype specification. These data demonstrate that CD244 signaling influences ILC differentiation by modulating the development of CD34+α4β7+CD48+ progenitors into ILC2 cells.

Fig. 8 CD244 activation enhances ILC2 development.

(A) Freshly isolated UCB-derived CD34+ HSCs were stained for CD244 and CD48 (representative of n = 4 experiments). (B) CD244 in CD34+CD48 cells (aqua) and in CD34+CD48+ cells (red) for day 1 and 3 UCB-derived CD34+ HSCs (representative of n = 4 experiments). (C) SAP mRNA expression in day 5 CD34+CD48 and CD34+CD48+ cells (n = 4 per group). (D and E) Anti-CD244 and anti-CD48 blocking antibodies or isotype IgG controls were added to differentiating CD34+α4β7+CD48+ cultures to block CD244 signaling. (F and G) CD34+α4β7+CD48+ cells were plated on anti-CD244– or IgG-coated plates twice for 4 days before 21 days of differentiation to activate CD244 signaling via cross-linking. Cells were stained for ILCs, and the percentage (D and F) or absolute number (E and G) of CD94+ NK cells, CD294+ ILC2, and CD117+ ILC3 is shown (n = 4 per group). (H) CD244+ cultures are shown for cells transfected with control or CD244 gRNA (n = 3 per group). (I) Representative ILC2 (CD294) staining for cultures transfected with control or CD244 gRNA. KO, knockout; WT, wild type. (J) Dot plot showing generation of ILC2s from cultures that express (control gRNA) or lack (CD244 gRNA) CD244. (C to H) Means ± SD, paired t tests (C, F, G, and H) or one-way ANOVA (D and E), and P values are depicted.


The exact developmental trajectory that human HSCs take to become ILCs is not fully defined, and human common ILCPs have not yet been characterized (17). To decipher ILC development in mice, transcription factor reporter and knockout animals have been used (16). Given that this approach is not possible in humans, we used scRNA-seq, immunophenotyping, in vivo adoptive transfer, and functional analysis to investigate the developmental trajectories of human HSCs. We describe developmental intermediates with progressively narrowing capacities to give rise to various NK and ILC subsets. All human ILCs are believed to arise from CILPs that are downstream of CLPs (20, 42). We found that human CILPs enriched within the Linα4β7+ compartment of CD34+ progenitors. Although these cells are rare in freshly isolated UCB, they expand with cell-specific cytokines. CD52 is expressed by mature lymphocytes, and CD34+ HSCs lack this protein (29, 30, 43). However, after several days in culture, a portion of CD34+ progenitors express CD52, which is concurrent with α4β7 and CD48 acquisition. CD34+ cells that express α4β7, CD48, and CD52 were present in tonsillar tissue, but were not found in UCB or peripheral blood, suggesting the requirement of secondary lymphoid tissues for the differentiation of these progenitors (42). In murine models, CILP gave rise to CHILPs and NKPs, with the former developing exclusively into ILCs and not NKs, and the latter showing the opposite characteristics (1, 16). In contrast, studies using freshly isolated human tonsils provide evidence that these same intermediates are not necessarily occurring in humans (44, 45). Here, we show that human CILPs are further defined by CD48 and CD52 expression and can be delineated into four progenitor populations with unique gene expression and developmental potential. The data here show that CD34+α4β7+LinCD48CD52+ cells express Tbet and give rise to an NK population that has cytotoxic features and most closely resembles CD56dim NK cells. We also found that the cells derived from CD34+α4β7+LinCD48+CD52 cells are TbetRORγt+ and give rise to LTi-like ILC3 cells. In this manner, these populations appear to be restricted progenitors. However, the CD34+α4β7+LinCD48+CD52+ fraction expressing Tbet, GATA3, and RORγt develops into a few different cell types, including CD56bright NK cells, ILC1, ILC2, and ILC3 cells. We have made an exhaustive attempt to further resolve this fraction into progenitors destined to become either NK cells or ILCs. Despite differential expression of multiple candidate molecules (AMICA1, CCR7, CD44, CD53, CD63, CD99, HLA-A, IL2RG, KRT1, and NOTCH1/2,) at a transcript level, we were unable to identify these and others as suitable antigens by using flow cytometry. Although our study does not quite formally define NKPs and ILCPs as “restricted” cell types, the finding that two NK cell subsets develop from distinct progenitors is consistent with the work of others (44, 45).

Data obtained using scRNA-seq analysis indicated that there were distinct cell subpopulations with gene clusters that resembled mature NK cells and/or ILCs. These clusters could be found within each of the four progenitor populations (based on CD48 and CD52), but only those derived from the CD34+α4β7+LinCD48+CD52+ fraction gave rise to the ILC1, ILC2, and ILC3 cells. In this manner, these results leave open the possibility of a human CHILP that gives rise to only ILC1, ILC2, or ILC3 cells, as described in mice (1, 16). Similarly, the developmental potential of the CD34+α4β7+LinCD48−/+ cells was determined in vivo using humanized mice, showing that CD34+α4β7+LinCD48 cells engraft only in the bone marrow and give rise to NK cells, whereas CD34+α4β7+LinCD48+ cells give rise to tissue-associated NK cells, ILC1s, ILC2s, and ILC3s in the spleen, liver, and lung. Both CD34+α4β7+LinCD48 and CD34+α4β7+LinCD48+ cells do not develop into T cells, B cells, or monocytes in vivo in the immunodeficient mice.

Human NK cells are one of the best characterized subsets of innate lymphocytes and two subsets of NK cells (CD56bright and CD56dim cells) have been identified, but their developmental relationship is still poorly understood (3638, 44, 45). Consistent with previous studies (4447), we found two NK cell populations with similar expression of common NK cell surface markers such as CD56 and CD94, but variations in NKp30, NKp44, NKp46, and 2B4 expression. Both NK cell populations produced IFN-γ, but differed in their propensity for degranulation in response to K562 targets, reminiscent of CD56bright and CD56dim populations. The in vitro developmental system used allowed us to identify that these NK cell subpopulations are derived from two separate progenitors, with the resulting cells having differing transcriptomes, likely due to their distinct developmental origins (2, 3, 48). Likewise, these two NK subpopulation may also represent the already described circulating and tissue-associated NK cells (44, 49, 50), as shown by our in vivo data that CD34+α4β7+CD48+CD52+ progenitors, but not CD34+α4β7+CD48CD52+ progenitors, give rise to NK cells in the tissues such as liver and lung. Comparison of ILC3s derived from the CD34+α4β7+LinCD48+CD52+ and CD34+α4β7+LinCD48+CD52 progenitors showed variations in overall transcriptomic expression and function. Such differences in human ILC3 subsets exist between the NCR+ ILC3 and the LTi-like ILC3 (12, 31), but the reasons for these are largely unknown. Unlike CD34+α4β7+LinCD48+CD52 ILC3 precursors, we find the CD34+α4β7+LinCD48+CD52+ ILC3 precursors to express both Tbet and GATA3, suggesting that these may modulate RORγt, as evidenced by less RORγt expression and IL-22 production in the CD34+α4β7+LinCD48+CD52+-derived ILC3 cells. Previous reports indicate that GATA3 antagonizes RORγt and enhances Tbet during NCR+ ILC3 development (51, 52). NCR+ ILC3 development also depends upon IL-23 signaling, leading to Tbet up-regulation, which inhibits RORγt expression (31). The developmentally programed expression of IL-23R in CD34+α4β7+LinCD48+CD52+ ILC3 precursors, but not in CD34+α4β7+LinCD48+CD52 ILC3 precursors, agrees with this report. Our findings suggest that CD34+α4β7+LinCD48+CD52+- and CD34+α4β7+LinCD48+CD52-derived ILC3s are distinct subsets of ILC3s. Last, the expression of GATA3 and the production of IL-13 by ILC3s, both observed here, likely have implications for ILC3 plasticity (13, 14).

The SLAM family receptors distinguish HSCs and hematopoietic progenitors (22). We show that the SLAM family receptor CD48 not only is a surface marker for human ILCPs but also participates in CD244 signaling that, in turn, influences ILC development. All HSC progenitors expressed CD244, with relatively higher expression in the CD34+α4β7+LinCD48+ fraction. CD48 is the primary ligand for CD244 and is present on murine hematopoietic cells, with the exception of quiescent long-term HSCs (40, 53). The HSCs from CD48 knockout mice display low engraftment and differentiation potential (53). CD244 signaling modulates the proliferation and activation of mature lymphocytes, such as T cells and NK cells (39, 40). We reasoned that differential expression of CD48 by CD34+α4β7+LinCD48+ progenitors might indicate the necessity of CD244 signaling for their differentiation. Confirming this hypothesis, progenitors lacking CD244 showed a complete absence of ILC2s but the generation of NK cells was unchanged, which is consistent with a previous finding that NK cell development is intact in CD244 knockout mice (54). Conversely, augmented CD244 signaling enhanced ILC2 development, but CD244 blockade prevented ILC2 differentiation and increased the generation of NK cells. These findings suggest that CD244 signaling induces ILC2 development from ILCPs and that ILC2 cells attenuate NK cell development or expansion, perhaps involving inhibitory factors such as TGFB1 and IL-10. In support of this concept, CD244 signaling led to the development of two ILC2 subsets, one of which expressed IL-10 upon stimulation with TGF-β. Recent murine studies report “regulatory” ILCs that influence the function of other ILCs by IL-10 production (9, 10, 15).

Like all research, there are limitations to this work. We have largely used an in vitro differentiation system to obtain sufficient cells to perform experiments including scRNA-seq at various stages of development. Whether these conditions sufficiently mirror the events that occur in vivo could be questioned. However, we were able to confirm the presence of CD48 and CD52 on progenitors in tonsillar tissue, but not in blood or UCB, pointing the likely importance of the local tissue microenvironment in cell differentiation and the associated challenges in mimicking this. Moreover, we were not able to further resolve the CD34+α4β7+LinCD48+CD52+ that gives rise to multiple ILC types (ILC1, ILC2, and ILC3 cells) and NK cells. However, the presented scRNA-seq and single-cell culture data would support the existence of common ILCPs independent of additional progenitor populations with restricted differentiation potential.

In summary, CD48 and CD52 expression by CD34+α4β7+ HSCs marks human ILCPs and NKPs. Using scRNA-seq, we characterize the NK cell and ILC progenitors and demonstrate that various subpopulations of these cells have distinct developmental origins. CD244 triggering favors ILC2 development at the cost of NK cell differentiation. Together, these findings define the developmental trajectories of specific human ILC subtypes and suggest that these subtypes are developmentally programmed and are not only a consequence of alternative activation states or ILC plasticity.


Study design

This study was designed to investigate the developmental intermediates of human ILCs as they grow from HSCs. De-identified human peripheral blood, UCB, and tonsil tissues, all of which were deemed exempt by University of Colorado Institutional Review Board (COMIRB), were used to perform this work. Animal experiments were approved by the ethical committee at University of Colorado Anschutz, Institutional Animal Care and Use Committee (IACUC) protocol (00251).

Isolation and expansion of CD34+ HSCs

Mononuclear cells were isolated by density gradient centrifugation using Lymphoprep (STEMCELL). UCB-derived CD34+ HSCs were positively enriched using a MACS CD34+ enrichment kit (Miltenyi). The cells (purity, >95%) were suspended (5 × 104 cells/ml) in StemSpan II cell culture medium (STEMCELL) supplemented with 1% penicillin + streptomycin, SCF (100 ng/ml, R&D Systems), FLT3L (100 ng/ml, STEMCELL), TPO (50 ng/ml, R&D Systems), and LDL (10 μg/ml, STEMCELL) and cultured in 24-well plates for 5 days of expansion. After 5 days of expansion, the cells were expanded threefold, on average, whereas the proportion of CD34+ cells remained >95%.

Differentiation of CD34+ HSCs

After 5 days of culture, the expanded cells were considered for further differentiation experiments. Where specified, expanded CD34+ HSCs were FACS-sorted into different subsets including CD34+α4β7+, CD34+α4β7, CD34+α4β7+CD48+/−, and CD34+α4β7+CD48+CD52+/−. For up to 28 days of differentiation, cells were cultured in B0 differentiation medium [as previously described (25)], supplemented with SCF (20 ng/ml, R&D Systems), IL-3 (5 ng/ml, STEMCELL), IL-7 (20 ng/ml, R&D Systems), IL-15 [10 ng/ml, National Institutes of Health (NIH)], IL-23 (10 ng/ml, R&D Systems), and FLT3L (10 ng/ml, STEMCELL). In some experiments (specified in figure legends), cells were also cultured in the presence or absence of preplated and irradiated EL08.1D2 stromal cells. After a week of culture, IL-3 was excluded from the B0 differentiation medium supplements. For plating progenitors on the stromal cells, 100 progenitor cells were plated per well of 96-well plates on the irradiated EL08.1D2 cells in 150 μl of B0 differentiation medium. Alternatively, cells were also plated without stroma using 96-well U-bottom plate and 1 × 103 cells were cultured per well. Culturing, maintenance, and preparation of irradiated stromal layer of EL08.1D2 cells on 96-well plate culture were as described earlier (55).

In vivo generation of ILCs in NSG immunodeficient mice model

CD34+α4β7+CD48 and CD34+α4β7+CD48+ progenitors were FACS-sorted from day 5 expanded CD34+ HSCs, as shown in Fig. 1A. CD34+α4β7+CD48 or CD34+α4β7+CD48+ progenitors (5 × 105) were injected via tail vein into 10 (n = 5 per group) sublethally irradiated (3 Gy) 4-week-old mice. Mice were intraperitoneally injected with IL-2, IL-7, IL-15, SCF, and IL-23 (300 ng each) four times in weekly bases and sacrificed for analysis 4 weeks after infusion. Isolation of lymphocytes from mice organs was as previously described (20).

CD244 cross-linking, blocking, and knockout

To study potential stimulatory effects of CD244 activation in CD244-expressing CD48+ progenitors, anti-CD244 antibody clone C1.7 (Thermo Fisher Scientific) was used to initiate cross-linking. For this purpose, anti-CD244 or isotype immunoglobulin G (IgG) antibody was coated as 2 μg/ml in phosphate-buffered saline (PBS) on flat-bottom 96-well culture plates for 2 hours at room temperature. After blocking by 5% fetal bovine serum–containing culture medium and three cycles of washing with PBS, cells were plated using B0 differentiation medium on the coated plates to facilitate CD244 cross-linking. After 48 hours of culture, cells were collected and transferred to a newly coated plate. Cells were finally collected after a total of 96 hours of cross-linking and plated for further differentiation in 96-well U-bottom cell culture plate off-stroma. Alternatively, to activate CD244 by its natural ligand, human CD48-expressing OP9 stromal cells were generated using lentivirus transduction. The progenitor cells (100 cells) were plated per well of 96-well plates on the layer of irradiated CD48-expressing OP9 cells in 150 μl of B0 differentiation medium. To further investigate the stimulatory effects of CD244 signaling, both CD244 and its primary ligand, CD48, were blocked in differentiation cultures using neutralizing antibodies (5 μg/ml) against CD244 (clone eBioPP35, eBioscience) and CD48 (clone eBio156-4H9, eBioscience). Moreover, CD244 was deleted from CD34+α4β7+CD48+ progenitors using the CRISPR-Cas9 system. For this purpose, 1 × 106 cells were electroporated in Amaxa 4D-Nucleofector system (Lonza) with a complex of Cas9 enzyme, tracrRNA, and custom CD244 or control gRNA. The Alt-R Cas9 Nuclease V3, universal tracrRNA, CD244 gRNA, and nonhuman control gRNA were purchased from Integrated DNA Technologies. The electroporated cells were plated as 1 cell per well of 96-well plates on the layer of irradiated OP9 cells in 150 μl of B0 differentiation medium for further differentiation.

scRNA-seq sample processing

UCB-derived CD34+ HSCs from two donors were expanded separately for 5 days as described above, and CD34+α4β7+ cells were sorted into four populations: CD48CD52, CD48CD52+, CD48+CD52, and CD48+CD52+ (as in Fig. 3A) for transcriptome analysis. Days 1, 5, 9, 13, 18, and 23 differentiating CD34+ cells from three donors were also analyzed and used as a reference sample that contains the NKP, ILCP, and their progeny. Terminally differentiated CD56+CD94+ NK cells from two donors were also analyzed to compare CD48- and CD48+-derived NK cells. A total of 15,827 cells from seven donors (table S3 for metrics) were captured using the Chromium Single Cell 3′ Solution (10x Genomics). The resulting libraries were then sequenced as paired-end 151–base pair reads on the Illumina NovaSeq 6000 platform at the University of Colorado’s Genomics Core Facility. Read mapping and quantification was performed using Cell Ranger (10x Genomics).

scRNA-seq analysis

Expression data for CD34+α4β7+ cells (n = 3203) and differentiating CD34+ cells (n = 10,959) with more than 250 unique genes were imported into an R environment; the data were filtered and analyzed using the R package Seurat (56). A total of 31,443 genes were retained after excluding genes expressed in less than one cell. CD34+α4β7+ cells (3181) and differentiating CD34+ cells (10,890) remained for further analysis after excluding cells based on the following criteria: Unique molecular identifiers (UMI) counts greater than 60,000, mitochondrial percentages greater than 12%, or a UMI to unique gene count ratio less than 1 or greater than 8. This filtering criterion was chosen for the purpose of removing doublets or cells that lysed prematurely during library preparation. Gene counts were then normalized for each cell by dividing the total number of counts within each cell, multiplying by a scaling factor of 10,000, adding one, and then taking the natural log. Samples were then integrated, using the functions provided by Seurat, to improve downstream analysis using UMAP (uniform manifold approximation and projection). Cell cycle state was determined for each cell, using Seurat’s CellCycleScoring function, and subsequently regressed out. Cells were then projected onto the top 30 principal components derived from principal components analysis (PCA). The dimensionality was then further reduced to two dimensions using UMAP. Thirteen clusters were demarcated in the UMAP space by a shared nearest neighbor modularity optimization-based clustering algorithm. A similar protocol was used to analyze the CD48- and CD48+-derived NK cells. Data for 932 CD48+- and 721 CD48-derived NK cells with 20,465 genes were present after importing the data into R. A total of 923 CD48+-derived NK cells and 712 CD48-derived NK cells remained after excluding cells based on the following criteria: unique gene counts greater than 7500, UMI counts greater than 60,000, mitochondrial percentages greater than 20%, or a UMI to unique gene count ratio less than 1 or greater than 9. Initial dimensionality reduction was performed using 23 principal components before UMAP.

Quantitative polymerase chain reaction

To perform quantitative polymerase chain reaction (qPCR), total RNAs were extracted from cells using RNeasy kit (Qiagen) and reverse-transcribed into complementary DNAs (cDNAs) using the iScript Advanced cDNA Synthesis Kit (Bio-Rad). The qPCR and analyses were performed as described before (27). The details of the primer assays used for this study are depicted in table S4.

Flow cytometry

Flow cytometry was used to analyze α4β7+ ILC progenitors, CD117+ ILC precursors, as well as mature ILCs. The gating strategy for ILCs was as shown (fig. S1). To evaluate the intracellular IL-13, IL-22, and IFN-γ expressions in ILCs, cells were stimulated with 10 ng/ml phorbol 12-myristate 13-acetate (PMA) (Sigma-Aldrich) + ionomycin (1 μg/ml) (Sigma-Aldrich) (10 ng/ml) or IL-12 + IL-18 (for ILC1, R&D Systems), IL-25 + IL-33 (for ILC2, Shenandoah), and IL-1β (STEMCELL) + IL-23 (R&D Systems) (for ILC3) (10 ng/ml) overnight and in the presence of brefeldin A (2 μg/ml) (BD Biosciences) for the last 4 hours. For intracellular staining, cells were first stained for surface markers and fixed, followed by permeabilization and staining for intracellular proteins. For CD107a degranulation assay, sorted NK cells were incubated with K562 cells at effector:target of 5:1, and the detail of the assay is previously described (57). All flow cytometry data were acquired in LSR II (BD Biosciences) and analyzed using FlowJo (BD Biosciences) or Kaluza (Beckman Coulter) analysis software. As negative controls, fluorochrome-conjugated isotype-matched antibodies from the respective companies were used. Viability of cells was analyzed using flow cytometry with the fixable viability dye eFluor 780 (eBioscience). The detail of the antibodies used for surface or intracellular protein staining is as mentioned in table S5.

Statistical methods

One-way analysis of variance (ANOVA) and paired t tests were used to compare differences between categorical values. For scRNA-seq analysis, Wilcox test was applied on any gene that had an absolute natural log fold change of at least 0.25 and was expressed in more than 20% of cells in at least one of the groups being compared. A P value of 0.05 was used as a cutoff before selecting differentially expressed genes shown in figures. IPA was performed using significantly differentially expressed genes. Pathways were considered significant at P ≤ 0.05.


Fig. S1. Representative ILC staining strategy.

Fig. S2. Generation of ILCs from CD34+α4β7+ HSC progenitors.

Fig. S3. Surface expression of CD48, CD52, CD117, and α4β7 by peripheral blood– and tonsil-derived CD34+ HSCs.

Fig. S4. Gene expression profiling in four subsets of hematopoietic progenitors.

Fig. S5. RNA-seq and single-cell culture demonstrated in vitro differentiation of NK cells and ILCs from CD34+ HSCs.

Fig. S6. Similar expression of NK cell–associated markers in CD34+α4β7+CD48+- and CD34+α4β7+CD48-derived NK cells.

Fig. S7. Differentially activated pathways in two distinct ILC3 subsets.

Fig. S8. The effect of CD244 signaling in the development of ILCs.

Table S1. Top 10 differentially expressed genes for each cluster within the terminally differentiated CD34+ HSC-derived ILCs.

Table S2. Table showing differentially expressed genes for clusters 3 to 6 that are common to both terminally differentiated cells and CD34+α4β7+ hematopoietic progenitors.

Table S3. Description of samples used for scRNA-seq analysis.

Table S4. Lists and description of primer assays used in the study.

Table S5. Lists and description of antibodies used in the study.

Data file S1. Raw data.


Acknowledgments: We thank the clinical immunology cell sorting facility, the genomics and microarray core, and the cancer center shared resource cores (P30CA046934) (University of Colorado) for technical assistance. The α4β7 antibody is a donation from the NIH. Funding: This work was supported by a grant from the NIH (1R01AI100879). Author contributions: D.M.T., M.R.V., and K.J. designed the study. D.M.T., A.M.Y., T.S., D.J., S.S., J.L., K.W., L.C., and R.W. performed experiments and edited the manuscript. D.M.T., G.D.T., and K.J. performed statistical analysis. D.M.T., M.R.V., K.J., and G.D.T. wrote the manuscript. M.R.V. directed the study. Competing interests: D.M.T. and M.R.V. are inventors on patent application (application no. 63/084,440) held/submitted by the regents of the University of Colorado, a body corporate that covers compositions and methods for producing and using ILCs to treat health conditions. All other authors declare that they have no competing interests. Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Materials. The scRNA-seq datasets were deposited in the GSE accession: GSE160009.
View Abstract

Stay Connected to Science Immunology

Navigate This Article