Conventional NK cells develop from HSCs in the bone marrow, through a sequential order of intermediate progenitors. The first progenitor to arise from HSCs is the lymphoid-primed multipotent progenitor, which then gives rise to the common lymphoid progenitor (CLP) (18). The earliest NK lineage committed progenitor to arise from CLPs is known as pre-pro NK (19), which was subdivided into pre-pro A and pre-pro B (19, 20). Differing only in c-kit (CD117) expression, the relationship between pre-pro A and B remains unclear and requires further investigation. Pre-pro NK cells then differentiate into the NK progenitor (NKP) (19, 21). NKPs give rise to immature NK (iNK) cells that either undergo further development within the bone marrow (22) or enter the periphery and develop into mature NK cells (23, 24).

As the early stages of murine NK development are still poorly defined, the developmental pathway outlined above is by no means the definitive model. Heterogeneity within existing progenitor populations, along with the discovery of new distinguishing cell markers, have led to the identification of new sub-populations and, therefore, refinements to the developmental pathway of NK cells. For instance, the common innate lymphoid progenitor (CILP) was found to possess the capacity to give rise to all lineages within the ILC family, of which NK cells are the founding member, but not B and T cells, thus making it an earlier progenitor than the pre-pro NK (25). As the CILP is expresses α₄β₇, it is also alternatively referred to as the α-lymphoid precursor (αLP) (25). However, it has been postulated that there could be an even earlier progenitor population, as there were only about 50 CILPs per mice, and only 2.5% of CILPs efficiently developed into all ILC lineages (25).

The early innate lymphoid progenitor (EILP) is proposed to be the earliest known progenitor for ILCs and was identified using a TCF1 (T-cell factor 1, encoded by Tcf7) transgenic mouse strain that expresses a green fluorescent protein reporter (26). Like CILPs, EILPs give rise to all ILC lineages both in vivo and in vitro, albeit more efficiently (26). Nonetheless, EILPs have not been shown to differentiate into CILPs and hence, the relationship between these two progenitors remains unresolved. As most of the surface markers used to identify the EILP and CILP were different, a detailed comparison of the surface marker phenotype between the two progenitors may also provide further insight.

The discovery of pre–pro NKPs was the outcome of efforts to better understand why only 8–40% of NKPs had solely NK cell potential (19). Similar studies resulted in the identification of a pre-NKP population that preceded a “streamlined” population of NKPs known as refined-NKP (rNKP) (27). Even though many parallels have been drawn between the pre-NKP and pre-pro NK, likewise with NKP and rNKP, it remains to be determined if these populations are exactly identical (20). A summary of the surface markers that are expressed on the various progenitors are provided in Table 1.

Natural killer cell maturation is a process by which lineage committed NK cells acquire their full effector functions. This process is also accompanied by the expression of different cell surface markers, which have helped in the identification of different NK cell maturation subsets. At present, most studies use CD11b and CD27 to divide cNK cells into three maturation subsets: immature (Imm), mature 1 (M1), and mature 2 (M2). Low (lo) and high (hi) CD11b expression divides cNK cells into immature and mature subsets, respectively (22). Heterogeneity in CD27 expression further delineates the mature NK compartment into CD27^hi and CD27^lo subsets, which has also been referred to as M1 and M2 NK subsets (24, 28). CD27 and CD11b expressing M1 NK cells have also been termed double-positive NK cells (24). The three subsets differ in proliferative and cytotoxic capacity. In general, NK cells lose proliferative potential and produce less cytokine, but become more cytotoxic against target cells as they mature (22, 24, 28).

Apart from CD27 and CD11b, markers such as KLRG1 (23, 29), CD62L (30), MCAM (31), CD49b (21), CD43 (32, 33), Ly6C (34), DNAM1 (35), and CD160 (36) have further dissected maturing NK cells into various phenotypic subsets (37, 38). NK cells that express CD160 exhibit enhanced IFN-γ production (36), while the opposite is true for mature NK cells that express higher levels of Ly6C (34), and KLRG1 (39). Like Ly6C, MCAM is also more highly expressed on mature NK cells, although MCAM⁺ and MCAM⁻ NK cells differ in their ability to kill target cells rather than cytokine production (31). Interestingly, the expression of DNAM1 appears to be independent of NK cell maturation that is defined by CD27 and CD11b, as DNAM-1⁺ and DNAM-1⁻ NK cells were observed in both the immature and mature NK compartments (35). As the correlation between these markers have not been studied in detail, the relationships between these phenotypic subsets remain unclear and warrants further investigation. A summary of the surface markers that are expressed on the various mature NK cell subsets are provided in Table 1.

The TFs that are involved in regulating NK cell lineage specification include ETS proto-oncogene 1 (ETS1), NFIL3, and TCF1. ETS1 is a key regulator of early NK cell development as ETS1-deficient mice have normal CLP numbers but lack NK cells (45). Further investigation into the impact of ETS1 deficiency on NK lineage specification revealed a reduction in pre-pro NK, pre-NKP and rNKP, and mature NK cell numbers, thereby supporting the role of ETS1 in NK cell lineage specification (46). ETS1 is believed to impact early NK cell development by regulating the expression of ID2 and TBET, which are also important TFs for NK cell development (46).

TCF1 and NFIL3 are also key regulators of NK cell lineage specification, as marked reductions in pre-pro NK, pre-NKP, and rNKP numbers were observed within the bone marrow of NFIL3- and TCF1-deficient mice, although only the former mouse strain exhibited an additional reduction in CILP numbers (25, 26, 47, 48). NFIL3 appears to be dispensable for mature NK cell as their numbers remained unchanged after its deletion in mature NKp46⁺ NK cells (49). Despite reduced progenitor and mature NK cell numbers in the bone marrow of TCF1-deficient mice, mature NK cells have been found in the periphery at frequencies comparable to littermate controls (26, 50). As most NK cells are derived from bone marrow precursors at steady state, an investigation into the source of NK cells in TCF1-deficient mice might perhaps shed some light on alternative pathways of NK development. The mechanism by which TCF1 mediates NK lineage specification remains poorly understood. However, TCF1 has been shown to regulate T-lineage specification by promoting the expression of genes, such as Gata3, Bcl11b, Il2ra, and Cd3e, that are critical for T cell development (51). Similarly, the mechanism by which NFIL3 specifically mediates lineage specification in NK cells remains unclear, although NFIL3 was found to promote the expression of ID2 and EOMES (47), transcription factors known to be involved in the later stages of NK cell development (52, 53). The role of NFIL3 in ID2 expression remains to be clarified given a contradicting report that ID2 expression is normal in NK cells lacking NFIL3 (48).

The role of ID2 in the lineage specification of NK cells is unclear, due to contradicting reports following deletion of the encoding gene Id2. While Yokota et al. (54). reported poor reconstitution of NK cells following bone marrow transplantation, implying a defect early on during NK cell development, Boos et al. (55) did not observe any reduction in NKP and iNK cell numbers. A recent demonstration of low ID2 levels in CLPs but consistently high levels in pre-pro NK, NKP, and immature and mature NK cells lends further support for the hypothesis that ID2 could indeed be important for NK lineage specification/maintenance (53). Interestingly, ID2 has been found to suppress T and B cell development through heterodimer formation with the E-box protein E2A (55–57), although how the interaction specifically promotes commitment to the NK cell lineage remains unknown.

A greater number of TFs have been shown to play a role in NK cell maturation. These factors include ID2, TBET, EOMES, Zinc finger E-box-binding homeobox 2 (ZEB2), Thymocyte selection-associated high mobility group box (TOX), IKAROS family zinc finger 3 (AIOLOS), Interferon regulatory factor 2 (IRF2), B lymphocyte-induced maturation protein 1 (BLIMP1), Forkhead box O1 (FOXO1), Kruppel-like factor 2 (KLF2), and GATA-binding protein 3 (GATA3). An overview of the expression these TFs during NK cell maturation is presented in Figure 1.

The role of ID2 in NK cell maturation is better understood than its role in early NK cell development. A recent study provided additional insight into the underlying mechanism of ID2 by demonstrating that it modulates the expression of E2A target genes (i.e., Socs3, Tcf7, and Cxcr5) by titrating E-protein activity, thereby controlling the responsiveness of NK cells to IL-15 that is crucial for survival (53, 58, 59).

TBET and EOMES are members of the T-box family of transcription factors that appear to regulate distinct checkpoints in NK cell maturation. TBET- and EOMES-deficient mice exhibited a similar phenotype where NK cell numbers were reduced in the all lymphoid tissues (52, 60, 61), except in the bone marrow of the former where there was an increase in NK cell numbers (60, 61). Detailed analyses of the bone marrow from TBET-deficient mice revealed that the increase in NK cell numbers was due to an accumulation of iNK cells, which the authors attributed to a developmental block (60). However, the possibility of a defect in NK cell trafficking remains unaddressed, given that TBET plays a role in the expression of sphigosine-1-phosphate receptor 5 (S1P₅) that is required for NK cell egression from the bone marrow (62). Nonetheless, both TFs are crucial for maturation, as mice that are deficient for both have a systemic lack of NK cells despite normal progenitor numbers (52, 61). As TBET and EOMES are, respectively, required by immature and mature NK cells, the two TFs are believed to function in a sequential manner for NK cell maturation (52, 60). Quite fittingly, both TBET and EOMES were shown to be essential for IL-15 responsiveness by enforcing high CD122 expression, with Il2rb (gene encoding CD122) being a direct target of EOMES (63).

Similar to that of TBET-deficient mice, NK cell numbers in the bone marrow of mice deficient for ZEB2 or BLIMP1 were higher than littermate controls (64, 65). The unusual accumulation of NK cells within the bone marrow of ZEB2-deficient mice was due to reduced S1P₅ expression (64), while enhanced proliferation was found to be the underlying cause in BLIMP1-deficient mice (65). Profound losses in terminally mature NK cells were observed on closer examination of the bone marrow NK cells from both strains (64, 65). The lack of mature NK cells in ZEB2-deficient was further attributed to poor responsiveness to IL-15, which resulted in poor survival (64).

The lack of mature NK cells was reported in mice that were deficient for TOX (66), AIOLOS (67), IRF2 (43, 44), KLF2 (68), or GATA3 (69, 70). A similar lack of mature NK cells in FOXO1-deficient mice was reported most recently (41), although this remains to be clarified against an earlier study, which instead found an accumulation of mature NK cells (42). Interestingly, the iNK cells in AIOLOS-deficient mice retained their expression of KLRG1, which is typically expressed on terminally mature NK cells (67). NK cells from IRF2-deficient mice have been shown to undergo accelerated apoptosis, indicating a role for IRF2 in regulating NK cell survival as well as maturation (43). Unlike the foregoing TFs, the temporal requirement for myeloid elf-1-like factor (MEF) by maturing NK cells has not been determined, as only an overall reduction in NK cell numbers was reported (71).

FOXO1 was recently found to be directly involved in the initiation of autophagy in iNK cells, most likely via cytosolic interactions between FOXO1 and the autophagy protein ATG7 (41). KLF2 appears to regulate NK cell maturation via a different mechanism, influencing the expression of homing receptors such as CD62L (i.e., L-selectin) on maturing NK cells, thereby dictating their access to IL-15 signaling that is essential for survival (68). Current knowledge of how GATA3 regulates NK cell maturation is limited to disturbances in expression of the TFs ID2, TBET, and NFIL3 (70). Nevertheless, GATA3 has also been shown to regulate NK cell egression from the bone marrow in a CXCR4-dependent manner, and also NK cell proliferation in response to IL-15 via CD25 expression (70).

Natural killer cell effector function is also regulated by many of the TFs outlined above. FOXO1 has also been proposed to negatively regulate NK cell effector function, as its absence was correlated with augmented IFN-γ production in response to murine cytomegalovirus (MCMV) infections and anti-metastatic activity against the B16F10 mouse melanoma cell line (42). Augmented anti-metastatic activity against the same melanoma cell line also was reported in mice deficient for either BLIMP1 or AIOLOS, despite the lack of any significant impact on cytokine production (65, 67). On the other hand, MEF is required for normal cytokine production and cytotoxicity, as it positively regulates IFN-γ and perforin expression, which corresponds to poorer cytotoxicity against tumor cell targets in MEF-deficient mice (71). Normal cytotoxicity but reduced IFN-γ production has been observed in IRF2-deficient mice and mice specifically lacking GATA3 in NK cells (43, 70). NK cell function is also regulated by TBET and EOMES, as TBET has been shown to bind to the regulatory regions of genes encoding granzyme B and perforin, while the expression of EOMES as NK cells mature is associated with increased transcription of mRNA (52, 60).