|Home | About | Journals | Submit | Contact Us | Français|
Natural killer (NK) cells are innate lymphocytes that are critical for host protection against pathogens and cancer due to their ability to rapidly release inflammatory cytokines and kill infected or transformed cells. In the 40 years since their initial discovery, much has been learned about how this important cellular lineage develops and functions. We now know that NK cells are the founding members of an expanded family of lymphocyte known as innate lymphoid cells (ILC). Furthermore, we have recently discovered that NK cells can possess features of adaptive immunity such as antigen specificity and long-lived memory responses. Here we will review our current understanding of the molecular mechanisms driving development of NK cells from the common lymphoid progenitor (CLP) to mature NK cells, and from activated effectors to long-lived memory NK cells.
NK cells, like B and T cells, are a lymphocyte lineage derived from the CLP , and like B cells, are thought to develop primarily in the bone marrow , although other sites of development, such as the liver and thymus, have also been proposed (reviewed in ). However, unlike the antigen receptors of B and T cells, NK cell receptors are germ line encoded and do not require gene rearrangement by RAG recombinase , though recent work has suggested that RAG plays an unexpected cell-intrinsic role in NK cell development . NK cells also undergo an “education” process during development where they acquire the ability to recognize lack of self MHC class I, or “missing-self”, a feature that facilitates their surveillance of target cells that have down-regulated MHC class I during infection or malignancy . NK cells rely on both cytokines and transcription factors to promote and control their development. Cytokine signaling from interleukin (IL)-15 is critical for the development of NK cells and is required throughout their lifetime [7,8]. Transcription factors such as Nfil3 and PU.1 are necessary for development of early NK cell progenitors [9-12], whereas Id2, Tox, and others are important later in development [13-15]. Eomes and T-bet are among factors that then control the final stages of NK cell maturation [16,17]. In the periphery, the activation and differentiation of NK cells are regulated by a plethora of transcription factors mediating distinct effector functions. This review will outline current knowledge about the stages of NK cell development and the factors driving each stage.
The CLP is characterized by expression of IL-7Rα (CD127), c-kit (CD117), Sca-1, and Flt-3 (CD135), as well as the lack of common lineage markers such as CD3, CD4, CD8, CD19, Ter119, Gr-1 and NK1.1 (Figure 1) . From the CLP, cells develop into NK cell precursors (NKP), which are defined by expression of the IL-15 receptor β chain (CD122), and lack of common lineage markers, including the NK cell markers NK1.1 and DX5 (CD49b) (Figure 1) . This NKP population has been further refined based on the co-expression of CD27 and CD244, with the majority of these cells also expressing IL-7Rα . An intermediate population between the CLP and NKP termed “pre-NKP” has also recently been defined as lineage negative, CD244+ c-kitlow IL-7Rα+ Flt-3− and CD122− [18,19]. However, recent work suggests that this population is heterogeneous, composed of true NK-committed precursors as well as PLZF- and α4β7 integrin-expressing ILC precursors (ILCP) (Figure 1) . A precursor of this pre-NKP population also capable of producing all ILC lineages (including NK cells) has recently been identified by expression of the transcription factor Tcf-1 . From the CD122+IL-7Rα+ NKP stage, cells develop into immature NK (iNK) cells, which lose expression of IL-7Rα and acquire expression of NK1.1 but do not yet express CD49b (Figure 1) . As immature NK cells gain expression of CD11b, CD43, Ly49 receptors, and CD49b (DX5), they also gain functional competence in cytotoxicity and production of interferon (IFN)-γ , and egress from the bone marrow.
The peripheral NK cell pool can be delineated by expression of CD27, with CD27lo/− NK cells being more cytotoxic and producing more cytokines than CD27high NK cells . These mature peripheral NK cell populations have more recently been further refined into four stages of maturation, defined by sequential upregulation of CD11b expression followed by downregulation of CD27, with the most immature NK cells being CD27−CD11b− and the most mature NK cells being CD27−CD11b+ . During viral infection or pro-inflammatory cytokine exposure, mature peripheral NK cells can differentiate into effector and long-lived memory NK cells (reviewed in ). During the CD8+ T cell response to viral infection, at least two different effector cell populations are thought to be generated: KLRG1hi short-lived effector cells (SLECs) and KLRG1lo memory precursor effector cells (MPECs) . Recent evidence suggests that a similar paradigm exists in the resting NK cell pool, with virus-specific KLRG1− NK cells exhibiting a greater capacity to generate memory NK cells than their KLRG1+ counterparts . In accordance with this finding, another recent study found that RAG expression during NK cell ontogeny was correlated with lower expression of KLRG1 and a greater memory potential .
Lineage commitment to either an adaptive or innate lymphocyte cell fate is determined by a complex network of transcription factors (Figure 2). For example, Notch signaling through the ligands Jagged1 and Jagged2 preferentially drives NK cell development from the CLP [28-30], whereas delta-like ligands (DLL) promote T cell development . Moreover, thymocytes can be diverted into an NK cell-like fate if the Notch1-dependent transcription factor Bcl11b is ablated during T cell development [32-34], suggesting active suppression of the NK cell fate. Similarly, early B cell factor 1 (Ebf1) and Pax5 promote the B cell fate by suppressing expression of ILC and T-cell promoting transcription factors Notch1, Tcf-1, Gata3, and Id2 . Even within the innate lymphocyte lineages, differential expression of specific transcription factors give rise to distinct cell fates. For example, although both NK cells and non-NK cell “helper” ILCs require the transcription factors Id2 [13,36,37] and Nfil3 [9,10,38-42] for their development, only the helper ILC lineages require Gata3 for development [43-45]. These differential requirements are consistent with recent studies indicating that ILCs are not derived from the same CD122+ precursor as NK cells, but rather arise from an IL-7Rα+, α4β7+, Id2-expressing precursor, referred to as the common helper innate lymphoid cell precursor or “CHILP” (reviewed in ).
Nfil3 (also known as E4BP4) is a critical factor in NK cell lineage commitment. Originally identified as a circadian clock gene , Nfil3 is widely expressed in many hematopoietic and non-hematopoietic cells, and is expressed as early as the CLP stage in developing lymphocytes . Early studies in Nfil3-deficient mice revealed a specific loss of NK cells, whereas numbers of B cells, CD4+, and CD8+ T cells were normal [9-11]. Later studies revealed that Nfil3 expression is only required in developing NK cells through the NKP stage; conditional deletion during the iNK stage does not impact NK cell numbers, cytokine production, or response to viral infection . Nfil3 expression is thought to be driven by IL-15, as IL-15 induces Nfil3 expression in NK cells and ectopic Nfil3 expression can partially rescue NK cell development in vitro in the absence of IL-15 signaling [9,50]. IL-15 is thought to drive Nfil3 expression through the kinase PDK1 and its downstream target mTOR  and mice with an NK cell specific deletion of mTOR have a block in maturation of bone marrow NK cells and a severe lack of NK cells in peripheral organs . However, several recent studies have shown that certain tissue resident NK cells may be Nfil3-independent [52-54].
B cells, T cells, and most ILCs require IL-7 for their development, whereas NK cells [37,55,56] and type 1 ILCs  do not. The first step towards an NK cell fate coincides with gain of CD122 and loss of IL-7Rα, reflecting a shift from IL-7 to IL-15 dependence. IL-15 signals through the transcription factor STAT5, and thus mice lacking Stat5b are deficient in NK cells . Deleting Stat5a/b in NKP also results in complete lack of NK cells . Similarly, the transcription factor Runx3 and its binding partner Cbfβ can promote CD122 expression, and NKP deficient in these factors fail to produce peripheral NK cells in fetal liver chimeras . An NKp46+ cell-specific deletion of Runx3 also results in a lack of peripheral NK cells . T-bet and Eomes have also been shown to cooperate to promote expression of CD122 and mice lacking both these transcription factors are deficient in NK cells as well as memory CD8+ T cells .
PU.1, a member of the Ets family of transcription factors (reviewed in ), is important in the development of T and B cells, monocytes, dendritic cells, and granulocytes (reviewed in ). Although PU.1-deficient fetal liver cells are able to generate NK cells, chimeric mice have reduced numbers of NKP and immature NK cells . PU.1-deficient NK cells also have increased expression of another Ets family factor, Ets-1, a finding that has led to suggestions that Ets-1 can compensate for lack of PU.1 in driving the NK cell lineage . Although Ets-1 is likely expressed prior to the NKP stage, it appears to be required at a later point in development than PU.1. Peripheral NK cells are severely decreased in Ets-1-deficient mice  due to an arrest at the NKP stage . Furthermore, residual NK cells in Ets-1-deficient mice were refractory when stimulated through activating receptors but hyperresponsive to proinflammatory cytokines, suggesting chronic basal stimulation . Ets-1 was shown to promote other transcription factors critical in NK cell development, including Id2 and T-bet . Similarly, mice lacking the Ets family member Mef have decreased NK cell numbers, as well as a defect in IFN-γ secretion and perforin expression .
In addition to the early role for Id family transcription factors in suppressing the adaptive lymphocyte fate while promoting innate lymphocyte development, these factors are also important later in the development of NK cells. Id2-deficient mice have a cell-intrinsic lack of peripheral NK cells  that was found to be due to an arrest at the iNK stage , indicating that Id2 is important in the transition from immature to mature NK cell. Both Id2 and Id3 are expressed in NKP, and Id2 continues to be expressed in NK cells through the mature NK cell stage. In addition, both Id2 and Id3 can promote NK cell development in culture [68,69] and Id2 is thought to be downstream of Nfil3, as ectopic Id2 expression can rescue NK cell development in Nfil3 deficient progenitors [9,48].
The T-box family of transcription factors is critical in several aspects of NK cell development and maturation. One family member, T-bet, is thought to regulate expression of S1P5, a receptor required for NK cell egress out of lymph nodes and bone marrow . T-bet-deficient mice lack mature NK cells in the bone marrow and periphery, exhibiting an arrest at the iNK stage during development . A more recent study suggests that T-bet stabilizes an immature (TRAIL+DX5−) NK cell state and that loss of T-bet results in higher expression of Eomes, another T-box transcription factor . Eomes is required for transition to the mature (DX5+) NK cell stage and acquisition of Ly49 receptors . However, there is some question as to whether this TRAIL+ population truly consists of immature NK cells or whether it represents a distinct lineage of ILC1 (reviewed in [3,71]). Loss of T-bet also results in reduced expression of another transcription factor, Blimp1, which is similarly required for progression from the iNK stage . The transcription factor Gata3 is thought to promote T-bet expression, as Gata3-deficient NK cells have reduced T-bet expression and are immature in phenotype similar to T-bet-deficient NK cells . Conversely, the forkhead box family transcription factors Foxo1 and Foxo3 suppress NK cell maturation by repressing T-bet . Foxo3-deficient mice have normal numbers of NK cells but an increase in KLRG1+ NK cells, suggesting a role for Foxo3 in suppressing terminal maturation of NK cells .
The transcription factor Tox is required for transition from the iNK to mNK stage, as Tox-deficient mice have a severe defect in mature NK cells but no defect in immature NK cells or NKP . Tox is believed to be downstream of Nfil3, as transduction of Nfil3-deficient bone marrow with Tox-expressing retroviruses was able to rescue NK cell and ILC development in recipient mice . Aiolos, an Ikaros family member, is also required for terminal maturation of NK cells; Aiolos-deficient mice have an accumulation of CD27+CD11b− and CD27+CD11b+ NK cells but a loss of the most mature CD27−CD11b+ NK cell subset . Irf2 is similarly required for promoting mature NK cells, as Irf2-deficient mice maintain normal numbers of immature NK cells in the bone marrow but lack the most mature CD11b+ subset and circulating NK cells [77,78]. Runx3, important in promoting CD122 expression , has likewise been shown to be important late in NK cell development, promoting the expression of Ly49 receptors, CD11b, and CD43 .
The STAT family of transcription factors contains members that are phosphorylated downstream of pro-inflammatory cytokine receptors and form homo- or hetero-dimers that translocate to the nucleus to induce gene transcription (reviewed in ). During viral infection, type I IFNs and downstream STAT1 have been shown to enhance NK cell cytotoxicity (Figure 3) [81,82], and shield activated NK cells from cell death via an NKG2D-dependent fratricide mechanism . IL-12 and downstream STAT4 are required for NK cell production of IFN-γ [81,82,84], and NK cell expansion and memory generation after mouse cytomegalovirus (MCMV) infection . IL-33, IL-18, and MyD88 are also important for optimal expansion during viral infection, but are not required for memory cell formation [86,87]. In addition, IL-12, IL-18, and type I IFNs together drive expression of the transcription factor Zbtb32, which promotes NK cell proliferation after MCMV infection by antagonizing Blimp-1 . The aryl hydrocarbon receptor (AhR) is another nuclear factor required for optimal cytotoxicity of NK cells . Similarly, NK cells deficient in either of the transcription factors C/EBP or MITF have reduced cytotoxicity and IFN-γ secretion [90-93]. MITF may regulate cytotoxicity through interactions with MEF and PU.1 at the perforin promoter [67,94].
Control of the apoptosis pathway is thought to be involved in regulating NK cell memory formation, as the anti-apoptotic molecule Bcl-2 is downregulated in NK cells following MCMV infection [88,95]. The pro-apoptotic factor Bim controls the formation of memory NK cells, with Bim-deficient NK cells failing to contract normally following MCMV infection and displaying lower levels of memory-associated cell surface markers . A recent study found that NK cells accumulate damaged mitochondria after MCMV infection, and that a small subset that cleared these mitochondria by autophagy preferentially survived to form memory NK cells, a process dependent on BNIP3 and BNIP3L . Additional factors that may be specifically required for memory formation remain to be elucidated.
As is evident by the number of groundbreaking studies discussed in this review, the development of NK cells is a highly dynamic process and new discoveries occur each week. Still, there is so much we do not yet understand. Future work will elucidate the mechanisms regulating the various factors described above, how the factors interact with each other, and how they might be involved in disease processes.
We thank members of the Sun lab for helpful discussions. In particular we thank Aimee Beaulieu, Clair Geary, and Tim O'Sullivan, who read and commented on this review. We apologize to those whose work we were unable to discuss due to space limitations. T.L.G. is supported by a fellowship from the National Institute of Allergy and Infectious Diseases (F31 AI114019). J.C.S. is supported by the Searle Scholars Program, the Cancer Research Institute, and National Institutes of Health grants AI085034 and AI100874. Our laboratory is also supported by National Institutes of Health/National Cancer Institute Cancer Center Support Grant (CCSG) P30CA008748.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.