|Home | About | Journals | Submit | Contact Us | Français|
Natural killer (NK) cells play critical roles defending against tumors and pathogens. We show that mice lacking both transcription factors Eomesodermin (Eomes) and T-bet failed to develop NK cells. Developmental stability of immature NK cells constitutively expressing the death ligand TRAIL depended on T-bet. Conversely, maturation characterized by loss of constitutive TRAIL expression and induction of Ly49 receptor diversity and integrin CD49b (DX5+) required Eomes. Mature NK cells from which Eomes was deleted reverted to phenotypic immaturity if T-bet was present or downregulated NK lineage antigens if T-bet was absent, despite retaining expression of Ly49 receptors. Adult, hepatic and fetal hematopoiesis restricted Eomes expression and limited NK development to the T-bet-dependent, immature stage, whereas medullary hematopoiesis permitted Eomes-dependent NK maturation in adult mice. These findings reveal two sequential, genetically separable checkpoints of NK cell maturation, the progression of which is metered largely by the anatomic localization of hematopoiesis.
NK cells derive from hematopoietic progenitors. A precursor that lacks expression of NK antigens but expresses CD122 (IL-2Rβ and IL-15Rβ) is believed to be committed to the NK lineage and restricted from other blood cell fates (Rosmaraki et al., 2001). NK precursors generate NK antigen-bearing (NK1.1+NKp46+) NK cells that can undergo further phenotypic maturation and express the integrins CD49b (DX5+) and CD11b (Chiossone et al., 2009; Kim et al., 2002; Walzer et al., 2007). Mature, DX5+ NK cells express a diverse repertoire of Ly49 family receptors, responsible for educating NK cells to self-MHC and for enabling a broad specificity against microbial components (Kim et al., 2002; 2005; Orr and Lanier, 2010). Development of adult murine NK cells is thought to take place predominantly in the bone marrow, though a variety of other tissues in mice and humans, such as the liver, lymph node, and thymus, may support NK cell development (Di Santo, 2006; Freud and Caligiuri, 2006; Luther et al., 2011; Takeda et al., 2005; Vosshenrich et al., 2006).
The transcription factors ID2, TOX, and E4BP4 (Nfil3) are thought to specify the earliest stages of NK cell development (Aliahmad et al., 2010; Boos et al., 2007; Gascoyne et al., 2009; Kamizono et al., 2009). Other transcription factors, including Blimp-1, Ets-1, GATA-3, IRF-2, MEF, and PU.1 play specific roles at distinct stages of NK cell development and maturation (Barton et al., 1998; Colucci et al., 2001; Kallies et al., 2011; Lacorazza et al., 2002; Lohoff et al., 2000; Samson et al., 2003). The precise hierarchy of transcription factors governing NK cell maturation, however, is incompletely understood.
The highly homologous T-box transcription factors Eomesodermin (Eomes) and T-bet direct fate and function in cytotoxic lymphocytes. Eomes and T-bet redundantly regulate differentiation of CD8+ effector T cells (Intlekofer et al., 2005; 2008; Pearce et al., 2003). Eomes and T-bet also appear to have non-redundant functions in specifying the fate of CD8+ T cells (Banerjee et al., 2010; Intlekofer et al., 2007). It has been suggested that some aspects of terminal NK cell maturation are dependent on T-bet (Jenne et al., 2009; Townsend et al., 2004). We now report an additional role for T-bet in stabilizing the immature NK cell fate. The consequence of complete deletion of Eomes on NK cell development had not yet been determined. We show that maturation of NK cells to the DX5+ stage characterized by acquisition of a diverse repertoire of activating and inhibitory Ly49 family receptors is dependent on Eomes. Deletion of Eomes from mature NK cells caused reversion to a more immature state. Loss of both T-bet and Eomes from mature NK cells resulted in loss of classical NK antigens. Progenitors lacking both Eomes and T-bet could not support NK lineage development. Our data support a model in which expression and function of T-bet and Eomes define key, genetically separable molecular checkpoints of NK cell maturation.
We first examined the expression of Eomes protein at the single-cell level in murine NK cells. The majority of NK cells expressed Eomes (Eomes+), but we detected a minor population of Eomes− NK cells in each organ examined (Figure 1A; Figure S1A). The Eomes− population was most enriched in the liver. A subset of less mature NK cells that express the death ligand TRAIL and lack CD49b are present in neonates and preferentially reside in the adult liver (Kim et al., 2002; Takeda et al., 2005). Eomes− cells were characterized by TRAIL expression, whereas Eomes+ cells expressed CD49b (DX5+) (Figure 1A). Eomes− and Eomes+ NK cells also expressed different repertoires of homing receptors, consistent with their disparate anatomic localization. Eomes− NK cells expressed Integrin αv and the chemokine receptors CXCR3 and CXCR6 at greater levels relative to Eomes+ NK cells (Figure 1B). Eomes− NK cells expressed substantially lower levels of the S1P receptors S1P1 and S1P5 than did Eomes+ NK cells (Figure 1C). Eomes− NK cells also underwent less steady-state proliferation in vivo compared to Eomes+ NK cells (Figure S1B).
To understand the requirements for Eomes in development and homeostasis of NK cells, we analyzed mice harboring floxed alleles of Eomes (Intlekofer et al., 2008) and expressing hematopoietic-specific Cre recombinase under control of Vav regulatory elements (Stadtfeld, 2004). EomesFlox/Flox, Vav-Cre+ mice had substantial reduction of NK cells in spleen and blood (Figure 1D; Figure S1C). We observed a more modest reduction in the numbers of NK cells in liver, lymph node, and bone marrow of EomesFlox/Flox, Vav-Cre+ mice. Despite the differences in NK cell number, total cellularity of NK-containing organs in wild-type and EomesFlox/Flox, Vav-Cre+ mice were similar (Figure S1D).
The phenotype of NK cells from EomesFlox/Flox, Vav-Cre+ mice approximated that of Eomes− NK cells from wild-type mice, expressing TRAIL but lacking CD49b (DX5−) (Figure 1E; Figure S1E). The requirement for Eomes to develop DX5+ NK cells appeared to be cell-intrinsic. In wild-type plus EomesFlox/Flox, Vav-Cre+ mixed bone marrow chimeras, we observed that EomesFlox/Flox, Vav-Cre+ marrow contributed inefficiently to the NK populations in the blood, spleen, and bone marrow, which were composed predominantly of mature, DX5+ NK cells (Figures S1G and S1H). In the liver, however, EomesFlox/Flox, Vav-Cre+ marrow competitively contributed to the TRAIL+ NK pool. The limited contribution of Eomes-deficient marrow to the steady-state adult NK cell compartment is consistent with prior suggestions that the major proliferative burst in NK cell maturation occurs at the DX5+ stage and that TRAIL+DX5− NK cells represent a minor subset in adult mice (Kim et al., 2002; Takeda et al., 2005).
DX5+ NK cells express a full complement of Ly49 family receptors and can express the integrin CD11b (Kim et al., 2002). Compared to Eomes+ NK cells, EomesFlox/Flox, Vav-Cre+ and wild-type Eomes− NK cells exhibited a limited repertoire of Ly49 receptors, expressing Ly49A, Ly49D, Ly49G2, and Ly49H at markedly reduced frequencies (Figure 1F; Figure S1F). EomesFlox/Flox, Vav-Cre+, wild-type Eomes−, and wild-type Eomes+ NK cells did, however, express Ly49C and/or Ly49I at similar frequencies. Few EomesFlox/Flox, Vav-Cre+ and wild-type Eomes− NK cells expressed high levels of CD11b, whereas Eomes+ NK cells were predominantly CD11bhi (Figure 1G). Thus, expression of Eomes appears to demarcate two subsets of adult NK cells: less mature, Eomes−TRAIL+ NK cells and more mature, Eomes+DX5+ NK cells. Eomes seems essential for NK cells to mature past the TRAIL+ stage and become DX5+ with a diverse repertoire of Ly49 receptors.
Fetal and neonatal NK cells lack markers of maturity and express limited Ly49 receptor repertoires (Kubota et al., 1999; Takeda et al., 2005), somewhat akin to NK cells in adult EomesFlox/Flox, Vav-Cre+ mice. We found that NK cells from neonatal wild-type mice have minimal expression of Eomes (Figure 1H). Postnatal NK cells in the spleen initiated Eomes expression earlier and more efficiently than did hepatic NK cells. The delayed onset and anatomic bias of Eomes expression is consistent with the prior suggestion that DX5+ NK maturation becomes evident post-natally and preferentially outside the liver (Takeda et al., 2005). Whether hepatic signals restrict Eomes or whether signals present in other sites promote Eomes expression remains to be determined.
Expression of Eomes by postnatal NK cells corresponds to the ontogenic shift in hematopoiesis from the liver to the bone marrow (Keller et al., 1999). Whether adult medullary hematopoiesis possesses an Eomes−TRAIL+ intermediate stage prior to maturation of Eomes+DX5+ NK cells was not known. Adult NK development is seemingly characterized by an immature CD27lowCD11blow NK1.1+NKp46+ cell giving rise to a CD27hiCD11blo NK1.1+NKp46+ cell, which can mature further and upregulate CD11b (Chiossone et al., 2009; Kim et al., 2002; Walzer et al., 2007). We found acquisition of TRAIL occurs at the earliest stage of definitive NK cell development, as the majority of medullary CD27loCD11blo NK1.1+NKp46+ cells express TRAIL (Figure 1I). The next stage (CD27hiCD11blo) contains both TRAIL+ and DX5+ NK cells, consistent with prior studies suggesting that NK cells become DX5+ prior to expressing CD11b in the bone marrow (Kim et al., 2002). Reciprocal expression of TRAIL and Eomes was also a consistent feature of medullary NK cell maturation (Figure 1I). Expression of Eomes and loss of TRAIL thus defines both the ontogenic appearance of mature NK cells and the acquisition of mature NK markers by developing adult NK cells.
Eomes appears to direct NK maturation past the immature, TRAIL+Integrin αv+CD11blo state to the mature, DX5+CD11bhi state. Whether Eomes− NK cells are the direct predecessors of Eomes+ NK cells has not been determined. Analyses of NK cell developmental in bone marrow and cell transfer experiments using defined intermediates are compatible with a model wherein TRAIL+Integrin αv+ NK cells may give rise to DX5+ NK cells (Chiossone et al., 2009; Kim et al., 2002; Takeda et al., 2005). To address whether TRAIL+ NK cells can upregulate Eomes while maturing into DX5+ NK cells, we sorted hepatic TRAIL+ NK cells to 97% purity and transferred them into unirradiated, immunodeficient Il2rg−/−Rag2−/− mice, which lack NK cells and other lymphocytes.
One (Figure 2A) or two (Figure 2B) weeks following transfer, the livers, spleens, and bone marrow of recipients were analyzed for presence of NK cells. Transferred immature, TRAIL+ NK cells underwent in vivo maturation characterized by repression of TRAIL as well as induction of CD49b (DX5+), Eomes, and Ly49 receptors (Figures 2A and 2B). NK cell maturation and induction of Eomes was restricted in liver, promoted in spleen, and favored most in bone marrow. Conversion of TRAIL+ NK cells into DX5+Eomes+ NK cells in immunodeficient mice is consistent with prior findings that purified neonatal or adult TRAIL+ NK cells can downregulate TRAIL and induce CD49b (DX5+) when transferred into sublethally-irradiated, immunocompetent recipients (Takeda et al., 2005). We observed analogous, yet less efficient, conversion of Eomes−TRAIL+ to Eomes+DX5+Ly49+ cells using immunocompetent recipients (Figure S2A). Purified, hepatic DX5+ NK cells transferred to Il2rg−/−Rag2−/− mice remained DX5+, Eomes+, and Ly49+ after adoptive transfer (Figures 2A and 2B), consistent with prior experiments showing stability of CD49b expression (Takeda et al., 2005). Acquisition of Eomes expression in the mature, DX5+ stage of NK cell development thus appears to be a stable event.
To assess whether Eomes is essential for the conversion of TRAIL+ to DX5+ NK cells, we purified TRAIL+ NK cells from EomesFlox/Flox, Vav-Cre+ donor mice and transferred the cells to Il2rg−/−Rag2−/− recipients. In contrast to wild-type TRAIL+ donor NK cells, EomesFlox/Flox, Vav-Cre+ TRAIL+ donor NK cells neither downregulated TRAIL efficiently nor stained brightly with DX5 (Figure 2C). After transfer, EomesFlox/Flox, Vav-Cre+ TRAIL+ donor cells did exhibit Ly49D, Ly49G2, and/or Ly49H expression at a diminished but detectable frequency relative to wild-type donor cells, which may be consistent with prior suggestions that Ly49 receptor acquisition can precede CD49b induction and cellular expansion in medullary NK cell development (Kim et al., 2002). Together, these data are consistent with a model in which induction of Eomes is essential for immature, TRAIL+ NK cells to become mature, DX5+ NK cells with robust expression of diverse Ly49 receptors.
Because TRAIL is expressed by the least differentiated adult NK cells in the bone marrow (Figure 1I), we tested whether medullary TRAIL+ NK cells can give rise to mature, DX5+ NK cells. TRAIL+ CD27loCD11blo and TRAIL+ CD27hiCD11blo NK cells were sorted from the bone marrow (at least 95% purity) and adoptively transferred into Il2rg−/−Rag2−/− mice. Transferred medullary TRAIL+ NK cells matured into TRAIL−Eomes+ NK cells and did so to a greater extent in the spleens than in the livers of recipient mice (Figure 2D). Additionally, immature, TRAIL+ NK cells converted to mature NK cells as measured by expression of CD27 and CD11b, consistent with previous adoptive transfer studies (Chiossone et al., 2009). These data support a model in which medullary NK precursors give rise to cells that express NK antigens and which are CD27loCD11blo and TRAIL+Eomes−. In the bone marrow environment, TRAIL+Eomes− NK cells readily undergo Eomes-dependent maturation to become DX5+Eomes+ NK cells that progress through defined stages of CD11b induction and subsequent CD27 repression. In contrast, Eomes-dependent maturation by TRAIL+Eomes− NK cells is restricted in the hepatic environment.
To investigate the role of Eomes in maintaining NK cells in the mature state, we temporally deleted Eomes from mature NK cells. Splenic NK cells with floxed alleles of Eomes were isolated and treated directly ex vivo with Cre-recombinase fused to the transduction domain of the HIV TAT protein (TAT-Cre) (Wadia et al., 2004). Treated or sham-treated cells were adoptively transferred into unirradiated Il2rg−/−Rag2−/− mice. From recipients of sham-treated cells, we retrieved Eomes+DX5+ NK cells that expressed diverse Ly49 receptors and Eomes−TRAIL+ NK cells that displayed a paucity of Ly49 receptor expression (Figures 3A and 3B). From recipients of TAT-Cre-treated cells, we retrieved an additional population of Eomes− cells that expressed Ly49 receptors, which presumably represented formerly mature, Eomes+DX5+ NK cells that underwent Cre-mediated deletion of Eomes (Figures 3A and 3B). The stability of all Ly49 family members examined in cells lacking Eomes protein suggested that maintenance of Ly49 receptor expression may not depend on Eomes, even though establishment of a diverse Ly49 repertoire is Eomes-dependent.
In contrast to the stability of Ly49 expression, we found temporal deletion of Eomes resulted in loss of other maturity markers. A substantial proportion of Ly49+ NK cells that lost expression of Eomes became TRAIL+ and DX5− in the liver, spleen, and bone marrow of recipient mice (Figure 3B). We also observed partial derepression of Integrin αv and CXCR6 in hepatic but not splenic Ly49+ NK cells after loss of Eomes (Figure S3). These findings suggest that Eomes directly or indirectly represses the markers of immaturity TRAIL, Integrin αv, and CXCR6, while inducing the maturity marker CD49b (DX5+). Eomes+DX5+ NK cells normally retain their TRAIL−DX5+ identity (Figure 2), but loss of Eomes can cause mature cells to downregulate CD49b (DX5−) and upregulate TRAIL, despite maintaining expression of Ly49 receptors (Figure 3B). The reversion in phenotypic attributes after temporal deletion of Eomes complements the prior findings suggesting that Eomes−TRAIL+ NK cells can be in vivo intermediates that give rise to mature, Eomes+DX5+ NK cells (Figure 2). Together, these experiments provide forward and reverse evidence that acquisition of Eomes elaborates and maintains mature attributes on the immature NK cell foundation.
To evaluate the role of T-bet in NK cell development, we compared the phenotypes and patterns of T-box protein expression in NK lineages of EomesFlox/Flox, Vav-Cre+ and Tbx21−/− mice. Tbx21−/− mice contained predominantly NK cells expressing CD49b (DX5+), a diverse Ly49 repertoire, and high levels of Eomes (Figures 4A and 4B; Figures S4A, S4B, and S4C). Eomes−TRAIL+ NK populations were seemingly absent from Tbx21−/− mice, while TRAIL+ cells represented the predominant NK population of EomesFlox/Flox, Vav-Cre+ mice. T-bet was not required for expression of TRAIL itself, however, as Tbx21−/− NK cells readily induced TRAIL following in vitro stimulation (Figure S4D). Tbx21−/− and EomesFlox/Flox, Vav-Cre+ mice thus have reciprocal maturational phenotypes: Tbx21−/− mice lack immature (TRAIL+DX5−) NK cells, while EomesFlox/Flox, Vav-Cre+ mice lack mature (TRAIL−DX5+) NK cells.
T-bet is normally expressed in both Eomes− and Eomes+ NK cells (Figure 4B; Figure S4A), which is consistent with roles for T-bet in both immature and mature NK cell stages. In addition to playing an essential role in development or maintenance of the immature, Eomes−TRAIL+DX5− subset (Figures 4A and 4B; Figure S4A and S4B), T-bet controls repression of CD27 and c-kit and induction of CD43 and KLRG1 by NK cells (Figure S4E), as has been previously suggested (Jenne et al., 2009; Kallies et al., 2011; Townsend et al., 2004). Although mature, DX5+ NK cells can develop without T-bet, the most terminal stages of maturation are incomplete in the absence of T-bet. Using wild-type plus Tbx21−/− mixed bone marrow chimeras, we found that the previously described requirement for T-bet in repressing CD27 among mature NK cells was predominantly cell-intrinsic (Figures S4F and S4G). This finding contrasts with the recent observation that T-bet-deficient NK cells might be capable of repressing CD27 in a wild-type environment (Soderquest et al., 2011), suggesting that the effect of T-bet could also be partially cell-extrinsic. We also observed the newly described requirement for T-bet in development or maintenance of the immature Eomes−TRAIL+DX5− NK population to be cell-intrinsic (Figures S4F and S4G).
T-bet appears required to develop or stabilize immature NK cells, and immature NK cells are the predominant NK cells in the perinatal period (Takeda et al., 2005). We, therefore, assessed the NK compartment of neonatal T-bet-deficient mice. Tbx21−/− neonates possessed substantially reduced NK cell numbers relative to age-matched, wild-type control mice (Figures 4C and 4D). The NK cells present in Tbx21−/− neonates expressed Eomes (Figure S4H). Expression of TRAIL on Eomes+ NK cells of Tbx21−/− neonates was higher than on wild-type Eomes+ NK cells, which suggests that DX5+ NK cells in adult Tbx21−/− mice may transit through a TRAIL+ intermediate. In contrast to the paucity of NK cells in neonatal Tbx21−/− mice, tissues from neonatal EomesFlox/Flox, Vav-Cre+ mice contained comparable numbers of NK cells to littermate controls (Figure S4I), consistent with the apparent restriction of Eomes expression in neonatal NK cell development (Figure 1H).
To address whether stability of TRAIL+ NK cells depends on T-bet, we temporally deleted T-bet by treating hepatic NK cells with floxed alleles of Tbx21 with TAT-Cre ex vivo and transferring cells into Il2rg−/−Rag2−/− mice. From recipients of sham-treated cells, we recovered both immature, Eomes−TRAIL+ NK cells and mature, Eomes+ NK cells (Figure 4E). From recipients of TAT-Cre-treated cells, we detected a sizeable population of NK cells that lost expression of T-bet (Figure S4J). TAT-Cre-treatment markedly diminished the presence of Eomes−TRAIL+ NK cells while not seemingly affecting the recovery of mature, Eomes+ NK cells (Figure 4E). Whether by maintaining its survival or curbing its progressive maturation, these findings suggest that T-bet stabilizes the Eomes−TRAIL+ stage of NK cell development. Accelerated maturation of Eomes−TRAIL+ NK cells to Eomes+DX5+ NK cells after loss of T-bet might be possible owing to the finding that NK cells from which T-bet was acutely (Figure S4J) or chronically (Figure 4B; Figure S4A) deleted both exhibited increased levels of Eomes expression. These data support a model in which immature, neonatal and adult Eomes−TRAIL+ NK cells depend on T-bet for developmental stability.
T-bet and Eomes seem to be required for stability or development of sequential (or alternative) stages of NK cell maturation. We intercrossed EomesFlox/Flox, Vav-Cre+ mice with Tbx21−/− mice to yield Tbx21−/−EomesFlox/Flox, Vav-Cre+ animals, lacking both Eomes and T-bet in the hematopoietic compartment. We could not detect NK cells in any organ of Tbx21−/−EomesFlox/Flox, Vav-Cre+ animals (Figure 4F), which is consistent with the hypothetical addition of the two complementary, Tbx21−/− and EomesFlox/Flox, Vav-Cre+ phenotypes (Figure 4A). Despite absence of NK antigen-expressing NK cells, we found that NK antigen-negative, CD122hi precursors of NK cells were present in Tbx21−/−EomesFlox/Flox, Vav-Cre+ mice (Figure S4K). The presence of NK precursors in Tbx21−/−EomesFlox/Flox, Vav-Cre+ animals is consistent with the low levels of T-bet and Eomes normally expressed in wild-type precursors (Figure S4L).
We next temporally deleted floxed alleles of Eomes from mature NK cells already lacking T-bet and either adoptively transferred them into immunodeficient mice or cultured them in Interleukin-2 in vitro. T-bet-deficient, mature NK cells that underwent temporal deletion of Eomes maintained Ly49 receptor expression but downregulated the NK antigens NK1.1 and NKp46 (Figure 4G) as well as TRAIL and CD49b (DX5), the identifying markers of immaturity and maturity of NK cells, respectively (Figure 4H). In contrast, T-bet-sufficient control cells from which Eomes was deleted maintained NK-lineage identity despite downregulating CD49b (DX5) and upregulating TRAIL. Expression of CD122 on acutely double-deficient NK cells was minimally affected (Figure S4M). These data are consistent with a model in which T-bet and Eomes are dispensable for development of CD122-expressing NKPs but are essential for stable development of NK antigen-expressing NK cells.
Prior analyses suggest that both TRAIL+ and DX5+ NK cells can lyse YAC-1 target cells and produce Interferon-gamma (IFN-γ) when stimulated in vitro (Takeda et al., 2005). T-bet-deficient mature NK cells are thought to possess lytic and cytokine-secreting function (Townsend et al., 2004). We found that Eomes−TRAIL+ and Eomes+DX5+ NK subsets from wild-type mice, as well as Tbx21−/− NK cells, EomesFlox/Flox, Vav-Cre+ NK cells, and T-bet-deficient NK cells with temporally deleted alleles of Eomes were all poised for substantial effector gene expression and function when challenged with a variety of stimuli in vitro and in vivo. (Figure 5 and Figure S5). Mature Eomes+DX5+ NK cells expressed modestly more Perforin (Prf1) mRNA than immature Eomes−TRAIL+ NK cells (Figure 5A; Figure S5A), consistent with a role for Eomes in inducing Prf1 (Cruz-Guilloty et al., 2009; Intlekofer et al., 2005; Pearce et al., 2003). Immature Eomes−TRAIL+ NK cells expressed increased Granzyme C (Gzmc) mRNA (Figure 5A), and elevated levels of Granzyme B (GzmB) and TNF-α protein, consistent with the phenotype of EomesFlox/Flox, Vav-Cre+ NK cells (Figures 5B and 5C; Figures S5C and S5D). Immature and mature wild-type NK cells, as well as Tbx21−/− NK cells, EomesFlox/Flox, Vav-Cre+ NK cells, and T-bet-deficient NK cells with temporally deleted alleles of Eomes, were all able to degranulate and express IFN-γ (Figures 5B and 5C; Figures S5B, S5C and S5D).
We also found that EomesFlox/Flox, Vav-Cre+ and Tbx21−/− mice were able to clear target cells missing self in vivo (data not shown), which is consistent with findings that Eomes− and Eomes+ NK cells expressed NKG2A (Figure S5E) and Ly49C and/or Ly49I (Figure 1F), which bind MHC Class I ligands in C57BL/6 mice (Brennan et al., 1996a; 1996b; Liu et al., 2000) and arm or license NK cells to kill targets (Kim et al., 2005; Raulet and Vance, 2006; Vance et al., 1998). NK cells, thus, appear to acquire substantial effector function once beyond the NK progenitor stage. Preservation of functions in NK cells temporally deficient in both T-box transcription factors suggest that effector gene expression may be epigenetically stable, or regulated by factors other than T-bet and Eomes (Barton et al., 1998; Lacorazza et al., 2002; Lohoff et al., 2000; Samson et al., 2003).
Infection with murine cytomegalovirus (MCMV) results in activation and expansion of NK cells in mice (Arase et al., 2002; Dokun et al., 2001; Lee et al., 2001). We tested whether infection with MCMV would induce EomesFlox/Flox, Vav-Cre+ NK cells to express markers of maturity and activation, CD49b (DX5+), CD11b, and KLRG1 (Robbins et al., 2004; Zafirova et al., 2009). MCMV infection induced expansion of MCMV-specific Ly49H+ NK cells in wild-type, Tbx21−/−, and EomesFlox/Flox, Vav-Cre+ (Figure 5D). Ly49H+ NK cells from infected mice of each genotype expressed high levels of CD11b and KLRG1, although the frequency was highest in cells from wild-type and lowest in cells from EomesFlox/Flox, Vav-Cre+ mice (Figure 5E). The majority of the Ly49H+ NK cells in MCMV-infected EomesFlox/Flox, Vav-Cre+ mice did not repress TRAIL or express CD49b (DX5+). Those Ly49H+ EomesFlox/Flox, Vav-Cre+ NK cells that were CD11bhi and KLRG1hi (and thus appeared most activated) were more likely to be DX5+ and TRAIL−, although it is not possible to determine whether this represents stable, Eomes-independent acquisition of a mature phenotype (Figure 5F). The failure to detect Ly49D on EomesFlox/Flox, Vav-Cre+ NK cells suggests that infection with MCMV does not completely bypass the requirement for Eomes in NK cell maturation.
Developing NK cells are believed to derive from lymphoid progenitors (Spits and Di Santo, 2011). Restriction of B cell potential and commitment to the innate lymphoid lineage appears to be mediated by ID2 (Boos et al., 2007; Carotta et al., 2011). ID2-expressing NK precursors may give rise to canonical NK cells in an E4BP4- and TOX-dependent fashion (Aliahmad et al., 2010; Carotta et al., 2011; Gascoyne et al., 2009; Kamizono et al., 2009). Previous studies have elucidated various phenotypic stages of NK maturation (Chiossone et al., 2009; Kim et al., 2002; Takeda et al., 2005), but the molecular events that control passage through these stages are not fully understood.
We found that NK cell development is profoundly impaired in the absence of both Eomes and T-bet. The phenotypes of Tbx21−/− mice and EomesFlox/Flox, Vav-Cre+ mice were not, however, identical. The most striking defect in each case was the loss of a distinct and non-overlapping phenotypic subset from that lacking in the other T-box factor-deficient mouse. The Tbx21−/−EomesFlox/Flox, Vav-Cre+ phenotype approximated the addition of Tbx21−/− and EomesFlox/Flox, Vav-Cre+ phenotypes. Eomes or T-bet may also be essential to maintain the lineage identity of NK cells since temporal loss of both T-box factors seems to abrogate expression of NK1.1 and NKp46 in cells that were formerly mature NK cells.
Our data support a model wherein NK precursors may upregulate T-bet and transit to an immature, NK1.1+NKp46+ Eomes−DX5−TRAIL+T-bet+ state. This developmental sequence seems to be the predominant, if not exclusive, manner of NK cell development in embryogenesis and in neonates, when hematopoiesis is predominantly occurring in the liver (Keller et al., 1999). Immature, Eomes−DX5−TRAIL+T-bet+ NK cells persist in adulthood, predominantly as a stable lineage in the liver and as a more mercurial intermediate that transits to a mature stage in the bone marrow (Kim et al., 2002; Takeda et al., 2005). Importantly, the immature, DX5−TRAIL+ NK cell identity appears strictly dependent on T-bet but not Eomes for its stability.
In adult hematopoiesis, which occurs predominantly in the bone marrow, NK precursors appear to give rise to a similarly immature-appearing, NK1.1+NKp46+ CD27loCD11blo Eomes−DX5−TRAIL+T-bet+ stage (Chiossone et al., 2009; Kim et al., 2002; Walzer et al., 2007). In contrast to hepatic hematopoiesis, the bone marrow environment appears permissive for Eomes−TRAIL+ NK cell intermediates to acquire expression of Eomes, yielding subsequent stages of NK cell maturation that we now show are strictly dependent on the function of Eomes. In the adult, reciprocal TRAIL and Eomes expression appears to bisect bone marrow NK cell development akin to the ontogenic shift in NK cell development observed postnatally. Relatively immature CD27hiCD11blo NK cells (Chiossone et al., 2009) contain both immature (TRAIL+Eomes−) and mature (TRAIL−Eomes+) NK cells. The apparent overlap of the Tbx21−/− and EomesFlox/Flox, Vav-Cre+ phenotypes in the CD27hiCD11blo medullary stage is resolvable into non-overlapping TRAIL+ and DX5+ CD27hiCD11blo substages, based on the findings that bone marrow from Tbx21−/− and EomesFlox/Flox, Vav-Cre+ mice lacks TRAIL+ and DX5+ NK cells, respectively. The Eomes-dependent phenotypic properties of mature, Eomes+DX5+TRAIL−Tbet+ NK cells, including repression of TRAIL and expression of CD49b (DX5+), also require ongoing Eomes activity to retain this identity. One of the unresolved features of the present model is whether all mature, DX5+TRAIL− NK cells transit through a DX5−TRAIL+ intermediate stage or whether the Eomes-dependent DX5+ state is an alternative lineage to the T-bet-dependent TRAIL+ state.
Despite TRAIL re-induction and loss of CD49b, the Ly49 repertoire appears to remain fixed after the loss of Eomes. Chromatin modifications and DNA methylation changes may mark the activity of Ly49 family members (Rouhi et al., 2006; 2007). The apparent heritability of Ly49 maintenance, together with the detection of Ly49 family expression following proliferation of EomesFlox/Flox, Vav-Cre+ TRAIL+ NK cells in lymphopenic mice and the prior suggestion that Ly49 acquisition precedes CD49b expression and cell division in the bone marrow (Kim et al., 2002), suggest an epigenetic basis for the action of Eomes at the Ly49 locus. Ly49 genes may be poised for induction prior to acquiring Eomes, but Eomes-dependent modifications, direct or indirect, may be essential to establish heritable Ly49 gene activity that can persist after loss of Eomes.
Why the fetal and neonatal NK repertoire consists predominantly of immature NK cells remains unresolved. The paucity of Ly49 receptors on early neonatal NK cells (Kubota et al., 1999) can now be linked to the restriction on Eomes induction in the hepatic milieu. Future experiments involving enforced Eomes expression within the immature, TRAIL+ stage will, thus, be informative for determining whether there is an undesirable consequence of maturation of the Ly49 repertoire in early development. Activation or tolerization of fetal NK cells through exposure to maternal components missing paternal MHC molecules might be reasons that repertoire selection of NK cells is delayed until after birth (Kumar et al., 1997; Tripathy et al., 2008). Humans and mice are highly susceptible to devastating, often fatal infection with herpesviruses in the perinatal period, owing partly to deficiency in NK cell-mediated antiviral responses (Anzivino et al., 2009; Bukowski et al., 1985). If more mature or more abundant NK cells are not catastrophic to fetus or mother, future studies may be directed toward mobilizing NK cells to clear viral pathogens in utero and in newborns.
The medullary and splenic microenvironments seem to drive NK expression of Eomes, while the liver appears non-permissive for Eomes induction. The micro-environmental signals upstream of Eomes induction, however, remain to be elucidated. We observed derepression of Eomes in steady-state T-bet-deficient NK cells and in NK cells that acutely lost T-bet. It is possible that T-bet directly or indirectly plays a role in repressing expression of Eomes. Lowest amounts of T-bet appear to be found in NK cells of the bone marrow, which may be mechanistically linked to the observation that TRAIL+ NK cells acquire Eomes most readily in the bone marrow following adoptive transfer.
Eomes is required for mature NK cells to express a diverse repertoire of Ly49 receptors, including Ly49H, which is essential to mediate resistance against MCMV infection (Arase et al., 2002; Lee et al., 2001). While Eomes+DX5+ NK cells are likely to be essential for resistance to some infections, it remains to be determined whether TRAIL+ NK cells serve unique functions in other aspects of host defense. Apart from their obvious differences in Ly49 repertoire, TRAIL+ and DX5+ NK cells express some similar but some dissimilar effector components. Future studies will be directed toward determining whether TRAIL+ and DX5+ NK cells kill targets in the same manner, whether both subsets are capable of memory responses, and whether effector-memory functions are dependent on T-bet, Eomes, or both (Cooper et al., 2009; Paust et al., 2010; Sun et al., 2009).
The transcription factors Eomes and T-bet are highly homologous and have both redundant and non-redundant roles in CD8+ T cell differentiation and function (Banerjee et al., 2010; Gordon et al., 2011; Intlekofer et al., 2005; 2007; 2008; Kinjyo et al., 2010; Pearce et al., 2003; Weinreich et al., 2010). Their overlapping and non-overlapping target genes have not yet been completely defined. The present study reveals that these two factors are molecular determinants that control transit through unique NK cell maturational checkpoints. Understanding how expression of T-bet and Eomes is regulated and how they, in turn, regulate distinct target genes should offer new avenues to customize CD8+ T and NK cellular therapy against infection and cancer.
Mice were housed in specific pathogen-free conditions and used in accordance with University of Pennsylvania Institutional Animal Care and Use Guidelines. C57BL/6 EomesFlox/Flox, Tbx21−/− (Intlekofer et al., 2005; 2008), and Vav1-Cre+ mice (Stadtfeld, 2004) were bred to obtain EomesFlox/Flox, Vav-Cre+ mice and EomesFlox/Flox,Tbx21−/−, Vav-Cre+ mice, used between 4–16w of age.
Livers were harvested by dissecting the inferior vena cava and aorta. Liver was perfused thoroughly with ice-cold PBS + 2mM EDTA + 1% fetal bovine serum. Once no visible blood remained, the liver was harvested, and liver lymphocytes were obtained by mechanical dissociation of the liver and passage through a 70-micron strainer. The liver was resuspended in 40% Percoll and underlaid with 60% Percoll (GE Healthcare). Lymphocytes were isolated at the interphase following centrifugation. Hepatic and other lymphocytes were stained with LIVE/DEAD fixable dead cell stain kit (Invitrogen), prior to staining with antibodies. Cells were analyzed on an LSR II coupled to FACSDiva software (BD Biosciences) and data were analyzed with FlowJo software (Treestar, Inc.). Sources of mAbs are provided in the Supplemental Experimental Procedures. CXCL16-Fc (Matloubian et al., 2000) was detected by a biotin-conjugated, anti-human Fc antibody (Jackson Immunoresearch).
Wild-type (CD45.1+) plus EomesFlox/Flox, Vav-Cre+ (CD45.2+) or wild-type (CD45.1+) plus Tbx21−/− (CD45.2+) FACS-purified blood progenitors (Lineage−c-kithighSca-1high) were used as donor cells. 1×103 donor cells were injected i.v. into sublethally-irradiated (450 rads) Il2rg−/−Rag2−/− recipients (Similar results were observed by transplanting 1×107 unsorted bone marrow cells). Organs of recipients were analyzed 6–10wk post transplant.
For MCMV infection, mice were infected with 2.5×107 pfu i.p. To analyze degranulation, bulk cells were stimulated with 5 ng/mL PMA and 50 ng/mL Ionomycin, in the presence of Brefeldin A (Golgiplug, BD), monensin (Golgistop, BD), and anti-CD107a. To analyze cytokine production, cells were stimulated with 5 ng/mL PMA and 50 ng/mL Ionomycin, in the presence of Brefeldin A.
Hepatic or medullary lymphocytes were isolated and depleted of T, B, granulocyte, and erythroid lineages with mAbs against CD4, CD5, CD8, CD19, Gr-1, and Ter119 (Biolegend). Lineage− cells were stained with mAbs against CD3ε, CD49b (DX5), NK1.1, NKp46, and TRAIL. TRAIL+DX5− and TRAIL−DX5+ NK cells were sorted on a FACSAria II (BD). Purified TRAIL+ or DX5+ NK donor cells were injected i.v. into separate, unirradiated Il2rg−/−Rag2−/− recipients. For transfer of hepatic TRAIL+ NK cells, 1×105 hepatic TRAIL+ NK cells or 1×105 hepatic DX5+ NK cells were transferred to recipient mice. The TRAIL+ NK cells for transfer were all hepatic. The DX5+ NK cells for transfer were 50% splenic and 50% hepatic. For transfer of medullary TRAIL+ NK cells, at least 1×104 of the following FACS-purified, medullary NK cell subsets were transferred into immunodeficient recipients: TRAIL+ CD27lowCD11blow, TRAIL+ CD27highCD11blow, DX5+ CD27highCD11blow, and DX5+ CD27high/lowCD11bhigh. Post-sort purity of transferred cells was at least 93%.
Splenic (for Eomes deletion experiments) or hepatic (for T-bet deletion experiments) lymphocytes were isolated and depleted of T, B, granulocyte, and erythroid lineages. Lineage− cells were stained with mAbs against CD3ε, CD122, NK1.1, and NKp46. 1×106 NK cells were sorted on a FACSAria II and washed thoroughly in serum free medium prior to addition of TAT-Cre (50μg/mL) or medium alone (sham). NK cells were then incubated for 45min at 37C. The reaction was quenched by adding media containing 20% FBS, followed by thorough washing. Treated or sham-treated donor cells were injected intravenously into separate, unirradiated Il2rg−/−Rag2−/− recipients. For in vitro temporal deletion studies, splenic or hepatic lymphocytes were isolated and depleted of T, B, granulocyte, and erythroid lineages, but NK cells were not FACS-purified prior to treatment with either TAT-Cre or medium alone. 1×107 splenic or 1×106 hepatic lymphocytes entered the TAT-Cre reaction. After the reaction was quenched and cells were washed, they were cultured separately in 2000 U/ml rIL-2 for 5d. To analyze function of acutely double-deficient NK cells in vitro, cells were stimulated after 5d culture with 5 PMA /Ionomycin for 2h to purge preformed granules and cytokines. Cells were then stimulated with 5 ng/mL PMA and 50 ng/mL Ionomycin for 2h in the presence of Brefeldin A, monensin, and anti-CD107a.
Splenic and hepatic lymphocytes were isolated and depleted of T, B, granulocyte, and erythroid lineages. Lineage− cells were stained with mAbs against CD3ε, CD49b (DX5), NK1.1, NKp46, and TRAIL. TRAIL+ and DX5+ NK cells were sorted on a FACSAria II. RNA was isolated with the miRNeasy kit (Qiagen), and cDNA was synthesized with the SuperScript VILO cDNA synthesis kit (Invitrogen). Q-PCR reactions were carried out with a TaqMan 7900HT machine and TaqMan gene expression assays (Applied Biosystems). Relative amplification values were calculated by normalizing to amplification of hypoxanthine guanine phosphoribosyl transferase (HPRT).
Statistics were calculated with Prism software (GraphPad Software). To test for significant differences in the number of NK cells between 2 groups of adult mice, a Student’s t test with a two-tailed P value was performed. Differences in numbers of NK precursors among the 4 genotypes were analyzed with a one-way ANOVA with a Tukey’s post-comparison test. Numbers of neonatal NK cells among aged-matched wild-type and Tbx21−/− mice were analyzed with a paired t test. Statistical significance was reached at p<0.05.
We thank M. Matloubian for providing the CXCL16-Fc reagent; and A. Bhandoola, T. Kambayashi, J. Maltzman, J. Orange, J. Wherry, M. Paley, M. Ciocca, and B. Barnett for assistance and discussion. Supported by NIH grants AI042370, AI076458, and AI061699 and the Abramson Family (S.L.R.) and NIH grant NCI T32 CA09140 (S.M.G).
The authors declare no competing financial interests or other conflicts of interest.
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.