SHIP1 Intrinsically Regulates NK Cell Signaling and Education Resulting in Tolerance of a MHC-I Mismatched BM Graft in Mice

Abstract

Natural killer (NK) cells are an important component of host immune defense against malignancy and infection. NK cells are educated by MHC class I ligands in order to ensure self-tolerance while also promoting lytic competency against altered-self and damaged-self targets. However, the intracellular molecular events that culminate in tolerance and functional competency of the educated NK cells remain undefined. Mice with germline deficiency in SHIP1 were shown to have a defective NK cell compartment. However, SHIP1 is expressed in all hematopoietic lineages and consequently several hematolymphoid phenotypes have already been identified in certain cell types that are the result of SHIP1 deficiency in cells in separate and distinct lineages – cell extrinsic phenotypes. Thus, it was previously impossible to determine the NK cell intrinsic role of SHIP1. Herein, through the creation of the first NK cell specific deletion of SHIP1, we show that SHIP1 plays a profound NK lineage intrinsic role in NK cell homeostasis, development, education and cytokine production and is required for in vivo mismatched bone marrow (BM) allograft rejection as well as for NK memory responses to hapten.

Introduction

Natural killer (NK) cells are an innate lymphocyte cell that through production of cytokines and direct target cell lysis are important in host defense against viral infection and malignancy. (1) At a cellular level, NK cell activation is regulated by the balance of signaling from germline-encoded activating and inhibitory receptors. (2) Activating receptors recognize endogenous danger molecules and pathogen encoded ligands, whereas inhibitory receptors recognize self-molecules such as MHC class I (MHC-I). (2) To prevent auto-reactivity NK cells undergo an education process that prevents activation of an NK cell that lacks expression of an inhibitory receptor capable of host MHC-I recognition. Inhibitory receptor binding of self MHC-I ligands is thought to endow NK cells with functional competency, which is considered arming or licensing. (3, 4) Others have proposed that NK cells which fail to express MHC-I binding inhibitory receptors experience chronic activation and thus become disarmed. (5) Moreover, there is evidence that both arming and disarming could regulate NK cells in distinct processes. (6) The molecular features of both arming and disarming also remain uncharacterized. Thus, although a great deal has been learned about the receptors and ligands that determine the regulation of NK cell activation and education, there is a significant deficit in our understanding of the intracellular events that culminate in NK cell education, licensing and disarming.

NK cells have recently been shown to possess memory capacity to a range of stimuli including memory responses to CMV, (7) to haptens (8) and viral particles (9). The NK cells responsible for the memory response to haptens and viral particles reside in the liver and are not renewed from adult hematopoietic stem cells (HSC) in the bone marrow (BM). (10) This liver memory NK cell population appears to be a unique lineage of NK cells which express CXCR6 (9), Thy1.2 and Ly49C/I (8) but are DX5⁻CD49a⁺ (10). NK cell memory to mCMV infection is mediated by a Ly49H⁺ splenic NK subset that requires the transcription factor Zbtb32 to regulate their mCMV-induced proliferation. Intriguingly, Zbtb32 is not required for maintenance of the hapten-specific memory NK cell subset. (11) In addition signaling through DNAM-1 and STAT4 is required for the generation of NK cell memory to mCMV. (12, 13) However, the role of these molecules in hapten and viral particle associated NK memory has not been defined.

Mice with germline deficiency in SH2 domain-containing inositol-5'-phosphatase 1 (SHIP1) have a severely defective NK cell compartment (reviewed in (14)). NK cells from these mice have a skewed natural killer cell receptor repertoire (NKRR), (15, 16) decreased IFNγ production following activation, (16) decreased killing of tumor targets (17) and an inability to reject MHC class-I (MHC-I) mismatched bone marrow allografts (15, 18). However, while SHIP1 appears to be required for natural cytotoxicity and IFNγ production in mice, SHIP1 may limit antibody dependent cellular cytotoxicity (ADCC), at least in human NK cells. (19, 20) It is presently unclear if NK cell defects in SHIP1 deficient mice are due to an intrinsic role of SHIP1 in NK cells or if the NK cell phenotype is due to the inflammatory cytokine millieu present in these mice (these mice develop a Crohn’s disease like phenotype and succumb to pneumonia typically within 8 weeks after birth), (21) or a requirement for SHIP1 expression in trans as SHIP1 expression is also required for the proper function of T cells (22, 23), B cells (24), regulatory T cell formation and homeostasis (25), dendritic cell function (26), myeloid derived suppressor cell homeostasis (26, 27), megakaryocyte progenitor cell formation (28), M2 macrophage homeostasis (29), basophil degranulation (30), hematopoietic niche cell function (31) and mesenchymal stem cell fate determination. (32)

To assess the intrinsic role of SHIP1 in NK cells we created the first NK cell conditional knockout of SHIP1. (33) Herein we show that SHIP1 plays a prominent and lineage intrinsic role in NK cell development, NKRR formation, cytokine production, NK cell hapten specific memory, NK cell education and acute bone marrow allograft rejection.

Material and Methods

Mice and genotyping

SHIP^flox/flox mice express normal levels of SHIP, but the SHIP proximal promoter and 1st exon are flanked by loxP recombination signales (floxed), such that SHIP expression is ablated when Cre recombinase is expressed in the cell. SHIP^flox/flox mice were originally created on a 129/Sv genetic background and have been backcrossed to C57BL/6 mice 11 times resulting in mice that are greater than 99.9% C57BL/6 (15). NKp46^iCre/+ transgenic mice have been previously described (34). Genotyping of Cre transgenic mice was performed by PCR using primers detecting the Cre sequence (P1, 5′-GGAACTGAAGGCAACTCCTG -3′; P2, 5′- CCCTAGGAATGCTCGTCAAG - 3′; P1, 5′-TTCCCGGCAACATAAAATAAA -3’). All animal experiments were approved by the SUNY Upstate Medical University Institutional Animal Care and Use Committees.

Ex vivo IFNγ production assay

Six well plates were coated overnight at 4°C with antibody, anti-NK1.1 (PK136), anti-NKp46 (29A1.4) or anti-NKG2D (A10) (eBioscience, San Diego, CA). After sacrifice splenocytes were harvested and red blood cells lysed using ACK lysis buffer (eBioscience, San Diego, CA). 6 × 10⁶ cells were incubated either in an antibody coated, unstimulated or PMA and ionomycin treated well for five hours in the presence of GolgiPlug (BD Biosciences, San Diego, CA). Following incubation Fc receptors were blocked with anti-CD16/CD32 (2.4G2) antibody (BD Biosciences, San Diego, CA). Invitrogen fixable aqua live/dead stain was used according to manufacturer’s instructions for dead cell exclusion. NK1.1 (PK136) (for all stimulation conditions other than samples stimulated with anti-NK1.1 antibody) or NKp46 (for samples stimulated with anti-NK1.1 antibody) and CD3ε (145-2C11) surface markers were stained (eBioscience, San Diego, CA). Cells were fixed and permeabilized using BD Cytofix/Cytoperm™ kit (BD Biosciences, San Diego, CA) according to manufactures instructions. Fc receptors were blocked with anti-CD16/CD32 (2.4G2) antibody. Intracellular staining was performed using an anti-IFNγ (XMG1.2) antibody (eBioscience, San Diego, CA). Cells were analyzed using a BD LSRFortessa cell analyzer (BD Biosciences, San Diego, CA)

Acute BM rejection assay

NKp46^iCre/+SHIP1^flox/flox and SHIP1^flox/flox control mice were irradiated twice, 3–4 hours apart, receiving 550 rads both times (RS2000, Rad Source, Suwanee, GA). 2×10⁶ congenic, B6.SJL, and 2×10⁶ MHC class-I mismatched BALB/C (Taconic, Hudson, NY) BM cells were injected into the retro orbital sinus of each host. Seven days later mice were bled and red blood cells were lysed using ACK lysis buffer (eBioscience, San Diego, CA). Fc receptors were blocked with anti-CD16/CD32 (2.4G2) antibody (BD Biosciences, San Diego, CA). CD45.1 (A20), CD45.2 (104) and H2D^d (34-2-12) surface markers were stained (BD Biosciences, San Diego, CA). Cells were stained with DAPI for dead cell exclusion. Cells were analyzed using a BD LSRFortessa cell analyzer (BD Biosciences, San Diego, CA)

Flow cytometry

NKp46^iCre/+SHIP1^flox/flox and SHIP1^flox/flox control mice were sacrificed by CO₂ toxicity and splenocytes or hepatocytes were harvested. In the case of experiments using hepatocytes blood was flushed from the liver by ligation of the portal vein and perfusion of the inferior vena cava with at least 20mLs of cold PBS. Lymphocytes were isolated by density gradient centrifugation using a 37.5% Percoll solution. Red blood cells were lysed using ACK lysis buffer (eBioscience, San Diego, CA). Fc receptors were blocked with an anti-CD16/CD32 (2.4G2) antibody (BD Biosciences, San Diego, CA). Invitrogen fixable aqua live/dead stain was used according to manufactures instructions for dead cell exclusion. Surface markers were stained using NK1.1 (PK136), CD3ε (145-2C11), CD11b (M1/70), Ly49A (A1), Ly49G2 (4D11), Ly49I (YLI90), Ly49F (HBF-719), NKG2D (CX5), NKp46 (29A1.4), CD244.2 (2B4), Ly49D (4E5), Ly49H (3D10), CD27 (LG.3A10), NKG2AB6 (16a11), CD49a (Ha31/8), CD94 (18D3), DNAM-1 (10E5), KLRG1 (2F1), CD90 (30-H12), anti-PI (3,4,5)P₃ IgM (Echelon, Logan, Utah), pS2448 mTOR (O21-404) and pS473 Akt (M89-61). All antibodies were purchased from BD Biosciences, BioLegend, eBioscience (San Diego, CA) or from R&D Systems (Minneapolis, Minnasota). Ly49C (4LO3311) staining was performed using supernatant purchased from University of California San Franscisco and revealed using an Allophycocyanin-AffiniPure Goat Anti-Mouse IgG, Fcγ Subclass 3 Specific secondary antibody purchased from Jackson ImmunoResearch (West Grove, PA). Fixation and permeabilization was performed using the BD biosciences Cytofix/Cytoperm kit and cells were analyzed using a BD LSRFortessa cell analyzer (BD Biosciences, San Diego, CA). In all cases flow cytometry plots are shown after backgating for singlets, viable cells and when applicable, lymphocytes.

Contact hypersensitivity to hapten

As described by O’Leary et al., shaved abdominal skin of NKp46^iCre/+SHIP1^flox/flox mice and SHIP1^flox/flox controls was primed by painting with 50µL of 5% oxazolone dissolved in 1:1 methanol:acetone on days 0 and 1. One month later baseline ear thickness was measured using a micrometer to the first cartilage ridge. One ear was painted with solvent only to account for any acute solvent induced swelling and the other was painted with 20µL of 1% oxazolone. 24 hours later the thickness of both ears was measured. To account for acute hapten induced ear swelling ears of naïve mice were painted and subsequent swelling was measured at 24 hrs. Ear swelling was calculated as follows: ([thickness of hapten challenged ear at 24 hrs] – [thickness of hapten challenged ear at baseline]) – ([thickness of solvent challenged ear at 24 hrs] – [thickness of solvent challenged ear at baseline]) - ([thickness of non-primed mouse ear at 24 hrs] – [thickness of non-primed mouse ear at baseline]).

Statistical analysis

All statistical analyses were performed using the statistical software Prism (GraphPad, San Diego, CA). * indicates p<0.05, ** indicates P<0.01 and *** indicates p<0.001.

RESULTS

SHIP1 is required for normal NK cell maturation and homeostasis

To determine if defective NK cell behavior observed in mice with germline SHIP1 deficiency is due to an NK cell intrinsic requirement for SHIP1 expression rather than due to the general inflammatory milieu present in these mice we created the first NK conditional deletion of SHIP1. To do this we used NKp46-iCre mice (mice with an improved Cre recombinase transgene in the NKp46 locus) on a SHIP1^flox/flox C57BL6 background (15) to specifically delete SHIP1 in NK cells. NKp46-iCre mice have been described previously (33) and were shown to have NK cells that develop normally in all tissues, express a normal NKRR, produce cytokines normally and kill target cells as efficiently as wild type (WT) NK cells.

We did not observe any decrease in longevity or gross pathology in NKp46^iCre/+SHIP1^flox/flox mice. This differs dramatically from germline SHIP1^−/− mice that succumb to pneumonia and/or Crohn’s like disease by 6–10 weeks of life. (26, 35, 36) To confirm that SHIP1 is deleted specifically in NK cells we performed intracellular flow cytometry for SHIP1 and found that both T and B cells from NKp46^iCre/+SHIP1^flox/flox mice express normal levels of SHIP1 and that SHIP1 deficiency is indeed confined to NK cells (Fig. 1a–c).

Figure 1. NKp46^iCre/+SHIP1^flox/flox mice do not express SHIP1 selectively in NK cells.

Splenocytes harvested from NKp46^iCre/+SHIP1^flox/flox mice (black histogram) and SHIP1^flox/flox controls (grey histogram) were stained with anti-NK1.1, anti-CD3ε and anti-CD19 antibodies and, following fixation and permeabilization, were stained with either anti-SHIP1 antibody or isotype control (filled histogram) showing deletion of SHIP1 in (A) NK1.1⁺CD3ε⁻CD19⁻ viable NK cells but not in (B) CD19⁺CD3ε⁻NK1.1⁻ viable B cells or in (C) CD3ε⁺NK1.1⁻CD19⁻ viable T cells. These flow cytometry plots are representative of at least 3 NKp46^iCre/+SHIP1^flox/flox mice and 3 SHIP1^flox/flox controls with similar results.

To begin to analyze the effects of SHIP1 deficiency on NK cells we harvested splenocytes from NKp46^iCre/+SHIP1^flox/flox mice and SHIP1^flox/flox controls and performed flow cytometry to determine if SHIP1 is required for normal NK cell development and homeostasis. As shown in Figure 2a,b NKp46^iCre/+SHIP1^flox/flox mice have a modest, but significantly decreased frequency and absolute number of splenic NK cells indicating that SHIP1 is intrinsically required for regulation of NK cell frequency. Furthermore, terminal maturation of NK cells is also dependent on SHIP1 expression. Terminal maturation of peripheral NK cells is determined by coordinate analysis of CD27 and CD11b surface expression. (37) Inmature NK cells are CD27⁺CD11b⁻ (referred to in this paper as R1), whilst NK cell maturity is demarcated by loss of CD27 expression and acquisition of CD11b expression (CD27⁻CD11b⁺ cells; R3). NK cells pass through an intermediate stage (R2) where they express both CD27 and CD11b. Again, NKp46^iCre/+SHIP1^flox/flox mice have a modest, but significantly decreased, frequency of mature CD27⁻CD11b⁺ splenic NK cells and significantly increased numbers of CD27⁺CD11b⁻ splenic NK cells (Fig. 2c,d) indicating an NK cell intrinsic role for SHIP1 in promoting efficient NK cell maturation in the periphery.

Figure 2. NK cell expression of SHIP1 is required for NK cell development.

(A and B) Splenocytes harvested from NKp46^iCre/+SHIP1^flox/flox mice and SHIP1^flox/flox controls were stained with anti-NK1.1 and anti-CD3ε to determine (A) frequency of NK1.1⁺CD3ε⁻ viable NK cells amongst splenic lymphocytes and (B) absolute number of splenic NK cells. NK cell terminal maturation was assessed by coordinate analysis of anti-CD27 and anti-CD11b staining. Graphs summarize pooled data from at least four independent experiments including at least 3 NKp46^iCre/+SHIP1^flox/flox mice and 3 SHIP1^flox/flox controls in each experiment. One representative flow plot showing CD27 vs. CD11b staining from a NKp46^iCre/+SHIP1^flox/flox mouse and a SHIP1^flox/flox control is shown (C) on gated singlet, viable, lymphocyte, NK1.1⁺CD3ε⁻ populations. (D) Box and whisker plots indicating percentage of CD27⁺CD11b⁻ (R1), CD27⁺CD11b⁺ (R2) and CD27⁻CD11b⁺ (R3) NK cell populations summarize data from at least 3 NKp46^iCre/+SHIP1^flox/flox mice and 3 SHIP1^flox/flox controls and has been repeated in at least two other independent experiments.

SHIP1 deficient NK cell have a highly disrupted MHC-I receptor repertoire

We next hypothesized that SHIP1 is required for normal natural killer receptor repertoire (NKRR) expression as was first reported in germline SHIP^−/− mice. (15) To test this hypothesis we implemented a multicolor flow cytometry panel to determine the frequency of NK cells expressing a multitude of different NK cell receptors. We found a normal frequency of NK cells expressing most activating receptors analyzed including Ly49D, Ly49H, NKp46, NKG2D and CD94 and the SLAM family receptor 2B4 (Fig. 3a), most of which are dramatically reduced in germline SHIP^−/− mice. (15, 16) This attests to the power of conditional mutation analysis and indicates that SHIP1 may have a role in trans in regulation of the NKRR. We used Boolean gating to fully characterize inhibitory Ly49 receptor expression amongst 32 different NK cell subsets in a similar manner to that of Brodin et al. (38) Using this technique we found significantly increased frequency of NK cells expressing the MHC class-I binding inhibitory receptors Ly49A, Ly49F, Ly49G2 and Ly49C and significantly decreased frequency of NK cells expressing Ly49I (Fig. 3b). Interestingly, we observed an increase in the number of Ly49 inhibitory receptors expressed per NK cell and a decreased frequency of NK cells that either do not express an inhibitory Ly49 receptor or express only a single Ly49 receptor (Fig. 3c,e) which follows a repertoire disruption similar to that observed in β2m^−/− mice that have an improperly educated NK compartment owing to greatly diminished surface expression of MHC class I. (39, 40)

Figure 3. NK cell expression of SHIP1 is required for normal NKRR expression.

(A) Splenocytes harvested from NKp46^iCre/+SHIP1^flox/flox mice and SHIP1^flox/flox controls were stained with anti-Ly49H, anti-Ly49D, anti-CD94, anti-NKp46, anti-NKG2D, anti-2B4 and anti-DNAM-1 or (B) anti-Ly49A, anti-Ly49C, anti-Ly49F, anti-Ly49G2 and anti-Ly49I antibodies and analyzed by flow cytometry. All cells were also stained with anti-NK1.1, anti-CD3ε and a viability dye and the frequency of NK cells expressing the given receptor were determined after backgating for singlets, live cells, lymphocytes and NK1.1⁺CD3ε⁻ NK cells. (C) Boolean gating was used to determine the number of inhibitory, MHC-I binding Ly49 receptors expressed by each NK cell harvested from NKp46^iCre/+SHIP1^flox/flox mice and SHIP1^flox/flox controls. The results are expressed as the Log₂ ratio between these frequencies. (D) Splenocytes harvested from NKp46^iCre/+SHIP1^flox/flox mice and SHIP1^flox/flox controls were stained with anti-NK1.1, anti-CD3ε, anti-CD27, anti-CD11b and anti-KLRG1 antibodies and analyzed by flow cytometry. KLRG1 expression is shown amongst NK1.1⁺CD3ε⁻ NK cells as well as amongst R3 (NK1.1⁺CD3ε⁻CD27⁻CD11b⁺) mature NK cells. (E) The frequencies of 32 subsets of NK cells expressing different combinations of inhibitory Ly49 receptors, expression is indicated by a filled box, was determined using Boolean gating scheme in both NKp46^iCre/+SHIP1^flox/flox mice and SHIP1^flox/flox controls. The results are expressed as the Log₂ ratio between these frequencies. The graphs represent pooled data from at least 7 NKp46^iCre/+SHIP1^flox/flox mice and 7 SHIP1^flox/flox controls collected in at least two independent experiments.

Importantly, surface Killer cell lectin-like receptor G1 (KLRG1) expression is significantly decreased in NKp46^iCre/+SHIP1^flox/flox mice compared to SHIP1^flox/flox controls (Fig. 3d). KLRG1 is an NK cell inhibitory receptor capable of binding cadherin molecules (E, N, R) (41) and is preferentially expressed on mature NK cells. (33) The decrease in KLRG1 expression is not simply due to decreased CD11b⁺CD27⁻ mature NK cells in our mice as when KLRG1 expression is analyzed after backgating on the most mature (R3) NK cell population KLRG1 expression remains decreased (Fig. 3d) KLRG1 expression is significantly increased on NK cells following infection (42) whereas KLRG1 expression is significantly decreased on NK cells from β2m^−/− mice (39, 43) and from C57BL/6 mice with K^bD^b deficiency (44) and thus its expression is thought to demarcate educated NK cells (3). Thus, decreased DNAM-1 and KLRG1 surface expression and the increased frequency of NK cells expressing multiple inhibitory Ly49 receptors indicate an uneducated NK cell compartment is present in NKp46^iCre/+SHIP1^flox/flox mice.

Deficiency of SHIP1 results in an intracellular signaling profile consistent with an activated NK cell

We next wanted to determine if SHIP1 is required for regulation of signaling pathways important for NK cell activation. Given that SHIP1 catalyzes the hydrolysis of PI (3, 4, 5)P₃ to PI (3,4)P₂ and thus functions as a negative regulator of the PI3K pathway, we hypothesized that NK cells from NKp46^iCre/+SHIP1^flox/flox would have increased activation of several different signaling pathways. We found using intracellular flow cytometry that NK cells, but not T cells, from NKp46^iCre/+SHIP1^flox/flox mice had significantly increased levels of PI (3, 4, 5)P₃ and of phosphorylated Akt (p-Akt S473) and mTOR (p-mTOR S2448) (Fig. 4a–c). The Akt/mTOR signaling pathway promotes NK cell activation and survival. Thus, the intracellular signaling profile of NKp46^iCre/+SHIP1^flox/flox NK cells is consistent with a state of sustained activation even during normal homeostasis.

Figure 4. SHIP1 deficient NK cells have increased phosphorylation of activating signaling pathways.

Splenocytes harvested from NKp46^iCre/+SHIP1^flox/flox mice (black histograms) and SHIP1^flox/flox controls (grey histograms) were stained with anti-NK1.1, anti-CD3ε and a viability dye. Following fixation and permeabilization cells were stained with (A) anti-PI (3, 4, 5)P₃ (B) anti-pS473 Akt, and (C) anti-pS2448 mTOR antibodies and analyzed by flow cytometry. Graphs are expressed as the median fluorescence intensity (MFI) (A, B and C) after backgating on singlets, live cells, lymphocytes and either NK1.1⁺CD3ε⁻ NK cells or NK1.1⁻CD3ε⁺ T cells. Graphs summarize data from 4 NKp46^iCre/+SHIP1^flox/flox mice and 4 SHIP1^flox/flox controls and similar data has been gathered in at least two independent experiments.

SHIP1-deficient NK cells are hyporesponsive to ligands for major NK activating receptors

Due to sustained activation state of NK cells in healthy, unmanipulated NKp46^iCre/+SHIP1^flox/flox mice we hypothesized that they might adopt a hyporesponsive or disarmed state, consistent with findings in mouse models where NK cells are chronically stimulated. (6, 39, 45) To assess this possibility, we measured IFNγ production from NK cells via intracellular flow cytometry following activation of NK cells using plate bound antibodies to crosslink NK1.1, NKp46 or NKG2D: three major NK activating receptors for cell-surface ligands. In each case, significantly fewer NK cells from NKp46^iCre/+SHIP1^flox/flox mice were able to produce IFNγ after activation compared to SHIP1^flox/flox controls (Fig. 5a,b). However, we did not observe a significant impairment for NKp46^iCre/+SHIP1^flox/flox NK cells stimulated with phorbol 12-myristate 13-acetate (PMA) and ionomycin (Iono) indicating SHIP1 deficiency in NK cells does not result in a distal signaling defect that impairs IFNγ production. NKG2D and NK1.1 surface expression are not significantly different between NKp46^iCre/+SHIP1^flox/flox and SHIP1^flox/flox controls, and while NKp46 surface expression is decreased in NKp46^iCre/+ mice (Fig. 5c), NKp46^iCre/+ NK cells produce IFNγ comparably to NKp46^+/+ NK cells after ex vivo activation with an anti-NKp46 antibody. (33) Thus, despite the normal representation in the SHIP1-deficient NK compartment and normal surface expression of NKG2D and NK1.1, crosslinking of three independent activating receptors results in defective activation of a key NK effector function.

Figure 5. SHIP1 is required for IFNy production.

(A) Splenocytes harvested from NKp46^iCre/+SHIP1^flox/flox mice and SHIP1^flox/flox controls were incubated on uncoated plates, plates coated with anti-NK1.1, anti-NKp46 or anti-NKG2D antibody, or uncoated plates but stimulated with PMA and ionomycin. The frequency of IFNγ production amongst NK1.1⁺CD3ε⁻ (for all samples not stimulated with anti-NK1.1 antibody) or NKp46⁺CD3ε⁻ (for samples stimulated with anti-NK1.1 antibody) was determined using flow cytometry with representative flow plots shown. (B) The graph summarizes pooled data from at least data from 8 NKp46^iCre/+SHIP1^flox/flox mice and 8 SHIP1^flox/flox controls gathered in at least two independent experiments. (C) Splenocytes were stained with anti-NK1.1, anti-CD3ε, anti-NKp46 and anti-NKG2D antibodies. Surface expression of the indicated activating receptor was determined after backgating on singlets, viable lymphocytes and NK1.1⁺CD3ε⁻ NK cells.

SHIP1 is required for NK cell mediated rejection of mismatched allogeneic bone marrow

To further test the functional competency of the NK cell compartment in NKp46^iCre/+SHIP1^flox/flox mice we asked whether these mice could efficiently reject an MHC-I mismatched BM allograft. This assay provides an in vivo measure of NK cytolytic function, and importantly also allows us to determine if SHIP1 expression is an intrinsic requirement for proper NK cell effector function in vivo, a question that cannot be addressed using chimeras consisting of WT and mutant BM. In this assay we transplanted equal numbers of congenic C57BL/6 SJL (CD45.1, H2^b; cells that should not be rejected) and MHC-I mismatched BALB/C (H2^d; cells that are rejected by normal, healthy mice) bone marrow cells into lethally irradiated NKp46^iCre/+SHIP1^flox/flox mice and SHIP1^flox/flox controls. One week after transplantation flow cytometric analysis of peripheral blood samples showed a significantly decreased frequency of MHC-I mismatched (H2^d) donor cells in SHIP1^flox/flox control mice as compared to mice with SHIP1 deficient NK cells (Fig. 6a,b). This shows for the first time that NK lineage expression of SHIP1 is required for acute rejection of an MHC-I unmatched bone marrow graft.

Figure 6. SHIP1 is required for NK mediated acute bone marrow graft rejection.

Lethally irradiated NKp46^iCre/+SHIP1^flox/flox mice and SHIP1^flox/flox controls received 2 million BL/6.SJL (CD45.1 congenic MHC-H2^b) and 2 million BALB/C (H2^d) bone marrow cells. (A) Flow cytometry performed after one week on peripheral blood samples comparing the frequency of viable MHC-H2D^d positive cells versus CD45.1 positive MHC-I matched bone marrow cells. (B) The ratio of MHC-H2^d+ cells compared to CD45.1⁺ cells is indicated in the graph which summarizes data from 4 NKp46^iCre/+SHIP1^flox/flox mice and 4 SHIP1^flox/flox controls and is representative of two independent experiments. Acute rejection” is calculated by [frequency of allogenic cells (H2D^d+)]/[frequency of congenic cells (CD45.1⁺)]

NK lineage expression of SHIP1 is required for hapten induced contact hypersensitivity

NK cells are capable of a mounting a memory response to certain viruses (7), viral particles (9) and to haptens (8). The NK cells responsible for memory response to hapten (contact hypersensitivity) are located in the liver, express Thy-1, require CXCR6 signaling and are DX5⁻ but are CD49a⁺ and express either Ly49C or Ly49I. (8–10) As noted above, while NKp46^iCre/+SHIP1^flox/flox mice do have a modest, but significant, reduction in NK1.1⁺CD3ε⁻ NK cell numbers. Importantly, while NKp46^iCre/+SHIP1^flox/flox mice do have significantly decreased frequency of NK1.1⁺CD3ε⁻ hepatic NK cells (Fig. 7a), these mice do not have any significant difference in hepatic memory NK cell frequency (NK1.1⁺CD3ε⁻DX5⁻CD49a⁺Thy1.2⁺Ly49C/I⁺) compared to SHIP1^flox/flox controls (Fig. 7b) or absolute number of hepatic memory NK cell numbers (Fig. 7c). This led us to determine that if these mice are able to mount a memory response to hapten any functional deficit in this assay cannot be attributed to decreased NK cell frequency and must be due to a requirement for SHIP1. To determine whether SHIP1 is required for NK memory responses to hapten, we primed NKp46^iCre/+SHIP1^flox/flox mice and SHIP1^flox/flox controls by painting shaved abdominal skin with oxazolone. One month later NKp46^iCre/+SHIP1^flox/flox mice had significantly less ear swelling after challenge with oxazolone as compared to SHIP^flox/flox controls (Fig. 7d) indicating that NK intrinsic expression of SHIP1 is required for a robust NK memory response that can elicit hapten induced contact hypersensitivity.

Figure 7. SHIP1 is required for memory response to hapten induced delayed type contact hypersensitivity.

Mice were sacrificed via CO₂ toxicity and livers were rapidly perfused with cold PBS to remove peripheral blood lymphocytes prior to liver resection. Lymphocytes were isolated from liver samples from NKp46^iCre/+SHIP1^flox/flox mice and SHIP1^flox/flox controls via Percoll density gradient centrifugation. Lymphocytes were stained with anti-NK1.1, anti-CD3ε, anti-Thy1.2, anti-DX5, anti-CD49a, anti-Ly49C (4LO3311 mAb) and anti-Ly49I (YLI-90 mAb) antibodies to determine memory NK cell frequency. (A) NK1.1⁺CD3ε⁻ hepatic NK cell and (B) NK1.1⁺CD3ε⁻Thy1.2⁺DX5⁻CD49a⁺Ly49C/I⁺ memory NK cell frequency amongst hepatic lymphocytes was determined with DAPI staining used for dead cell exclusion. Absolute NK1.1⁺CD3ε⁻Thy1.2⁺DX5⁻CD49a⁺Ly49C/I⁺ memory NK cell numbers are given in (C). The graphs summarize pooled data from 7 NKp46^iCre/+SHIP1^flox/flox mice and 8 SHIP1^flox/flox controls gathered during two independent experiments. (B) Shaved abdominal skin of NKp46^iCre/+SHIP1^flox/flox mice and SHIP1^flox/flox controls was primed by painting with oxazolone. Thirty days later one ear was challenged with solvent while the other with oxazolone. 24 hours following oxazolone challenged ear swelling was measured while accounting for acute swelling due to solvent induced irritation (see further details in the Methods). The graph represents pooled data from two independent experiments of at least 5 NKp46^iCre/+SHIP1^flox/flox mice and 5 SHIP1^flox/flox controls.

Discussion

In this study we used the first mouse model that offers NK cell specific deletion of SHIP1 to determine its role in NK cell development, education and effector function. Herein we show that NK cell intrinsic expression of SHIP1 is required for NK cell homeostasis, development, and receptor repertoire formation. Additionally, we have shown that SHIP1 is required for NK cell activation and education, thus providing the first evidence of a molecular signaling component required for these key NK cell milestones.

Previous studies examining the role of SHIP1 in NK cells have been performed using mice with germline deficiency in SHIP1. These studies have shown that mice require SHIP1 for normal NK cell homeostasis, normal IFNγ production and for rejection of a mismatched MHC-I BM allograft. Mice with germline deficiency of SHIP1 were shown to have increased numbers of NK cells (15) and BM chimeras made from WT and SHIP1 germline deficient bone marrow have normal numbers of NK cells (46) thus there is a role for SHIP1 in trans in limiting NK cell numbers, but NK cell intrinsic expression of SHIP1 is in fact required for normal NK cell homeostasis as our mice have a slight decrease in splenic NK cell frequency.

SHIP1 has also been shown to regulate NKRR expression. Mice with germline SHIP1 deficiency on a mixed S129/C57BL/6 genetic background were originally shown to have increased frequencies of Ly49I and LyC/I (using the 5E6 mAb) expressing NK cells and decreased frequencies of NK cells expressing Ly49G2, CD94 and Ly49D. (15) Germline SHIP1^−/− NK cells on a BL/6 background were later shown to express nearly all receptors analyzed at a decreased frequency including inhibitory receptors capable of binding MHC-I molecules Ly49A, Ly49C/I, Ly49F, Ly49G2, Ly49I, activating receptors capable of binding MHC-I molecules including Ly49D, and other activating receptors such as NKG2D, Ly49H and NKp46. (16) In the current study we show that, with the exception of DNAM-1, NK cell expression of SHIP1 is dispensable for regulation of expression of activation receptors and of inhibitory receptors that bind non-classical MHC-I molecules. However, we show here that SHIP1 is required to limit frequencies of most inhibitory MHC-I binding Ly49 receptors (all except Ly49I) and we observed an increased number of different Ly49 receptors expressed by individual NK cells in NKp46^iCre/+SHIP1^flox/flox mice as compared to SHIP^flox/flox controls. This pattern of receptor expression is also seen in mice with β2m deficiency. β2m is required for stable surface expression of MHC-I molecules and thus mice with β2m deficiency do not have educated NK cells because there is no ligand for Ly49 binding. Thus, this pattern of Ly49 expression indicates that SHIP1 is required for NK cell education.

A liver resident subset of NK cells was recently shown to mediate delayed type contact hypersensitivity in an antigen specific manner. (8) Specific signaling molecules required by these cells have only begun to be elucidated. Initial antigen specific proliferation to mCMV requires the transcription factor zbtb32 (11) and memory NK cells to mCMV require STAT4 signaling and DNAM-1 (12, 13), but signaling molecules and events required for NK cell memory to haptens have not been identified. For the first time our studies reveal a critical signaling component required for NK cell memory to hapten. NK cells have recently been shown to possess the capacity to form memory to the mouse herpes virus mCMV. (7) Formation of this type of NK cell memory could be a distinct molecular process and thus it is presently unclear if SHIP1 is required for formation of the NK cell memory response to mCMV.

Previous studies have indicated at least two different cellular mechanisms for NK cell education. First, NK cells must undergo an “arming” process where an inhibitory receptor expressed by the NK cell must bind to host MHC-I molecules in order to become functionally competent. (3) Secondly, NK cells experiencing chronic activation are “disarmed” or rendered hyporesponsive. (39) Several studies have shown that disarming appears to act as the dominant NK cell education process. (45, 47) We have shown that in the absence of SHIP1, NK cells have increased activation of the PI3K/Akt/mTOR axis. However, despite the increased activation of pathways important to NK cell activation and memory formation, NKp46^iCre/+SHIP1^flox/flox NK cells have defective effector functions indicating that chronic stimulation results in hyporesponsive NK cells similar to the disarmed NK cells from transgenic mice with constitutive expression of an activating ligand. (6, 48) NKp46^iCre/+SHIP1^flox/flox mice are healthy and do not express increased levels of activating ligands. Recently, Viant et al. described mice with NK cell conditional deletion of the tyrosine phosphatase SHP-1. (49) Following ligation of an NK cell inhibitor receptor, SHP-1 is recruited to the immunoreceptor tyrosine-based inhibition motif (ITIM) of that receptor (50) and attenuates NK cell activation. (49) NK cells from NCR1^iCre/+SHP-1^flox/flox mice have a very similar phenotype to NCR1^iCre/+SHIP^flox/flox NK cells. Both types of mice have increased frequency of inhibitory Ly49 expression by NK cells but do not have alteration in frequency or expression of the activating receptors Ly49D, Ly49H or NKG2D. Additionally, SHP-1 was shown to be required for NK cell terminal maturation in the periphery, and NK cells from NCR1^iCre/+SHP-1^flox/flox were hyporesponsive both in vitro and in vivo. Unlike SHIP1 deficient NK cells, SHP-1 deficient NK cell do not have increased activation of kinases such as Akt and mTOR but instead are unable to attenuate NK activation after ligation of an inhibitory receptor. Thus, NK cell deficiency of two different phosphatases, SHP-1, which is needed for inhibitory receptor signaling, and SHIP-1, which is needed to attenuate basal signaling from activating receptors during normal homeostasis, both result in hyporesponsiveness due to chronic activation.

It is possible that the in vivo functional defects reported above in NKp46^iCre/+SHIP1^flox/flox mice are attributable to the modestly reductions we observe in splenic NK cell frequency or frequency of CD11b⁺CD27⁻ mature NK cells. We believe this to be unlikely as the maturation and NK cell homeostasis defects are slight and thus are not sufficient to compromise the NK compartment as a whole. Further, NKp46^iCre/+SHIP1^flox/flox NK cells were hyporeactive in vitro favoring NK cell hyporeactivity as the explanation for in vivo functional defects we elucidate here. These findings represent the first molecular component that is essential to prevent basal signaling and thus chronic activation of NK cells and their disarming.