Influence of SHIP on the NK Repertoire and Allogeneic Bone Marrow Transplantation

See allHide authors and affiliations

Science  15 Mar 2002:
Vol. 295, Issue 5562, pp. 2094-2097
DOI: 10.1126/science.1068438


Natural killer cell (NK) receptors for major histocompatibility complex (MHC) class I influence engraftment and graft-versus-tumor effects after allogeneic bone marrow transplantation. We find that SH2-containing inositol phosphatase (SHIP) influences the repertoire of NK receptors. In adult SHIP−/− mice, the NK compartment is dominated by cells that express two inhibitory receptors capable of binding either self or allogeneic MHC ligands. This promiscuous repertoire has significant functional consequences, because SHIP−/− mice fail to reject fully mismatched allogeneic marrow grafts and show enhanced survival after such transplants. Thus, SHIP plays an important role in two processes that limit the success of allogeneic marrow transplantation: graft rejection and graft-versus-host disease.

Certain intracellular phosphatases contain SH2 domains that enable their recruitment to phosphorylated tyrosine residues in the cytoplasmic tails of inhibitory receptors (1). Following recruitment to the plasma membrane, these enzymes remove phosphate groups on other proteins or inositol phospholipids and thus counteract signaling cascades necessary for cell survival, proliferation or differentiation. SHIP is one of these signaling phosphatases (2–4). SHIP has been shown to limit the number of myeloid cells produced in vivo demonstrating it as a crucial mediator of survival signals in a hematopoietic lineage (5,6).

In order to examine the role SHIP plays in NK development and function, we generated mice with a targeted mutation in the SHIP gene resulting in SHIP-deficient mice (7). Analysis of the peripheral NK compartment at different stages of ontogeny (7) indicated NK cells develop normally in juvenile SHIP−/− mice (Fig. 1A). However, in adult mice an abnormal population of NK cells appears that expresses approximately 10-fold higher surface levels of the NK receptor, NK1.1 (NK1.1hi) (Fig. 1A). The NK1.1hi population lacks CD3 and thus is not an NK-T cell population. The appearance of the NK1.1hi population coupled with an increase in NK cells with a normal 2B4+NK1.1+ staining profile (NK1.1+cells) leads to a two- to threefold increase in peripheral NK cells in SHIP−/− adult mice (≥8 weeks), relative to wild-type littermates (Fig. 1B).

Figure 1

Increased NK cell numbers in SHIP−/− mice due to enhanced survival. (A) FACS analysis of splenic NK cells in SHIP+/+ and SHIP−/− littermates. Genotype and age of the mice at the time of sacrifice and analysis are indicated. (B) Absolute splenic NK cell numbers in SHIP+/+ and SHIP−/− mice at different ages. The values determined for SHIP−/− mice that are significantly different from that of their age-matched SHIP+/+ counterparts are indicated by the following symbols: *, P < 0.05 and †, P < 0.01. (C) Percentage of BrdU labeled splenic NK cells in SHIP+/+ (n = 3) or SHIP−/− (n = 3) mice immediately following 1 week on BrdU drinking water (Pulse) or after a 2-week chase (Pulse+Chase). P-values: **P = 0.2 for SHIP−/− versus SHIP+/+ in the “Pulse” analysis, *P = 0.05 for SHIP−/− versus SHIP+/+ in the “Pulse+Chase” analysis. Representative histograms for BrdU labeling of NK1.1+CD3 cells from +/+ and −/− animals in the Pulse+Chase group are shown.

We then examined whether the increased number of NK cells present in SHIP−/− mice is due to increased proliferation or due to enhanced survival (7). When SHIP−/−and SHIP+/+ mice were labeled for 1 week with BrdU and their NK compartments were analyzed, we found no significant difference in BrdU uptake (Fig. 1C) indicating NK cells in both genotypes have similar rates of production and proliferation. However, when mice were pulsed with BrdU for 1 week and the mice analyzed 2 weeks later, a significantly larger proportion of the SHIP−/− NK compartment retained the BrdU label (Fig. 1C), demonstrating that SHIP−/− NK cells survive longer in vivo than do their SHIP+/+ counterparts.

Murine NK cells detect MHC class I molecules using receptors encoded by the Ly49 or CD94/NKG2 genes (8). Expression of these MHC receptors is distributed among different NK subsets during the transition from neonate to adult. Because the number of peripheral NK cells increases in SHIP−/− mice during this period, we asked whether the relative representation of NK subsets expressing certain Ly49 and CD94 receptors might account for this increase (7). Indeed, the relative representation of several Ly49 receptors and CD94 was significantly altered in the SHIP−/− NK compartment of older mice when compared to SHIP+/+ littermates (Fig. 2) (7). However, SHIP−/− weanlings showed no skewing of their NK repertoire, relative to wild-type littermates (Fig. 2) (7). This repertoire distortion was most pronounced in mice ≥8 weeks of age and was found in both the NK1.1+ and the NK1.1hi populations. We found that expression of Ly49A+ and C/I+ was overrepresented in adult SHIP-/- mice, while Ly49D, G2 and CD94 were underrepresented (7). Because the overwhelming majority of the NK1.1+ and NK1.1hicell populations lacked Ly49I in adult SHIP−/− mice (7), then the majority of the Ly49C/I+NK cells express only Ly49C. Thus, the repertoire distortion in adult SHIP−/− mice leads to an NK compartment dominated by a subset of cells with the following repertoire: Ly49A+C+DG2ICD94. In vitro and in vivo studies indicated that Ly49C and Ly49A can bind ligands in the H2b haplotype of SHIP−/− mice; however, these two receptors also bind and transmit inhibitory signals from ligands in most or all H2 haplotypes (9–12). Therefore, SHIP deficiency leads to an NK inhibitory repertoire that is both self-specific and promiscuous for other ligands.

Figure 2

MHC class I receptors on peripheral NK cells in SHIP−/− mice. The mean percentage of peripheral NK cells expressing the indicated Ly49 or CD94 molecule after gating on 2B4+NK1.1+ cells. The age and genotype of the mice are indicated. Values determined for SHIP−/−mice that are significantly different from their age-matched SHIP+/+ littermates are indicated: *, P < 0.05; †, P < 0.01.

A potential explanation for the repertoire disruption seen in SHIP−/− NK cells is that SHIP is recruited to certain inhibitory receptors expressed by NK cells to oppose intracellular signals that mediate survival of specific NK subsets expressing these receptors. Indeed, SHIP binds the phosphorylated ITIM motif of Ly49A in vitro (13). These findings prompted us to examine whether SHIP associates in vivo with Ly49 receptors expressed by NK cells (7). This analysis revealed that SHIP is associated with Ly49A and Ly49C under physiological conditions (Fig. 3A), but not Ly49G2, Ly49F, or Ly49I (Fig. 3B). As further confirmation that the protein coprecipitating with Ly49A and Ly49C is SHIP, we analyzed NK lysates from SHIP+/+ and SHIP−/− mice (Fig. 3C). This analysis detected coprecipitation of SHIP with Ly49A and Ly49C only in the SHIP+/+ NK lysates, confirming the coprecipitating protein is SHIP.

Figure 3

SHIP is recruited to NK inhibitory receptors in vivo to oppose activation of Akt. (A) Western blot detection of SHIP in Ly49A and Ly49C immunoprecipitates. Immunoprecipitation with a murine IgG2a antibody (IgG2a) was analyzed as a negative control and SHIP was immunoprecipitated as a positive control (SHIP). (B) Western blotting for SHIP in other Ly49 immunoprecipitates (Ly49G2, Ly49F, and Ly49I). (C) Western blot analysis of SHIP in Ly49A and Ly49C immunoprecipitates prepared from lysates of SHIP+/+ (+/+) and SHIP−/−(−/−) NK cells. (D) Western blot analysis of Akt phosphorylation at Thr308 and total Akt protein in SHIP+/+ and SHIP−/− NK cell lysates. To control for equal loading the blot was re-probed with antibodies specific for β-actin, GAPDH, and α-tubulin.

Because SHIP limits the in vivo survival of myeloid cells by opposing the PI3K/Akt pathway (3, 4), we examined whether Akt is activated in SHIP−/− NK cells in vivo based on its phosphorylation at Thr308(14). We found that both Akt phosphorylation and total Akt protein levels were significantly increased in SHIP−/− NK cells relative to those in SHIP+/+ NK cells (Fig. 3D) (7). The increase in total Akt levels is surprising; however, primary B cell activation leads to increased Btk levels in a PI3K-dependent manner (15). This additional level of regulation may further amplify signals from PH domain-containing kinases (Akt, Btk) that are recruited to PI(3,4,5)P3. Consistent with the activation of the PI3K/Akt pathway in SHIP−/− NK cells, we find thatBcl-2 levels were increased in SHIP−/− NK cells and the p85 subunit of PI3K is recruited to Ly49A and Ly49C/I (16). Taken together, these findings suggest the interplay of SHIP and PI3K may influence the relative survival of NK subsets expressing MHC class I receptors capable of recruiting these enzymes. Interestingly, PI3K is recruited to human KIR and can activate Akt in human NK cells (17). Thus, despite their evolutionary divergence in how they bind MHC class I, murine Ly49 receptors and human KIR may recruit SHIP to limit the in vivo survival of NK subsets, just as both receptors recruit SHP-1 to limit NK effector functions (18, 19).

We speculated that the distortion of the NK repertoire toward inhibitory receptors with ligands in most MHC haplotypes might hamper responses to allogeneic targets in SHIP−/− mice. Indeed, transgenic mice with enforced expression of Ly49A are unable to reject allogeneic bone marrow (BM) grafts from H2d donors (20). Thus, we examined whether SHIP−/− mice on an H2b background could acutely reject fully H2 mismatched BM grafts (7). We first analyzed whether SHIP−/−mice reject fully allogeneic BM grafts from BALB/C mice whose H2d haplotype forms strong interactions with both Ly49A and Ly49C. Consistent with our hypothesis, we observed that BALB/C BM was not rejected by SHIP−/− mice, whereas their wild-type littermates could reject equivalent grafts (Fig. 4A). However, the failure of SHIP−/− mice to reject H2d grafts might also have been due to underrepresentation of the activating receptor, Ly49D, which also binds H2d. To rule out this possibility, we tested the ability of SHIP−/− hosts to reject BM grafts from A/SW(H2s)/Sn mice whose H2s haplotype possesses inhibitory ligands for both Ly49A and Ly49C (9), but not activating ligands for Ly49D. Consistent with our proposed mechanism, SHIP−/−, but not SHIP+/+ hosts, failed to reject H2s marrow grafts (Fig. 4B). In addition, pretreatment of mice with F(ab′)2 antibody fragments that block the Ly49C receptor prior to transplant partially restored the ability of SHIP−/− hosts to reject BALB/C BM grafts (Fig. 4C). These results demonstrate that overrepresentation of an inhibitory receptor contributes directly to the compromised ability of SHIP−/− hosts to reject allogeneic BM grafts. To exclude the possibility that the NK compartment was simply impaired, we examined the ability of SHIP−/− mice to reject MHC class I–deficient BM (7) as compared to their wild-type littermates (Fig. 4D). SHIP−/− mice showed complete rejection of β2m−/− BM grafts comparable to that seen in SHIP+/+ littermates. Thus, the NK compartment in SHIP−/− hosts was not broadly disabled.

Figure 4

Allogeneic BM engrafts SHIP−/− hosts in the absence of GVHD. (A andB) Analysis of allogeneic BM graft rejection in SHIP−/− hosts. Growth of H2d (BALB/C) (A) and H2s [A/SW(H-2s)/Sn] (B) BM in SHIP+/+ or SHIP−/− hosts as measured by splenic uptake of125I-dUrd. In (A), *P = 0.0001 for SHIP−/− versus SHIP+/+, and **P = 0.0633 for SHIP−/− versus syngeneic control. In (B), *P = 0.0006 for SHIP−/− versus SHIP+/+ and **P = 0.002 for SHIP−/− versus syngeneic control. (C) Growth of BALB/C donor BM in SHIP+/+ or SHIP−/− (H2b) hosts after pretreatment with blocking antibodies against Ly49C receptors [F(ab′)2] or with 0.2 ml of normal mouse serum (NMS) as a source of irrelevant mouse IgG. P-values: F(ab′)2-treated SHIP−/− versus NMS-treated SHIP−/−,P = 0.0476; F(ab′)2-treated SHIP+/+ versus NMS-treated SHIP+/+, P = 0.4654; Syngeneic control versus NMS-treated SHIP−/−, P = 0.4000; NMS-treated SHIP−/− versus NMS-treated SHIP+/+, P = 0.0179; F(ab′)2-treated SHIP−/− versus F(ab′)2-treated SHIP+/+,P = 0.0022. (D) Growth of β2m−/− BM in SHIP+/+ (C57BL6/J) or SHIP−/− (C57BL6/J) hosts, *P = 0.2894 for SHIP−/− versus SHIP+/+, **P = 0.0001 for SHIP−/− versus syngeneic control (D- donor; H- host; T- treatment). (E) Survival plots for SHIP−/− and SHIP+/+ recipients receiving fully H2 mismatched BM grafts.

To determine whether engraftment of MHC mismatched BM could lead to severe graft-versus-host disease (GVHD), we transplanted a cohort of SHIP−/− mice and their SHIP+/+ littermates with BM from BALB/C mice (7). The large majority of SHIP−/− mice survived transplant without developing GVHD, whereas less than half of the SHIP+/+ mice survived (Fig. 4E) (7). These findings suggest a previously unappreciated role for host NK cells in the initiation of GVHD. Potentially, SHIP−/− NK cells fail to produce inflammatory cytokines (γ-IFN, TNF-α) in response to allogeneic BM cells, thereby reducing the likelihood of a significant GVH reaction by donor T cells. Alternatively, other host cell types that contribute to GVHD, such as antigen-presenting cells (21), could also be altered by SHIP deficiency.

Although Ly49 inhibitory receptors prevent inappropriate killing by NK cells, the interaction of these receptors with self MHC ligands may also elicit signals that promote the survival or proliferation of these cells in vivo (22). SHIP may counteract these signals and thus prevent the expansion of NK subsets expressing more than one self-restricted inhibitory receptor. We propose that inhibiting SHIP activity prior to BM transplant could restrict the NK inhibitory repertoire, such that selecting a donor with an appropriate MHC ligand, or ligands, might enable engraftment in the absence of GVHD. Thus, inhibition of SHIP signaling should be explored as a means to increase both the efficacy and utility of allogeneic BM transplantation.

  • * These authors contributed equally to this work.

  • To whom correspondence should be addressed. E-mail: kerrw{at}


Stay Connected to Science

Navigate This Article