Natural Killer Cells Require Selectins for Suppression of Subcutaneous Tumors Free

Primary tumor growth is a complex process, involving many interactions between the tumor and surrounding tissue. A developing tumor influences and is influenced by its stroma, initiates angiogenesis, and interacts with both the adaptive and innate immune systems. The concept of the immune system surveying the body for nascent tumors and suppressing tumor growth dates to the beginning of the 20th century. Whereas the immune surveillance hypothesis has had a turbulent history (1, 2), it is now generally accepted that the immune system plays an important role in controlling cancer progression (1, 3–5).

Innate immune cells clearly participate in immune surveillance. There is clinical and experimental evidence to implicate macrophages (6, 7), mast cells (8, 9), eosinophils (10), dendritic cells (11), and natural killer (NK) cells (12, 13) in tumor interactions. Whereas many innate immune cells have both tumor-promoting and tumor-suppressive functions (14, 15), NK cells are generally considered tumor suppressive (1, 16, 17). Through a variety of positive and negative receptors, NK cells can recognize (2, 16) and eliminate tumor cells by several mechanisms (16, 18). In addition to their well-known cytotoxic functions, NK cells are also gaining recognition as key regulators of the adaptive immune response. Several recent studies indicate that NK cells can activate dendritic cells (19, 20) and influence T-cell differentiation (20, 21). NK cell–related cancer immunotherapies are being actively investigated (22, 23).

Although much progress has been made in understanding the processes of NK cell target recognition, target killing, and cross-talk with other immune cells, the cell adhesion mechanisms that mediate NK cell traffic to the tumor are still unclear (24, 25). NK cells are not tissue-resident cells in most organs. However, NK cells need contact with tumor cells to recognize and eliminate them (16). Therefore, cell adhesion mechanisms that would allow NK cells to exit the circulation and enter into tissues must play an important role in NK-mediated tumor surveillance.

Classic immune cell recruitment requires chemokines, integrins, and selectins. Selectins are members of the C-type lectin family of cell adhesion molecules. E-selectin is expressed on activated endothelium; P-selectin is rapidly translocated to the cell surface from intracellular storage granules in activated endothelium and in platelets; and L-selectin is constitutively expressed on leukocytes. Selectins mediate initial rolling of leukocytes along activated endothelium, as has been shown for neutrophils, monocytes, and T cells (26–28). The rolling step allows for activation of leukocyte integrins, leading to firm leukocyte adhesion to endothelium and subsequent extravasation into tissue (27, 28). Selectins are critical for immune cell trafficking, as revealed by defective leukocyte rolling and recruitment to lymphoid organs and to sites of inflammation in selectin-deficient mice (29–32). It is known that NK cells express L-selectin and selectin ligands and can use them for rolling in in vitro studies (24, 25). However, there is little prior work to indicate whether selectin-dependent recruitment mechanisms are important for NK cell traffic in vivo.

Previous experiments from our laboratory showed that human tumors injected s.c. into Rag2^−/− mice grow significantly larger in the absence than in the presence of selectins (33). This phenotype is largely transferable with bone marrow transplant. These results implicated selectins and the innate immune system in tumor immune surveillance. Because selectins are known to play a role in immune cell traffic, we hypothesized that enhanced tumor growth in the absence of selectins may be due to defective migration of tumor-suppressing immune cells.

In this article, we have extended the previous xenograft tumor model to immunocompetent C57BL/6 mice. In concordance with previous work on human cell lines, we found that several syngeneic murine tumor cell lines grow significantly better in selectin-deficient mice. We explored leukocyte traffic in selectin-deficient mice and discovered that NK cell recruitment to tumors is impaired in the absence of selectins. NK cells isolated from selectin-deficient mice appear otherwise normal and functional. In mice depleted of NK cells, either pharmacologically by antibody injection or genetically in NK-deficient GrzA-Ly49A transgenic mice (34), tumor growth is significantly enhanced. Increased tumor growth in the absence of NK cells is not augmented further by the absence of both NK cells and selectins, arguing that selectins and NK cells act in the same pathway to suppress tumor growth. Furthermore, the ability of NK cells to clear tumors in selectin-deficient mice is defective. These results suggest that NK cells act to suppress tumor growth in our system and depend on selectins to do so.

Mice. Mice missing all combinations of selectins were created in our laboratory (29, 31, 32) and back-crossed for eight to nine generations onto C57BL/6, BALB/c, or 129S4 backgrounds. Mice are referred to by the selectins they are missing: L^−/− for L-selectin knockouts and ELP^−/− for triple selectin knockouts. Immune-deficient mice missing the recombination activating gene 2 (Rag2^−/−) on C57BL/6 background were obtained from the laboratory of Dr. Chen [Massachusetts Institute of Technology (MIT), Cambridge, MA; ref. 35]. Mice in which NK cell development is disrupted by the insertion of the Ly49A NK receptor under the control of the granzyme A promoter (GrzA-Ly49Atg) backcrossed to the C57BL/6 background were obtained from the laboratory of Dr. Yokoyama (Washington University, St. Louis, MO; ref. 34).

Subcutaneous tumors. For tumor growth assays, B16F0 melanoma, LL/2 Lewis lung carcinoma, and CT26 adenocarcinoma cells were obtained from American Type Culture Collection (ATCC). MC38GFP colon carcinoma cells were obtained from the laboratory of Dr. Varki (University of California San Diego, La Jolla, CA; ref. 36). B16-puro and B16-puroRae1ε variants were generated by infecting B16F0 cells with lentiviral vectors containing appropriate constructs (empty vector or vector containing Rae1ε, respectively). The cells were infected and selected in parallel; Rae1ε expression was validated by fluorescence-activated cell sorting (FACS). The Rae1ε construct was a kind gift from Dr. Raulet (University of California Berkeley, Berkeley, CA). All tumor cell lines were cultured in DMEM, 10% heat-inactivated fetal bovine serum (FBS), glutamine, and penicillin/streptomycin (100× stock from Gibco-Invitrogen). For in vivo experiments, anesthetized 8- to 12-week-old mice were injected s.c. in the flank with 1 × 10⁴ tumor cells per mouse in HBSS (Gibco-Invitrogen) or in 50% HBSS/50% growth factor–reduced Matrigel (BD Biosciences) for Matrigel experiments. For time-course experiments, mice were monitored every 2 to 3 d by palpation for tumor presence at the injection site. Tumor surface area was calculated by multiplying the length of the tumor by its width. For end-point experiments, tumors were removed and weighed 21 d after tumor cell injection.

Immune cell purification from Matrigel. Matrigel “plugs” were removed on days 1, 3, 5, 7, and 10 after tumor injection; four to seven mice of each genotype (C57BL/6, L^−/−, ELP^−/−) were analyzed at every time point. The plug was mechanically dissociated and incubated with agitation at 37°C for 2 h in 0.2% autodigested collagenase I (Worthington Biochemical Corporation) in PBS. Digested samples were strained through a 70-μm mesh sieve. Larger pieces were mashed through the sieve with the rubber-capped end of a syringe plunger. A 100-μL sample was counted by hemocytometer to establish total cell number in the Matrigel plug. Recovered cells were pelleted by centrifugation and analyzed by FACS.

Flow cytometry. Cells harvested from Matrigel plugs or from mouse tissues were analyzed on a four-color FACSCalibur (BD Biosciences) flow cytometer supported by FlowJo software. Antibodies used were against CD3, CD4, CD8, NK1.1, Gr1, CD107a, NKG2D, NKG2A/C/E, Ly49C/I/F/H, Ly49A/D, MEL14 (L-selectin, CD62L), PSGL-1 (CD162), DX5 (α₂ integrin, CD49b), LFA-1 (α_L integrin, CD11a), Mac-1 (α_M integrin, CD11b), CD11c (α_X integrin), α_V integrin, α₅ integrin, and isotype controls (BD Biosciences). Pan-specific anti-mRae1 antibody was obtained from R&D Systems.

In vitro cytotoxicity assay. Yac-1 lymphoma cells were obtained from ATCC and cultured in RPMI medium/10% FBS, l-glutamine, and penicillin/streptomycin (100× stock from Gibco-Invitrogen). NK cells were isolated by positive selection with anti-DX5 beads (Miltenyi Biotech), plated at ∼1 × 10⁶ per well, and activated overnight in complete RPMI medium supplemented with 125 units/mL interleukin (IL)-2 (BD Biosciences) and 75 ng/mL IL-15 (Sigma). Activated NK cells were coincubated at varying effector/target ratios with target cells (1 × 10⁴ Yac-1 cells per well; 2 × 10³ RMA or LL/2 tumor cells per well) for 4 h. Target cell killing was assayed using the CytoTox Non-Radioactive Cytotoxicity Assay (Promega). In L-selectin blocking experiments, NK cells were pretreated with the L-selectin blocking antibody MEL14 or an isotype control antibody (BD Biosciences), which were present throughout the assay.

NK cell IFNγ production and degranulation. Purified splenic NK cells were left in culture medium (RPMI + 5% FBS) alone for 4 or 30 h, stimulated with LL/2 tumor cells, or stimulated with NK1.1 activating antibody or an isotype control antibody for 6 h as described previously (37). IFNγ secretion was assayed using the DuoSet ELISA Development System for IFNγ (R&D Systems). In some cases, NK cells were recovered after 6-h stimulation and analyzed by FACS for surface CD107a expression to assess degranulation in response to activating stimuli.

NK cell antibody depletion. TM-β1 NK-depleting antibody (against mouse IL-2 receptor β; ref. 38) and RR4-7 control antibody (against human T-cell receptor variant) were purified in the laboratory from hybridoma supernatants. Hybridoma cells were provided by Dr. Tanaka (Osaka University, Osaka, Japan) and Dr. Eisen (MIT, Cambridge, MA), respectively. For NK cell depletion experiments, antibodies were injected i.p. at 1 mg/mouse in 500 μL of sterile PBS. Antibody injection was done once, 3 d before tumor cell injection. The efficiency of NK cell depletion was monitored throughout the experiment by FACS analysis of small blood samples collected from the tail vein of antibody-treated mice.

Statistical analysis. All data have been subjected to statistical analysis. For graphic representations, data are presented as mean values for at least three individual experiments, with error bars representing SE. For determining the statistical significance of the differences in the data, two-tailed Student's t test analysis with the assumption of unequal variance was used. In cases where tumor incidence varied, statistical significance was additionally assayed by a χ² test. All statistically significant (P < 0.05) differences in the data are indicated by asterisks in the figures.

LL/2 tumor growth is enhanced in selectin-null mice. To investigate the mechanism of selectin-dependent tumor suppression in the context of an intact immune system, we extended our previous studies to fully murine models using inbred mouse strains and s.c. injection of syngeneic tumor cell lines. Growth of MC38GFP, B16F0, and LL/2 cells syngeneic to C57BL/6, as well as CT26 cells syngeneic to BALB/c, was assayed in selectin-deficient and in control selectin-replete mice (Fig. 1). Tumors were analyzed at 3 weeks after injection. In all cases, tumors arose more frequently and grew larger in selectin-deficient mice than in matched selectin-replete controls (Fig. 1). In both C57BL/6 and in BALB/c backgrounds, the tumors were largest and most frequent in ELP^−/− mice. Using LL/2 cells, which showed the clearest difference in tumor growth between selectin-deficient and control mice, we further explored tumor growth in mice missing single selectins and in mice missing both of the endothelial selectins (PE^−/−). Of the single selectin knockouts, L^−/− mice were most similar in tumor growth to the triple knockouts, P^−/− mice and PE^−/− double knockouts showed a mild tumor enhancement, and E^−/− animals showed no tumor growth (Fig. 1D). These findings are consistent with previous data using human cell lines (33) and indicate that selectin-dependent tumor suppression is an important phenomenon for a variety of tumor cell lines in multiple mouse strains.

Selectin-deficient mice are defective in NK cell recruitment to tumors. Having established an immune-competent fully murine model, we proceeded to investigate the causes of the differences in tumor growth. Selectins are known to be involved in immune cell recruitment; we hypothesized that selectin-deficient mice fail to reject tumors because tumoricidal immune cells are unable to reach the tumor. To obtain samples that could be readily harvested and assayed for immune cell recruitment, tumor cells were injected as a suspension in Matrigel, a commercial preparation of mouse tumor matrix (39, 40). LL/2 cells embedded in growth factor–reduced Matrigel were injected into the right flank of selectin-replete or selectin-deficient mice, whereas Matrigel alone was injected into the left flank of the same mice. Matrigel plugs were removed from the mice over a course of 10 days following injection and analyzed by FACS for immune cell content.

Matrigel enhanced tumor growth in all mice, but differences in tumor size between selectin-replete and selectin-deficient mice were preserved (Fig. 2A). Analysis of total immune cell infiltration into Matrigel plugs showed that the overall immune response to tumor-containing Matrigel was significantly more robust than the immune response to Matrigel alone (Fig. 2B). Matrigel plugs were assayed for the presence of neutrophils, CD4⁺ T cells, CD8⁺ T cells, macrophages, dendritic cells, and NK cells. Neutrophil recruitment proved neither tumor specific nor significantly impaired by the absence of selectins. No significant differences in T-cell recruitment were detected between selectin-deficient and control mice. We observed some reduction in tumor-specific recruitment of macrophages and dendritic cells in selectin-deficient mice (Supplementary Fig. S1). However, because defects in macrophage and dendritic cell recruitment were only present either in ELP^−/− or in L^−/− mice but not in both, whereas the tumor growth increase was observed both in ELP^−/− and in L^−/− mice, these differences were not pursued further.

Analysis of immune cell content of tumors in Matrigel showed a clear and consistent defect in tumor-specific NK cell recruitment both in ELP^−/− and in L^−/− mice. Early NK cell recruitment to tumors was similar in all mice and independent of tumor cell presence in Matrigel plugs. In Matrigel plugs without tumor cells, NK cell numbers slowly declined over time in all mice. However, NK cell recruitment to tumor cell–containing Matrigel plugs increased steadily in selectin-replete mice, peaking on day 7 after tumor injection. In contrast, NK cell recruitment to tumors in ELP^−/− and L^−/− mice was markedly reduced, particularly at the peak of NK cell infiltration (Fig. 2C).

These data show that, first, NK cells are recruited to Matrigel plugs containing tumor cells, but not to Matrigel alone, confirming that NK cell recruitment in our system is not a general reaction to Matrigel nor to tissue injury caused by the injection, but is a tumor cell–specific response. Second, the results show that NK cell recruitment to tumors is defective in ELP^−/− and L^−/− mice. These data establish NK cells as good candidates for mediating selectin-dependent tumor suppression.

NK cell development is unimpaired in the absence of selectins. To ascertain that the defect in NK cell recruitment to tumors in selectin-deficient mice is not due to an overall NK cell deficit in these animals, we performed a comparative analysis of NK cells from selectin-replete and selectin-deficient mice.

First, we confirmed that NK cells isolated from spleens of C57BL/6 mice express L-selectin and the main ligand for all three selectins, PSGL-1, as has been reported (24, 41–43). We found that L-selectin is highly expressed on splenic NK cells, whereas PSGL-1 is expressed at low levels, and that expression levels of L-selectin and PSGL-1 are not altered by overnight NK cell activation with cytokines (Fig. 3A). To investigate NK cell development in selectin-deficient mice, we performed FACS analysis of blood, bone marrow, and lymphoid organs. NK cells were present in the blood and bone marrow of the ELP^−/− and L^−/− mice in normal numbers (Fig. 3B). NK cell numbers in the lymph nodes of L^−/− mice were reduced, whereas NK cell numbers in the spleens of ELP^−/− mice were increased, in keeping with known disturbances of leukocyte recirculation in these mouse strains (30, 32). NK cell receptor repertoire remained largely unaltered in the absence of selectins (Supplementary Fig. S2). We next assayed selectin-deficient and selectin-replete mice for changes in NK cell integrin expression that are known to correlate to NK cell developmental stages (44). As with NK cells from selectin-replete animals, integrin α_V was down-regulated on NK cells from selectin-deficient mice as the cells exited the bone marrow, whereas the expression of integrin α_Mβ₂, which is a marker of mature activated NK cells, was up-regulated (Fig. 3C).

Overall, we did not find any large defects in NK cell number or development in the absence of selectins. It is therefore unlikely that such disturbances could account for the defect in NK cell recruitment to tumors in selectin-deficient mice. However, we did confirm that NK cells normally express L-selectin and a major selectin ligand, PSGL-1, which could potentially contribute to NK cell recruitment to tumors.

Selectin-deficient NK cells retain tumoricidal and secretory functions. To explore whether the inability to reach the tumor was the only functional impairment of selectin-deficient NK cells, we assayed their ability to kill tumor targets and secrete cytokines in response to stimulation in vitro. NK cells from ELP^−/− and L^−/− mice were as efficient as control NK cells at destroying Yac-1 and RMA tumor targets (Fig. 4A and B). Similarly, when L-selectin was blocked acutely using the MEL14 antibody, no defects were observed in NK cell killing of LL/2 targets (Fig. 4C). NK cells are one of the major secretors of the tumor-suppressing cytokine IFNγ (16, 45). We tested NK cells isolated from spleens of selectin-deficient mice for their ability to produce IFNγ in response to stimulation. Selectin-deficient NK cells did not differ significantly from selectin-replete counterparts in IFNγ production in response to LL/2 tumor cell stimulation or to NK1.1 antibody cross-linking of an activating NK receptor (Fig. 4D,, left). Selectin-deficient NK cells also displayed enhanced degranulation in response to both activating stimuli (Fig. 4D , right). In summary, NK cells from selectin-deficient mice seem to be functional, apart from the in vivo defect in recruitment to tumors.

Subcutaneous tumor growth is enhanced in the absence of NK cells. To test whether loss of NK cells contributes to increased tumor growth, we removed NK cells by two independent methods. Mice were injected with a TM-β1 antibody that selectively depletes NK cells for up to 3 weeks (38) or with a control RR4-7 antibody of the same isotype that is unreactive to mouse antigens. In a parallel set of experiments, we used GrzA-Ly49A transgenic mice that genetically lack functional NK cells (34, 46). NK depletion by either method was confirmed by monitoring NK cell numbers in the blood (Fig. 5A and B). The antibody depletion was thorough and stable, whereas a few NK cells were detected in GrzA-Ly49Atg⁺ mice. However, these cells are reportedly immature and of low cytotoxic potential (34, 46). We then implanted subcutaneous LL/2 tumors into mice lacking NK cells. Depletion of NK cells, either by antibody or genetically, led to significantly enhanced subcutaneous tumor growth in NK-depleted animals as compared with matched controls (Fig. 5C).

We next wanted to know whether removing NK cells could explain the increased tumor growth in selectin-deficient mice. We therefore depleted NK cells in selectin-deficient mice and monitored tumor growth. Depletion of NK cells from L^−/− mice with the TM-β1 antibody did not result in any synergistic enhancement of tumor growth, nor did tumors grow better in ELP^−/− GrzA-Ly49Atg⁺ mice than in ELP^−/− animals (Fig. 5D). Therefore, it is likely that the defect in tumor rejection that we observed in selectin-deficient mice is due to a defect in NK cell recruitment. To test whether loss of adaptive immune cells also affects tumor growth in selectin-deficient mice, we repeated the LL/2 tumor growth experiments in Rag2^−/− and ELP^−/− Rag2^−/− mice. In contrast to our results with NK cell depletion, the absence of T cells, B cells, and NK T cells did not promote tumor growth in control mice and slightly decreased tumor incidence in selectin-deficient mice (Fig. 5E).

NK cell depletion had a similar effect on tumor growth to that seen with loss of selectins. Furthermore, depletion of NK cells from selectin-deficient mice had no additional effect on tumor growth, showing that NK cells are unable to participate in tumor rejection in selectin-null mice, whereas they do so in selectin-replete controls. This indicates that NK cells are necessary for tumor rejection and may require selectins to function in antitumor immunity. The adaptive immune system does not seem to participate in tumor rejection in our system.

NK-mediated tumor rejection in vivo requires selectins. To test whether selectins are required for NK cell–mediated tumor rejection, we used a tumor rejection model that is strongly dependent on NK cells. The B16 C57BL/6 melanoma cell line does not express NK cell ligands and is fairly resistant to NK cell killing. However, B16 cells transfected with Rae1, a ligand to an activating NK cell receptor, NKG2D, are rejected specifically by NK cells (12). To test NK cell function in vivo, we generated matched B16 cell lines: a control cell line (B16-puro) and a cell line stably expressing Rae1ε (B16-Rae1ε). Rae1ε expression in the matched cell lines was confirmed by FACS analysis (Fig. 6A).

When injected into control mice, B16-puro cells grew sizeable tumors in all animals, whereas B16-Rae1ε cells were completely rejected, as expected. In addition, as previously reported (12), rejection was NK cell specific because both cell lines gave rise to tumors in GrzA-Ly49Atg⁺ mice or in mice treated with the TM-β1 antibody (Fig. 6A). When injected into ELP^−/− mice, however, B16-Rae1ε tumors were not rejected (18 of 18 mice). These results show a profound defect in NK cell–mediated tumor rejection in selectin-deficient mice (Fig. 6B). Of the single selectin knockouts, L^−/− mice and P^−/− mice displayed partial defects in B16-Rae1ε tumor rejection, whereas E^−/− mice did not show a significant defect.

These data confirm that selectin-deficient mice have a clear defect in NK-mediated tumor rejection in vivo. They also show that both L-selectin and P-selectin contribute to NK cell antitumor function in vivo, whereas E-selectin does not contribute significantly.

In this study, we used subcutaneous tumor models to explore the role of selectins in tumor surveillance by innate immune cells. Existing evidence indicated that, in the absence of selectins, subcutaneous growth of human tumor cell lines in Rag2^−/− mice is enhanced, and that the innate immune system is likely responsible for selectin-dependent tumor suppression (33). Here we report increased growth of several syngeneic tumor cell lines of different origins in selectin-deficient mice, on several genetic backgrounds, confirming that selectin-dependent tumor suppression is a reproducible and important phenomenon. This phenotype was most apparent in the absence of all three selectins. Of the single selectin knockouts, L-selectin contributed most to tumor suppression, P-selectin contributed somewhat, and E-selectin did not contribute significantly. The predominant role of L-selectin in tumor suppression suggested that leukocytes are the key mediators of the observed phenotype.

Selectins play a role in immune cell recruitment to tissues and to lymphoid organs. Exploration of immune cell recruitment to Matrigel-embedded tumors in selectin-deficient mice led to several important findings. First, we established that all mice, regardless of the presence or absence of selectins, exhibited a tumor-specific immune response. Second, we found fewer macrophages, dendritic cells, and NK cells recruited to tumors in the absence of selectins. The defect in NK cell recruitment in the absence of all three selectins, or of L-selectin alone, was particularly striking. To explore the possibility that other aspects of NK cell biology are also affected in the absence of selectins, we confirmed that NK cells are present in selectin-deficient mice in normal numbers and seem to develop normally and to be functional in in vitro antitumor cytotoxicity and cytokine secretion assays. Because NK cells are known to be tumoricidal, appeared normal in the absence of selectins, but failed to infiltrate tumors in selectin-deficient mice, they became our top candidates for mediating selectin-dependent tumor suppression.

NK cells proved essential for tumor suppression in our system because growth of several s.c. injected tumor cell lines was enhanced in two different NK cell–deficient mouse models. In contrast, the adaptive immune system did not actively promote tumor immunity. Tumor growth in Rag2^−/− mice lacking B cells, T cells, and NK T cells was, if anything, slightly inhibited. The slight inhibition of tumor growth in Rag2^−/− animals could be due to several factors, including increased NK cell numbers in Rag2^−/− mice and lack of regulatory T cells, which could normally act to suppress tumor immunity and have recently been shown to suppress NK cell function (47). To test directly the function of NK cells in selectin-deficient mice, we used a subcutaneous tumor model wherein tumor rejection is strongly dependent on NK cells. Selectin-deficient mice showed a severe defect in NK cell–dependent elimination of tumor targets in vivo, establishing that selectins are necessary for this process.

Taken together, our data provide evidence that selectins are necessary for tumor elimination, that NK cells are necessary for tumor elimination, and that NK cells require selectins to clear tumor targets in vivo. We also found that, in the absence of selectins, NK cells are defective in entering the tumor. The simplest model suggested by these findings is that in our system, NK cells require selectins, particularly L-selectin, and selectin ligands to interact with tumor vasculature, enter the tumor, and promote its rejection. Our findings support and extend previous reports, in which it has been shown that in L-selectin–deficient mice, NK cell numbers around primary or metastatic tumors are reduced (48).

The work described here leaves some open questions. One important caveat to the proposed model is that it is unclear whether the defect in NK cell function observed in selectin-deficient mice is cell-intrinsic or partly attributable to defects in other selectin-dependent cell(s) that regulate NK cell recruitment or function. NK cells are not the only innate immune cell type affected by the absence of selectins because we also observed some defects in tumor-specific recruitment of macrophages and dendritic cells in selectin-deficient mice. Thus, NK cells and their recruitment are necessary but may not be sufficient to explain selectin-dependent tumor suppression.

In addition, it is worth noting that whereas tumor growth seemed to be strongly dependent on NK cells in our system, other immune cells may be more important in different tumor models. Several factors could influence the immune response to tumors in our model. First, we inject fully transformed tumor cells in a procedure that is concomitant with some tissue damage. A naturally arising tumor that proceeds through all stages of transformation in vivo may well interact differently with its microenvironment and with the immune system than our tumor model does. Second, tumor growth in our system is assayed on a time scale of 3 weeks, which is a relatively short period. NK cells are early-response immune cells; it is possible that in models wherein tumor growth is slower, the adaptive immune response would have time to develop and would contribute to tumor rejection. Lastly, the location and anatomic origin of the tumor may influence the immune response to that tumor. Nonetheless, our results clearly show an important contribution of selectins and NK cells to tumor rejection. It would therefore be interesting to address the role of NK cells and selectins in spontaneous tumor models.

Furthermore, selectins are likely not the only adhesion molecules that are required for NK cell migration to the tumor. Chemokines and chemokine receptors, as well as integrins, may also be necessary. Of the chemokines, CX₃C chemokine (fractalkine) and its receptor CX₃CR1 have been implicated in NK cell recruitment to tumors (49). Of the integrins, one publication associated α_Lβ₂ and α_Mβ₂ integrins with NK cell tumor migration (48). Previous work from our laboratory also implicates α_Vβ₃ and α_Vβ₅ integrins in tumor suppression (33). It will be interesting to investigate the contribution of other adhesion molecules to tumor surveillance by NK cells.

Lastly, L-selectin is also important for immune cell traffic to lymphoid organs. Our findings do not address this aspect of NK cell migration, but it has been reported that NK cells can use L-selectin to enter lymph nodes in cancer (50). Although the main proposed role for NK cells at the lymph node is in regulating the adaptive immune response, and the adaptive immune system does not seem to control tumor rejection in our system, we cannot rule out the possibility that defects in NK cell traffic to lymphoid organs may play a role in tumor overgrowth in selectin-deficient mice.

The results reported here contribute toward a more complete understanding of tumor immunology, particularly in the poorly understood areas of immune cell migration in cancer and of innate immune response to cancer. We have examined the role of one family of adhesion molecules, the selectins, in mouse models of tumorigenesis and determined that selectins promote tumor immune surveillance in our experimental system. We have also uncovered some of the mechanisms by which selectins control tumor growth by showing that NK cell recruitment and in vivo tumor-suppressive functions are defective in the absence of selectins. These results strongly implicate selectin-mediated recruitment of NK cells in tumor surveillance. Understanding the migration and recruitment mechanisms of immune cells in cancer, whereas challenging, is both scientifically interesting and potentially clinically valuable, raising the possibility of targeting the immune response specifically to the tumor.

No potential conflicts of interest were disclosed.

Grant support: NIH, National Cancer Institute grants U54CA126515 and U54CA112967; Howard Hughes Medical Institute; and Ludwig Center at MIT and Biology Department MIT training grant (O. Sobolev). A. Lacy-Hulbert is a UK Research Council Fellow. R.O. Hynes is Howard Hughes Medical Institute Investigator.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank Dr. Toshiyuki Tanaka for providing the TM-β1 hybridoma cells, Dr. Herman Eisen for RR4-7 hybridoma cells, Dr. Ajit Varki for the MC38GFP cells, Dr. Wayne Yokoyama for the GrzA-Ly49A transgenic mice, Dr. David Raulet for the Rae1ε construct, and the MIT Division of Comparative Medicine and the Koch Institute FACS and histology facilities for their help.