Increased surveillance of cells in mitosis by human NK cells suggests a novel strategy for limiting tumor growth and viral replication

Viral-infected or tumor cells can become susceptible to lysis by natural killer (NK) cells via expression of stress-inducible proteins¹  or by decreased expression of self class I MHC protein.²  Many viruses, including human cytomegalovirus known to be targeted by NK-cell surveillance, encode genes that help drive cell-cycle progression.³  In addition, NK cells react more efficiently to concanavalin A-stimulated, proliferating MHC class I–deficient target cells than to nonactivated cells in vitro and in vivo.^4,5  However, whether cell-cycle progression could be used as an endogenous indicator for potential danger has not been considered. To test this, we compared the innate response of human NK cells to target cells in mitosis or other stages of the cell cycle.

Human peripheral blood NK cells were isolated as described previously.⁶  For T cells, peripheral blood mononuclear cells (PBMCs) were stained with FITC-labeled anti-CD3 monoclonal antibody (mAb; (BD Biosciences, San Jose, CA), positively enriched using anti-FITC microbeads (Miltenyi Biotec, Auburn, CA), and cultured for 1 week in IL-2–containing medium. Peripheral blood NK cells and target cells were coincubated at 37°C/5% CO₂, fixed, stained, and imaged.⁶  Cells were stained using anti–α-tubulin mAb (clone DM1A; Sigma) used at 0.5 μL per 3 × 10³ cells, followed by Alexa Fluor 488–labeled goat anti–mouse IgG (Molecular Probes, Paisley, United Kingdom), and were then washed in PBS. The fraction of mitotic and nonmitotic target cells bound by one or more NK cells was then assessed by confocal microscopy (TCS SP2; Leica, Heidelberg, Germany) using a 63 ×/1.32 numerical aperture oil objective. Images were acquired and analyzed using Leica confocal software version 2.61. Data were analyzed by multinomial logistic regression analysis with fixed model design (SPSS, Chicago, IL).

Target cells were treated with 30 U/mL Streptomyces hyalurolyticus hyaluronidase, 0.5 U/mL chondroitinase ABC, or 1.5 U/mL heparinase I (Sigma, St Louis, MO) for 40 minutes. NK cells were preincubated with 1 μg/mL heparin (Leo Pharmaceuticals, High Wycombe, United Kingdom), heparan sulfate, chondroitin sulfate A or B, or hyaluronan (Sigma) for 30 minutes before coincubation.

NK cells were incubated for 30 minutes on ice with 10 μg/mL anti-CD44 blocking mAb or control isotype-matched anti-CD44 mAb (BRIC 235 or KZ1, respectively⁷ ), anti-NKp46 (195314; R&D Systems, Minneapolis, MN), mouse IgG2b control (BD Biosciences), goat polyclonal anti-NKp46 (C-20; R&D Systems), goat IgG control (Jackson ImmunoResearch, West Grove, PA), or anti-NKG2D (149810; R&D Systems). Anti-NKp44 mAb was a gift from Carsten Watzl (Heidelberg, Germany).

Target cells (721.221; 0.5 × 10⁶/mL) were incubated overnight at 37°C, 5% CO₂ with 0.1 μM colchicine (Sigma) or 100 ng/mL nocodazole (Sigma). For comparison, incubation in 10 μM colchicine for 1 hour was used. HT1080 cells were enriched by shake-off.⁸ 

A variety of pathogenic target cells, namely, EBV-transformed MHC class I–deficient B cells 721.221, Jurkat T-cell leukemia cells, Burkitt lymphoma-derived Daudi cells, or myeloid leukemia K562 cells, were separately incubated with human NK cells derived from peripheral blood. Images of conjugates stained for α-tubulin were used to identify cells in mitosis by the characteristic mitotic spindle (Figure 1A). We found that NK cells conjugated to cells in mitosis significantly more frequently than to the same cells in other stages of the cell cycle (Figure 1B).

Next, 721.221 cells were treated for 18 hours with low concentrations of the microtubule-depolymerizing drug, colchicine, to increase the fraction of cells in mitosis from approximately 4% to approximately 75%. NK-cell degranulation, measured by NK-cell surface staining for CD107a (LAMP-1),⁹  was up to 4-fold greater in response to target cells arrested in mitosis compared with control-treated target cells (Figure 1C-D). A 1-hour treatment of target cells with a high concentration of colchicine, disrupting the target-cell cytoskeleton without arresting cells in mitosis, did not alter the NK-cell response (Figure 1C-D). Target cells arrested in mitosis using the drug nocodazole also triggered increased NK-cell degranulation (Figure 1E). To establish an assay independent of using drugs, the fraction of HT1080 fibrosarcoma cells in mitosis was enriched to about 40% to 50% by shaking off detached cells. NK-cell degranulation triggered by these cells was approximately twice that triggered by adherent HT1080 cells in other stages of the cell cycle (Figure 1F). Thus, susceptible target cells in mitosis trigger a heightened NK response compared to the same cells in other stages of the cell cycle.

We also found an increased binding of NK cells to lysis-resistant autologous T cells in mitosis (Figure 1G). However, NK-cell degranulation in response to autologous cells in mitosis remained low, even after 1 hour of coincubation (Figure 1H). Thus, mitosis triggers increased surveillance by NK cells but does not inevitably induce NK-cell responses; presumably, self-tolerance achieved through inhibitory NK-cell receptors remains dominant to allow healthy cells to undergo cell division.

We next set out to test for specific molecular recognition of cells in mitosis. Production of the pericellular matrix, including hyaluronan, increases to facilitate morphologic changes during mitosis.¹⁰  Treatment of 721.221 cells with hyaluronidase, which catalyzes degradation of hyaluronan, abrogated specific binding of cells in mitosis (Figure 2A). In contrast, increased binding of cells in mitosis was unaffected by degradation of other glycosaminoglycans (Figure 2A). Additionally, when NK cells were incubated with different glycosaminoglycans as putative blocking agents, only hyaluronan was able to prevent increased binding of NK cells to target cells in mitosis (Figure 2B). Thus, hyaluronan facilitates NK-cell recognition of cells in mitosis.

NK-cell CD44 was previously shown to bind hyaluronan on tumor cells and augment adhesion and cytotoxicity.¹¹  However, increased NK-cell binding of cells in mitosis was unaffected by a CD44-blocking mAb (Figure 2C) used at a concentration sufficient to completely inhibit the high-level binding of hyaluronan to BW5147 cells (data not shown). Thus, receptors other than CD44 on NK cells are involved in recognition of cells in mitosis.

Natural cytotoxicity receptors (NCRs) are major activating NK-cell receptors with unknown cellular ligands.¹²  Mice lacking Ncr1 (NKp46 in humans) are impaired in their ability to clear certain tumors.¹³  Here, blocking NCR NKp46 with either monoclonal or polyclonal antibodies reduced NK-cell binding to 721.221 (Figure 2D) or Daudi (Figure 2E) cells in mitosis to the same level as cells in other cell-cycle stages. Blocking NKG2D, an activating NK-cell receptor not involved in lysis of 721.221,¹⁴  did not interfere with binding of cells in mitosis.

Additionally, NCR expression varies in NK cells,¹²  and NK-cell clones that expressed high levels of NKp46 showed more pronounced recognition of target cells in mitosis (Figure 2F-G). Antibody-mediated blocking of NKp46 on NKp46^high NK cells reduced the binding of target cells in mitosis to a similar level as NKp46^low NK cells (Figure 2F). Antibody-mediated blocking of the NCR NKp44, which is expressed on activated NK cells,¹⁵  also abrogated specific binding of 721.221 cells in mitosis (Figure 2H). That blocking either NKp46 or NKp44 was sufficient to abrogate increased binding to cells in mitosis is consistent with different NCR functioning synergistically.^16,17  Thus, ligation of activating receptors involved in lysis of 721.221 facilitates increased binding of 721.221 cells in mitosis.

To test whether NKp46 directly binds hyaluronan, we used an enzyme-linked immunosorbent-type assay to determine if plate-bound hyaluronan bound immunoglobulin-fusion proteins of NCR or CD44 (positive control). Additionally, we tested if hyaluronan would stimulate reporter cells transfected to express a chimera consisting of the extracellular portion of NKp46 and the intracellular signaling domain of CD3 ζ, an approach used previously.¹⁸  Using these methodologies, we could not detect direct binding between NKp46 and hyaluronan (data not shown). Thus, although direct interaction between NCR and hyaluronan remains possible, alternatively, inside-out signaling via NCR could activate a hyaluronan-binding receptor¹⁹  or signaling via NCR and a hyaluronan-binding receptor could synergize to augment responses to cells in mitosis. Increased surveillance of cells in mitosis suggests an important novel strategy to safeguard tumor growth or viral replication.

This work was supported by the Human Frontier Science Program, the Medical Research Council, Fundação para a Ciência e a Tecnologia, the Biotechnology and Biological Sciences Research Council, the Wellcome Trust, and a Lister Institute Research Prize.