Blood Journal
Leading the way in experimental and clinical research in hematology

Human NK cells of mice with reconstituted human immune system components require preactivation to acquire functional competence

  1. Till Strowig1,*,
  2. Obinna Chijioke2,*,
  3. Paolo Carrega3,
  4. Frida Arrey1,
  5. Sonja Meixlsperger2,
  6. Patrick C. Rämer2,
  7. Guido Ferlazzo3,4,, and
  8. Christian Münz1,2,
  1. 1Laboratory of Viral Immunobiology, and Christopher H. Browne Center for Immunology and Immune Diseases, The Rockefeller University, New York, NY;
  2. 2Viral Immunobiology, Institute of Experimental Immunology, University Hospital of Zürich, Zürich, Switzerland;
  3. 3Giannina Gaslini Institute, Genoa, Italy; and
  4. 4Laboratory of Immunology and Biotherapy, Department of Human Pathology, University of Messina, Messina, Italy


To investigate human natural killer (NK)–cell reactivity in vivo we have reconstituted human immune system components by transplantation of human hematopoietic progenitor cells into NOD-scid IL2Rγnull mice. We demonstrate here that this model allows the development of all NK-cell subsets that are also found in human adult peripheral and cord blood, including NKp46+CD56 NK cells. Similar to human cord blood, NK cells from these reconstituted mice require preactivation by interleukin-15 to reach the functional competence of human adult NK cells. Mainly the terminally differentiated CD16+ NK cells demonstrate lower reactivity without this stimulation. After preactivation, both CD16+ and CD16 NK cells efficiently produce interferon-γ and degranulate in response to stimulation with NK cell–susceptible targets, including K562 erythroleukemia cells. NK-cell lines, established from reconstituted mice, demonstrate cytotoxicity against this tumor cell line. Importantly, preactivation can as well be achieved by bystander cell maturation via poly I:C stimulation in vitro and injection of this maturation stimulus in vivo. Preactivation in vivo enhances killing of human leukocyte antigen class I negative tumor cells after their adoptive transfer. These data suggest that a functional, but resting, NK-cell compartment can be established in immune-compromised mice after human hematopoietic progenitor cell transfer.


Natural killer (NK) cells are innate lymphocytes that are primarily thought to curb viral infections and tumor cell expansion until antigen-specific adaptive immune responses can be primed to eradicate these threats to human health.1 In contrast to adaptive lymphocytes like T and B cells, NK cells recognize their targets through germ line encoded receptors. These receptors transmit either activating or inhibitory signals.2,3 The activating receptors recognize primarily stress-induced molecules on infected and transformed cells, including major histocompatibility complex (MHC) class I–like molecules that serve as ligands for the activating NK-cell receptor NKG2D, PVR and Nectin-2 as ligands for the activating NK-cell receptor DNAM-1, and B7-H6, as well as ligands of still poorly defined identity for the natural cytotoxicity receptors (NCRs) NKp30, NKp46, and NKp44.4,5 Ligands for these activating receptors are up-regulated upon for example DNA damage or heat shock,6,7 but are also constitutively present on some hematopoietic cells, including myeloid dendritic cells (DCs),8 microglia,9 and activated macrophages.10 These activating signals are balanced by inhibitory receptor engagement, recognizing classical and nonclassical MHC class I molecules. In humans, killer immunoglobulin-like receptors (KIRs) recognize polymorphic determinants of classical MHC class I molecules, and C-type lectin receptors like the CD94/NKG2 heterodimer engage the nonclassical MHC class I molecule human leukocyte antigen (HLA)–E.11 The balance of transmitted activating and inhibitory signals decides if NK cells will mount effector functions against conjugated target cells.

The main effector characteristics of NK cells are cytokine secretion and cytotoxicity,12 and humans carry NK-cell subsets that preferentially mediate one or the other of these functions. CD56brightCD16KIR NK cells respond primarily with production of interferon-γ (IFN-γ), tumor necrosis factor, and granulocyte-macrophage colony-stimulating factor to activation, and only exert cytotoxicity after prolonged activation.13 In contrast, CD56dimCD16+KIR+ NK cells are constitutively loaded with perforin and granzymes and are the primary human cytotoxic NK-cell subset.14 While the latter population constitutes the majority of peripheral blood (PB) NK cells, CD56brightCD16KIR NK cells are enriched in human secondary lymphoid organs.15,16 They have been proposed to limit pathogen invasion and polarize adaptive immune responses at these sites.12,17 Thus cytotoxic NK cells patrol primarily the periphery, while immunoregulatory NK cells support Th1 polarization in secondary lymphoid organs.

The developmental pathways leading to the functionally distinct human NK-cell subsets are still being defined.18 So far 3 alternative pathways have been proposed. Originally, it was proposed that NK cells develop exclusively in the bone marrow from which they populate the periphery as constitutively reactive innate lymphocytes.1 After the discovery that the immunoregulatory human NK cells could acquire phenotypic and functional characteristics of cytotoxic NK cells,1922 it was proposed that CD56brightCD16KIR NK cells could home to secondary lymphoid tissues by virtue of their CD62L and CCR7 expression and mature to CD56brightCD16+KIR+ NK cells at these sites.23 Alternatively, CD56brightCD16KIR NK cells might also develop directly from CD34+ precursors in secondary24 or primary lymphoid organs.25 To shed some more light on the requirements for the development of the human NK-cell subsets and their functional competence, we investigated human NK-cell compartments reconstituted from CD34+ hematopoietic stem and progenitor cells in NOD-scid IL2Rγnull (hu-NSG) mice in comparison to human cord blood (CB) and adult blood NK cells. We describe here that both hu-NSG and CB NK cells require preactivation by monokines to reach functional competence as found in adult peripheral blood, and part of this recruitable activity lies dormant in terminally differentiated CD16+ NK cells.


Preparation of hu-NSG mice

NOD. Cg-Prkdcscid Il2rgtm1Wjl/SzJ (NSG) mice were obtained from The Jackson Laboratory and raised under specific pathogen-free conditions. Human fetal liver was obtained from Advanced Bioscience Resources. The tissue was minced and treated with 2 mg/mL collagenase D (Roche Diagnostics) in Hanks balanced salt solution with CaCl2/MgCl2 for 30 minutes at room temperature followed by filtering through 70-μm nylon cell strainers (BD Biosciences). CD34+ human hematopoietic stem and progenitor cells (HPCs) were isolated using the direct CD34 MicroBead kit (Miltenyi Biotec). Two- to 5-day-old NSG mice were irradiated with 100 cGy and injected intrahepatically with 1-2 × 105 CD34+ HPCs 6 hours after irradiation. The mice were bled 10-12 weeks after engraftment, and peripheral lymphocytes were analyzed by fluorescence-activated cell sorting as described previously26,27 to check for the reconstitution of the human immune system. Animal protocols were approved by the Institutional Animal Care and Use Committee of the Rockefeller University and the Cantonal Veterinary Office Zürich.

Isolation of blood mononuclear cells

PB from healthy adult volunteers was obtained as part of Institutional Review Board-approved protocols from the New York, Liguria, and Zürich Blood Centers. Human CB samples were kindly provided by the Liguria Cord Blood Bank (San Martino Hospital, Genoa, Italy) after delivery of the infant and after prior need for clinical samples had been satisfied. Samples were collected from full-term mothers upon their informed consent for research purpose, according to the declaration of Helsinki. CB was processed within 16 hours after birth to secure viability of cells. CB and PB were diluted 1:4 or 1:2 with phosphate-buffered saline (PBS), respectively, and mononuclear cells (CBMCs/PBMCs) were isolated by density-gradient centrifugation on Ficoll/Hypaque. Cells were either directly analyzed or cryopreserved for further analyses. Purified NK cells were obtained by negative selection using NK Cell Isolation kits (Miltenyi Biotec).

Isolation of cells from tissues of hu-NSG mice

For isolation of cells from spleen and lung, organs were minced and treated with 2 mg/mL collagenase D in Hanks balanced salt solution with CaCl2/MgCl2 for 30 minutes at room temperature followed by filtering through 70-μm nylon cell strainers. Livers of mice were perfused with PBS to remove blood for the isolation of liver-resident cells and subsequently single-cell suspensions were generated as described for spleen and lung. Single-cell suspensions from lymph nodes were generated by meshing the organs through 70-μm nylon cell strainers. Contaminating erythrocytes in cell suspensions were lysed using ACK lysis buffer or removed by density-gradient centrifugation on Ficoll/Hypaque, and remaining cells were washed and subjected to subsequent analysis and functional assays. Purified NK cells were obtained by negative selection using NK Cell Isolation kits, supplemented with anti–mouse CD45 antibodies (Miltenyi Biotec).

Generation of polyclonal NK-cell lines

To generate polyclonal NK cells from mice with reconstituted human immune system components, murine cells were depleted from splenocytes using anti–mouse CD45 microbeads, and human cells were subsequently cultured in medium with 100 IU/mL of interleukin-2 (IL-2). Cells were split upon becoming confluent and medium was replaced. After 2 weeks, cell cultures contained > 95% human NK cells (hCD45+NKp46+) as analyzed by flow cytometry.

Degranulation assay

To characterize the functional capacities of NK cells within bulk cell populations, we stimulated bulk cells (splenocytes, CBMCs, PBMCs) with target cells at an effector to target ratio of 1:1 for 6 hours in the presence of anti-CD107a antibody. To detect spontaneous degranulation and cytokine production, a control without target cells was included. After 1 hour, monensin (1 μg/mL; Sigma-Aldrich) was added to all samples. At the end of the incubation, cells were stained on ice with antibodies against human CD45, CD3, and NKp46, washed, fixed, permeabilized, and finally labeled with an anti–IFN-γ antibody. Cells were subsequently analyzed by flow cytometry.

Cytotoxicity assay

To evaluate the cytotoxic activity of NK cells against the K562 target cell line, cytotoxicity assays were performed as previously described.16 Briefly, target cells were labeled with PKH26 (Sigma-Aldrich) and then incubated with NK-cell lines at the indicated effector/target ratios. After 4 hours, cells were harvested; TO-PRO-3-iodide, a membrane-impermeable DNA stain, was added to each culture (0.5μM final concentration), and cells were finally analyzed by flow cytometry. Background and maximum TO-PRO-3-iodide staining were obtained by incubation of target cells with medium and detergent, respectively. The percentage specific lysis was calculated as [(% TO-PRO-3-iodide+PKH26+ cells in effector/target cell coculture − % TO-PRO-3-iodide+PKH26+ cells in medium)/(% TO-PRO-3-iodide+PKH26+ cells in detergent − % TO-PRO-3-iodide+PKH26+ cells in medium)] × 100%.

In vivo cytotoxicity assay

LCL721.221 and LCL721.45 were grown in RPMI, 10% fetal calf serum and penicillin/streptomycin supplemented with L-glutamine. After labeling with 5μM carboxyfluorescein succinimidyl ester (CFSE) for 10 minutes at room temperature, the reaction was stopped by the addition of fetal calf serum, cells washed twice in PBS and 5 × 106 cells of both lines injected intravenously into the tails of hu-NSG mice either with or without pretreatment with 50 μg poly I:C by intraperitoneal injection 18 hours before. Single-cell suspensions of recipient spleens were analyzed 12 hours after transfer after being stained with allophycocyanin (APC)–conjugated anti–HLA class I (w6/32; BioLegend) for the presence of CFSE-positive cells. Specific lysis was as calculated by gating on CFSE-positive cells as (% MHC class I positive − % MHC class I negative) × 100/% MHC class I positive.

Adoptive transfer of purified NK-cell populations

CD3NKp46+CD56 cells from pooled spleens of hu-NSG mice were sorted to high purity (> 98%) on a FACS Aria (BD Biosciences) and labeled with 5μM CFSE. Sorted cells (1-2 × 105) were adoptively transferred into recipient hu-NSG mice (reconstituted with same CD34+ HPCs as donor hu-NSG mice), which were killed 48 hours later. Recovered NKp46+CFSE+ cells from spleens were then analyzed for CD56 expression by flow cytometry.

Monoclonal antibodies and flow cytometry

Analysis of cell surface markers was performed using the following monoclonal antibodies (mAbs): V450-conjugated anti-CD45 (clone HI30), fluorescein isothiocyanate (FITC)– or PerCP-conjugated anti-CD3, APC-conjugated anti-NKp46, phycoerythrin (PE)–Cy7–conjugated anti-CD56, APC-Cy7–conjugated anti-CD16 and FITC- or PE-conjugated anti-CD107a (BD PharMingen), FITC-conjugated anti-CD19, PE-conjugated anti-CD158a, h, PE-conjugated anti-CD158b1/b2, j, PE-conjugated anti-CD127, and PE-conjugated anti-NKp46 (Immunotech-Coulter), Alexa Fluor 647 conjugated anti-CD107a (eBioscence), APC-conjugated anti-CD56, PE-conjugated CD117, FITC-conjugated anti-CD16, and FITC-conjugated anti-CD3 were from Miltenyi Biotech. The unlabeled anti-NKp46 mAb (clone BAB281) was generously provided by Prof Maria C. Mingari's laboratory (University of Genova, Genoa, Italy). Direct immunofluorescence staining was performed diluting fluorochrome-labeled mAb with 1 mg/mL human γ-globulin (human therapy grade from Biotest SRL) to block nonspecific FcR binding. Cells were then washed, and flow cytometric analysis was performed. For indirect immunofluorescence staining, the Alexa Fluor 633-conjugated isotype-specific goat anti–mouse Ab (eBioscence) was used. Negative controls included directly labeled and unlabeled isotype-matched irrelevant mAbs. IFN-γ expression analysis were performed by PE-conjugated reagents (BD Biosciences-PharMingen) or Pacific Blue anti–human IFN-γ antibody (BioLegend) after cells were fixed in 2% paraformaldehyde and permeabilized with 0.1% saponin.

Statistical analysis

Statistical analyses were performed using the Graph-Pad Prism Version 4.00 for Windows (GraphPad Software). Statistical significance was evaluated by 2-tailed Student t test.


NOD-scid IL2Rγnull mice reconstitute human NK cells from CD34+ hematopoietic progenitor cells

To reconstitute human immune system components in NSG mice, we injected mice neonatally with 105 human CD34+ hematopoietic stem and progenitor cells intrahepatically.27,28 After 3 months we monitored human B-cell, T-cell, monocyte, DC, and NK-cell reconstitution. As previously reported,27 around 60%-80% of mononuclear splenocytes were of human origin at this timepoint. Of these 50%-60% were B cells, 25%-30% were T cells (CD4:CD8 ratio 3:2), 1%-2% were plasmacytoid DCs, 1%-2% were conventional DCs, 3%-5% were monocytes, and 1%-3% were NK cells. We decided to characterize the human NK-cell distribution further. Compared with spleen, we found higher percentages of CD3NKp46+ NK cells in lung, blood, and liver, while below 1% of mononuclear cells of the bone marrow were NK cells (Figure 1A-B). In lung, the organ with the highest NK-cell fraction, up to 7% NK cells were detected. However, when taking total cell numbers for these organs into account, spleen and liver constituted the largest NK-cell reservoirs after reconstitution (Figure 1C). These data suggest that human NK cells reconstitute in NSG mice with human immune system components (hu-NSG), and they are present in all strongly blood perfused organs.

Figure 1

Distribution of human NK cells in reconstituted NSG mice. Frequency within human CD45 positive cells and total number of CD3NKp46+ human NK cells in spleen, blood, lung, liver, and bone marrow (BM) of NSG mice reconstituted with human immune system components 3 months after transfer of CD34+ hematopoietic progenitor cells. Representative flow cytometric staining from 1 mouse (A) and composite data from 7 mice (B-C) are shown. Numbers in plots represent frequencies within gates.

Hu-NSG mice and CB contain a NKp46+CD56 NK-cell subset

To characterize the NK-cell subsets that were developing in hu-NSG mice, we performed phenotyping of the splenic NK-cell compartment. We found KIR, CD94, NKG2D, NKp46, NKG2A, and CD16 expression on splenic CD3CD56+ NK cells (Figure 2A). In comparison to human spleen, whose NK-cell compartment is in the majority composed of CD56dimCD16+ NK cells (75%),16 the minority of hu-NSG–derived splenic NK cells belonged to this terminally differentiated subset (25%-40%), suggesting a more immature NK-cell compartment in hu-NSG mice compared with human adult spleen. Interestingly, in addition to CD56brightCD16 and CD56dimCD16+ NK cells, hu-NSG spleen contained a large subpopulation of NKp46+CD56 NK cells (Figure 2B). Part of these carried, in addition to NKp46, other NK-cell markers like NKG2A and CD16. Some of the NKp46+CD56CD16+ NK cells expressed KIRs (20%-30%), while minimal or no KIR expression was detected on CD16 NK cells (Figure 2C). Vice versa, a subset of CD3NKp46+CD56CD16 cells stained positive for CD127 and CD117 and could therefore constitute stage III NK-cell precursors.29 Because CD56 is up-regulated during human NK-cell activation,30 we considered the possibility that the NKp46+CD56 NK cells might be a resting NK-cell population prior to a history of pathogen exposure. To provide more evidence for this hypothesis, we investigated human CB to determine whether this NK-cell subpopulation also existed prior to human pathogen exposure. Indeed, we found within NKp46+ CB NK cells both CD56 and CD16 negative cells (Figure 2D). The frequency of these NKp46+CD56 NK cells was in contrast very low in adult blood (around 3%). NKp46+CD56 NK cells constituted around 15% of CB NK cells amounting in some CB samples to up to 30% of the human NK-cell compartment (Figure 2E). To determine whether indeed these CD3NKp46+CD56 cells can differentiate into CD3NKp46+CD56+ NK cells, we adoptively transferred purified and CFSE-labeled CD3NKp46+CD56 cells into hu-NSG mice reconstituted with autologous CD34+ HPCs by intravenous injection. After 48 hours, NKp46+CFSE+ cells were recovered from recipient spleens and analyzed for their CD56 expression (Figure 2F). All recovered cells had up-regulated CD56. Similarly, CD3NKp46+CD56 cells from hu-NSG mice up-regulated CD56 after activation in vitro (supplemental Figure 1, available on the Blood Web site; see the Supplemental Materials link at the top of the online article). These data suggest that spleens of recently reconstituted hu-NSG mice and human CB contain a NKp46+CD56 NK-cell population that constitutes around 50% of NK cells in hu-NSG mice and up to 30% of NK cells in human CB, and can differentiate into NKp46+CD56+ NK cells in vitro and in vivo.

Figure 2

NK-cell subset development in NSG mice in comparison to human adult and CB. CD3CD56+ human NK cells from spleens of reconstituted NSG mice were analyzed for KIR, CD94, NKG2D, CD16, NKG2A, and NKp46 expression (A). One representative of 12 mice is shown. CD3NKp46+ human NK cells from spleens of reconstituted NSG mice were analyzed for CD56, CD16, and NKG2A expression (B). A representative staining from 1 of 12 mice is shown. CD16, KIR, CD127, and CD117 expression on CD3NKp46+CD56 cells of hu-NSG mice was analyzed (C). One representative of 3 stainings is displayed. In comparison, CD56 and CD16 expression was analyzed on CD3NKp46+ NK cells from human adult (PB) and CB. One representative example (D) and composite data (E) of 8 experiments are shown. CFSE-labeled CD3NKp46+CD56 cells of hu-NSG mice were adoptively transferred into hu-NSG mice, which had been reconstituted with autologous CD34+ HPCs. CD56 expression on the recovered NKp46+ CFSE-labeled cells from recipient spleens was analyzed after 48 hours by flow cytometry (F). Gray shaded histogram of post-sort CD3NKp46+CD56 cells in comparison to black histogram of post-sort CD56+ (left panel). White histogram of recovered cells in comparison to gray shaded histogram of cells that were transferred (right panel). One representative of 2 experiments is shown. Numbers in plots represent frequencies within gates or quadrants.

NK cells from hu-NSG mice have a decreased ability for cytotoxicity and IFN-γ production compared with NK cells from human adults

The functional capacity of NK cells from hu-NSG mice was evaluated by ex vivo stimulation with 2 NK cell–susceptible human target cell lines, the erythroleukemia cell line K562 and the T-cell lymphoma line CEM, as well as by monokine stimulation (IL-12 plus IL-15). While K562 is recognized by a combination of NCR-, NKG2D-, and DNAM-1–mediated mechanisms, triggering all NK-cell subsets, CEM-mediated stimulation of NK cells is primarily mediated by NKG2D and affects mainly CD56brightCD16 NK cells.31 As a surrogate marker for cytotoxicity, degranulation of NK cells in response to these stimuli was assessed by CD107a staining, and intracellular IFN-γ staining was monitored as a representative measure for cytokine production (Figure 3A). NK cells from hu-NSG mice degranulated slightly less than human adult PB NK cells in response to K562 cells. Strikingly, a nearly 10-fold difference, was, however, seen in IFN-γ production in response to K562 cells (Figure 3B). Interestingly, NKG2D-mediated recognition of CEM was similar between NK cells from hu-NSG mice and human adult blood. When we analyzed K562 recognition by the different NK-cell subsets of hu-NSG mice, the functional defect was primarily confined to CD56dim and NKp46+CD56 NK cells (Figure 3C). Similarly, NK cells from CB degranulated less efficiently in response to K562 than human adult blood NK cells, and this deficiency was primarily confined to CD56dim and NKp46+CD56 NK cells (Figure 3D). The same NK-cell subsets also produced less IFN-γ, when restimulated with K562 cells (data not shown). These data suggest that the resting NK-cell compartment of hu-NSG mice degranulate and produce cytokines less efficiently upon cognate target recognition than human adult NK cells, and part of this deficiency is caused by the CD56dim and NKp46+CD56 NK-cell populations.

Figure 3

Functional activity of human NK cells from reconstituted NSG mice in comparison to human adult or CB NK cells. Degranulation, a surrogate marker for cytotoxicity, and cytokine production were assessed by surface CD107a and intracellular IFN-γ staining in response to ex vivo stimulation with medium, the erythroleukemia cell line K562, the T-cell lymphoma cell line CEM, and the monokines IL-12 plus IL-15 (IL-12/15). One representative staining (A) and composite data (B) of 12 mice in 3 experiments are shown. In the composite data human NK-cell reactivity of reconstituted NSG spleens was compared with 3 adult PBMC samples. In addition, the function of the human NK-cell subsets CD56brightCD16, CD56dimCD16+ and NKp46+CD56 from human adult (PBMC), hu-NSG (C) mice or CB (CBMC in panel D) were compared for degranulation. Composite data of 3 (C) and 6 (D) experiments are shown. Numbers in plots represent frequencies within gates.

IL-15 mediated preactivation can overcome the functional deficiency of NK cells from hu-NSG mice

Because NK-cell preactivation by mature DCs has recently been identified as crucial for full functional capacity of these innate lymphocytes,8,3235 we exposed hu-NSG splenocytes to the monokines IL-12 and IL-15, as well as the DC-maturing TLR3/mda5 ligand polyinosinic:polycytidylic acid [poly I:C or p(I:C)]. Degranulation in response to K562 and CEM was markedly enhanced by IL-15 or poly I:C preactivation (Figure 4A). In addition, IFN-γ production by hu-NSG NK cells after stimulation with K562 and CEM was boosted 5- to 10-fold after preactivation with IL-15 or poly I:C in vitro (Figure 4B). Poly I:C was only able to enhance degranulation and IFN-γ production by NK cells, when bystander cells, presumably mainly DCs, were present, while purified NK cells from hu-NSG mice or CB only increased their degranulation and cytokine production in response to K562 after preactivation with IL-15, but not poly I:C (Figures 4C-F). From these data, we conclude that NK cells from hu-NSG mice and CB have a similar functional capacity as human adult blood NK cells, but require preactivation to respond efficiently to cognate target cells, while adult NK cells have a higher constitutive reactivity, possibly due to preactivation by environmental stimuli.

Figure 4

Preactivation enhances human NK-cell function in spleen cells from reconstituted NSG mice and human CB. Splenocytes from reconstituted NSG mice (A-B) purified NK cells from hu-NSG mice (C-D) and human CB (E-F) were used untreated (medium) or preactivated with poly I:C [p(I:C)] or the monokines IL-12 and IL-15. These cultures were restimulated with medium alone, monokines, K562 cells, or CEM cells. Degranulation (CD107a; A,C,E) and cytokine production (IFN-γ; B,D,F) were evaluated after gating on NKp46-positive cells. Panels A and B represent composite data from 4 mice in 2 experiments, panels C and D represent composite data from 10 mice in 2 experiments, and panels E and F represent composite data from 3 experiments.

Preactivation increases perforin and granzyme B content in NK cells of hu-NSG mice

In addition to the functional parameters of degranulation and IFN-γ production, we also tested if the cytotoxic armament of hu-NSG NK cells increased upon preactivation in vitro. For this purpose we stained intracellularly for perforin and granzyme B in CD3NKp46+ NK cells of hu-NSG splenocyte cultures that had been exposed to poly I:C or the monokines IL-12 and IL-15. Both IL-15 and poly I:C addition were able to increase the perforin and granzyme B content of hu-NSG mice around 5-fold (supplemental Figure 2A-B). Analogous to the in vitro activation data, the frequency of perforin expressing cells was only increased after IL-15 stimulation of purified splenic NK cells, while poly I:C was not able to induce cytotoxic granules directly (supplemental Figure 2C). Therefore, preactivation of NK cells via bystander cell maturation with poly I:C or direct stimulation by IL-15 increases both degranulation capacity and cytotoxic effector molecule expression of NK cells from hu-NSG mice.

Preactivation boosts degranulation and cytokine production in terminally differentiated NK cells of hu-NSG mice

Because we had noted that primarily CD56dim and NKp46+CD56 NK cells of hu-NSG mice and CB were functionally deficient in comparison to adult PB NK cells, we wondered if IL-15–mediated preactivation would affect these NK-cell compartments. We confirmed that these NK-cell compartments of hu-NSG mice contained CD16+KIR+ terminally differentiated human NK cells (Figure 5A-B) and that CD16 expression was not altered on purified NK cells from these subsets during 1 day of IL-15 activation, while CD56 was up-regulated during this activation period (supplemental Figure 1 and data not shown). This allowed us to analyze degranulation and cytokine production of CD16+ terminally differentiated hu-NSG NK cells in response to K562 stimulation after IL-15 preactivation. Indeed, degranulation and IFN-γ production by these NK cells were significantly, around 5 fold, elevated after preactivation in response to K562 stimulation (Figure 5C-D). In addition, CD16 NK cells also up-regulated IFN-γ production and degranulation in response to K562 cells after exposure to IL-15. Thus, the functional deficiency of terminally differentiated NK cells in hu-NSG mice can be overcome by preactivation with IL-15.

Figure 5

Terminally differentiated CD16+ NK cells acquire functional capacity after IL-15–mediated preactivation. Distribution of KIR+ (A) and CD16+ (B) NK cells in splenic CD56bright, CD56dim, and CD56 NK-cell subsets of hu-NSG mice, as gated in the top panels of (A) in comparison to isotype control staining or (B) staining on marker negative CD3+ cells (gray shaded histograms). Degranulation (C) and IFN-γ production (D) of unstimulated and preactivated CD16+ and CD16 NK cells after coculture with K562 cells were analyzed. A representative experiment of 2 is shown in panels A and B, while panels C and D represent composite data of 2 independent experiments. Numbers in plots represent frequencies within gates or marker regions.

NK cells from hu-NSG mice display cytotoxic function against K562 cells

So far we tested degranulation and cytotoxic molecule expression as surrogate markers for NK-cell cytotoxicity of hu-NSG mice. To demonstrate cytotoxicity against K562 cells by human reconstituted NK cells from hu-NSG mice directly, we established NK-cell lines from hu-NSG mice (Figure 6A). These lines contained to more than 95% NKp46+ NK cells with partial KIR expression. They were comparable or slightly superior to NK-cell lines established from adult human blood with respect to perforin and granzyme B expression. We used the To-Pro-3 incorporation assay as a measurement of direct cytotoxicity.9,16,36 The NK-cell lines from hu-NSG mice killed K562 with high efficiency of more than 70% lysed cells already at a 10:1 NK:K562 ratio (Figure 6B). This suggests that NK cells from hu-NSG mice are functionally competent, but require cytokine preactivation to reach their full functional potential.

Figure 6

Cytotoxic ability of a human NK-cell line from reconstituted NSG mice. (A) Comparison of NKp46, KIR, perforin, and granzyme B (GrzB) expression of a reconstituted mouse-derived human NK-cell line with a line of human adult PB NK cells. (B) Cytotoxicity of the human NK-cell line from reconstituted NSG mice against K562. Loss of membrane integrity of PKH26-labeled K562 cells as a measure of cytotoxicity was assessed by To-Pro-3 iodide staining of DNA, which is blocked by intact cell membranes. Cytotoxicity was evaluated at the indicated NK:K562 (E:T) ratios. One representative of 6 experiments with 3 different NK-cell lines from 3 mice is shown. Numbers in plots represent frequencies within gates.

Poly I:C administration enhances human NK-cell functions in vivo

The experiments described in the previous paragraph documented the composition and functions of NK cells from hu-NSG mice ex vivo and after activation in vitro. However, to assess the capacity of these innate lymphocyte to become functional effectors in vivo, we injected poly I:C intraperitoneally. Eighteen hours after injection, hu-NSG mice were killed, and degranulation as well as intracellular IFN-γ content of CD3NKp46+ splenocytes was assessed ex vivo or after restimulation with monokines or the NK cell–susceptible cell lines K562 and CEM. Already ex vivo, NK cells demonstrated low levels of CD107a staining and intracellular IFN-γ content (Figure 7A-C). Moreover, they responded with 3-fold better degranulation and 5- to 10-fold increased IFN-γ production to restimulation with K562 and CEM (Figure 7B-C). In addition, when we adoptively transferred CFSE-labeled HLA class I deficient LCL721.221 cells alongside their parental HLA class I positive LCL721.45 cells into hu-NSG mice, LCL721.221 cells were killed in vivo within 12 hours, and this cytotoxicity was enhanced by prior poly I:C injection (Figure 7D-E). Cytolysis of LCL721.221 increased from 40% to 85% upon NK-cell preactivation in vivo. These data document that NK cells from hu-NSG mice can be rapidly, within 18 hours, preactivated by poly I:C injection in vivo. This should allow the study of human NK-cell functions against viral infections and tumors in vivo in hu-NSG mice.

Figure 7

Preactivation of human NK cells of reconstituted NSG mice in vivo. Splenocytes of untreated or poly I:C [p(I:C)]-injected reconstituted NSG mice were restimulated ex vivo with medium alone or the monokines IL-12 and IL-15, or K562 cells, or CEM cells. Both degranulation (B) and cytokine production (IFN-γ; A,C) were evaluated. One representative of 2 experiments (A) and composite data (B) from 8 mice are shown. CFSE-labeled LCL721.221 (HLA class I negative) and LCL721.45 (HLA class I positive) cells were injected intravenously into untreated or poly I:C–preactiveted hu-NSG mice. Twelve hours later, the composition of CFSE-labeled cells in the recipient spleens was analyzed by w6/32 (anti–HLA class I) staining. Analysis of HLA class I expression and CFSE prior to adoptive transfer (D) and after recovery after gating on CFSE+ cells (E). One representative of 3 experiments is shown. Numbers in plots represent frequencies within gates or marker regions.


Our study assessed human NK-cell reconstitution from CD34+ hematopoietic stem and progenitor cells in NSG mice. We found reconstitution of NK-cell subsets that are similar to human CB, differing by the additional presence of NKp46+CD56 NK cells from the subpopulations found in human adult blood. Functionally, these reconstituted NK-cell populations also resembled the lower spontaneous NK-cell activity found in CB, secreting lower levels of IFN-γ and degranulating less efficiently in response to susceptible tumor cell lines like K562. This deficiency was primarily found in terminally differentiated CD56dimCD16+ NK cells, the main cytotoxic NK-cell compartment in humans. However, CB and reconstituted NSG mouse NK cells could be preactivated directly with IL-15 or indirectly via poly I:C-mediated bystander cell maturation in vitro and in vivo to reach cytotoxicity and cytokine secretion levels similar or exceeding human adult PB NK cells. These findings suggest that resting NK cells with a similar composition as human CB NK cells can be reconstituted in hu-NSG mice, and are functionally competent with respect to mounting NK cell–effector functions after preactivation.

The human NK-cell reconstitution in hu-NSG mice is with respect to NK-cell subset distribution similar to human NK-cell development in BALB/c Rag2−/− γc−/− mice,22 in which also NKp46+CD56, CD56brightCD16KIR, and CD56dimCD16+KIR+ NK-cell populations could be observed. Interestingly, however, in all investigated organs except for bone marrow, the frequency of developing NK cells was 5- to 10-fold lower in reconstituted BALB/c Rag2−/− γc−/− versus hu-NSG mice.22,37 This suggests that while NK-cell development in the bone marrow progresses to similar levels in these 2 immunodeficient mouse strains, human NK cells survive and, therefore, accumulate to higher levels in NSG mice. The lower murine myeloid xenoreactivity mediated by enhanced cross-reactivity of NOD SIRP-α toward human CD4738 could be in part responsible for this better NK-cell reconstitution. However, human NK cells in BALB/c Rag2−/− γc−/− mice were able to delay K562 tumor growth in vivo39 and can be expanded in BALB/c Rag2−/− γc−/− mice by IL-15 or IL-15/IL-15Rα hybrid molecule administration, reaching then levels of steady-state NK-cell reconstitution in NSG mice.22 This expansion occurs at the expense of the CD16 negative NK-cell subsets and converts nearly all NK cells to CD16+ partially KIR+ NK cells. A similar NK-cell expansion could also be induced in reconstituted NSG mice by hydrodynamic delivery of IL-15 encoding expression plasmids, but the effect on NK-cell subset differentiation was not investigated in detail in this study.40 Therefore, we propose that the more efficient NK-cell subset reconstitution in hu-NSG mice recommends this model for studying immune responses mediated by these human innate lymphocytes in vivo.

In addition to this essential role of IL-15 for NK-cell development and accumulation, which was also previously reported for mouse NK cells,41,42 IL-15 has also been shown to be required for preactivation (termed “priming”) of mouse NK cells to reach their full functional potential.35 During viral and bacterial infections, this NK-cell preactivation is performed by DCs, possibly via IL-15 trans-presentation.35 Similarly in humans, DCs can activate NK cells to produce cytokines, proliferate, and acquire elevated cytotoxicity.8,32,33,4345 Thus reconstituted human NK cells of hu-NSG mice resemble resting NK cells, which like mouse NK cells from animals housed under pathogen-free conditions, require preactivation to become fully functional. In contrast, human adult NK cells display a higher constitutive reactivity, because they are probably partially preactivated through continuous pathogen encounter. Especially the CD56dimCD16+ NK cells required this preactivation step, while CD56brightCD16 NK cells demonstrated already significant reactivity against K562 cells ex vivo, quite similar to CD56brightCD16 NK cells from human adult blood. These findings are consistent with a previous report, which also suggested that deficiency in NK-cell reactivity of human CB was primarily confined to CD56dimCD16+ NK cells.46 This suggests that primarily terminally differentiated human NK cells require preactivation. We also identified IL-15 as a potent mediator of human NK-cell preactivation in hu-NSG mice, and poly I:C–mediated maturation for NK-cell preactivation probably primarily affected also human DCs to fulfill this function. Indeed, we have previously shown that poly I:C–mediated DC maturation renders these cells more efficient in NK-cell stimulation than other maturation stimuli,47 and these DCs can interact with both CD56brightCD16 and CD56dimCD16+ NK cells to transmit IL-15–dependent signals.48 Thus, hu-NSG mice with reconstituted human immune system compartments should allow studies on resting human NK cells, which are functionally competent.

The functional competence that we observed after preactivation, also questions further the requirement of NK licensing by somatic tissues for reactivity of this innate lymphocyte subset.49,50 Licensing describes functional competence of NK cells that sequentially up-regulate inhibitory NK-cell receptors during development until they express at least one for a self-MHC class I molecule. This model is supported by several lines of evidence. Firstly, transgenic expression of an inhibitory receptor decreases the frequency of NK-cell expression for other inhibitory receptors, but only in the presence of the cognate MHC class I molecule.51 Secondly, NK cells that do not express self-MHC reactive inhibitory receptors are functionally attenuated.52,53 Thirdly, NK cells of transporter associated with antigen processing-deficient patients who do not express MHC class I molecules at high levels on their cell surfaces due to defective peptide loading are functionally compromised.54,55 In contrast, during mouse cytomegalovirus infection, primarily unlicensed NK cells without an inhibitory receptor for polymorphic self-MHC class I molecules restrict virus load,50 and unlicensed NK cells acquire functional competence after cytokine activation.56 Likewise, NK cells lacking inhibitory receptors specific for MHC class I molecules are activated after Listeria monocytogenes infection in vivo, and then produce as much IFN-γ as NK cells carrying these inhibitory receptors.52 Considering these different possibilities, the functional competence of reconstituting human NK cells in hu-NSG mice suggests that licensing can either occur via inhibitory NK cell–receptor engagement (CD94/NKG2A or KIR) by HLA class I molecules on reconstituting hematopoietic cells, in addition to previously investigated stromal cells,57 or that unlicensed NK cells can gain similar functionality as licensed human adult blood NK cells after preactivation in vivo. Further studies on the influence of transgenic HLA class I expression in hu-NSG mice might yield valuable insights into this NK cell–development question.

Altogether, our data point toward functional competence of resting human NK cells that develop from CD34+ hematopoietic stem and progenitor cells in hu-NSG mice. We propose that this model could allow the study of NK-cell responses during infection with lymphotrophic human pathogens like the Epstein Barr virus.27 These studies could reveal the contribution of NK cells to initial pathogen restriction and priming of protective immune control. Such information could then be harnessed to develop vaccination approaches against human pathogens.


Contribution: T.S., O.C., and P.C. performed research; F.A., S.M., and P.R. contributed vital reagents; and T.S., O.C., G.F., and C.M. designed research and wrote the manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

The current affiliation for T.S. is Department of Immunobiology, Yale University School of Medicine, New Haven, CT.

Correspondence: Christian Münz, Viral Immunobiology, Institute of Experimental Immunology, University of Zürich, Winterthurerstr 190, CH-8057 Zürich, Switzerland; e-mail: christian.muenz{at}


We thank the Center for Microscopy and Image Analysis of the University of Zürich for expert technical support and Vinko Tosevski for expert technical advice with flow cytometric cell sorting.

This research was partly supported by grants from the National Cancer Institute (R01CA108609 and R01CA101741), from the Foundation for the National Institutes of Health (Grand Challenges in Global Health), and the Swiss National Science Foundation (310030_126995) to C.M., and by the Associazione Italiana per la Ricerca sul Cancro (AIRC) and the Ministero Italiano della Salute, Progetto Strategico Oncologia to G.F. S.M. is a recipient of a postdoctoral fellowship from the German Research Foundation (Deutsche Forschungsgemeinschaft).


  • * T.S. and O.C. contributed equally to this study.

  • G.F. and C.M. contributed equally to the work.

  • The online version of this article contains a data supplement.

  • The publication costs of this article were defrayed in part by page charge payment. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.

  • Submitted February 18, 2010.
  • Accepted July 22, 2010.


View Abstract