Effect of the in vivo application of granulocyte colony‐stimulating factor on NK cells in bone marrow and peripheral blood

Abstract Granulocyte colony‐stimulating factor (G‐CSF) has been widely used in the field of allogeneic haematopoietic stem cell transplantation (allo‐HSCT) for priming donor stem cells from the bone marrow (BM) to peripheral blood (PB) to collect stem cells more conveniently. Donor‐derived natural killer (NK) cells have important antitumour functions and immune regulatory roles post‐allo‐HSCT. The aim of this study was to evaluate the effect of G‐CSF on donors' NK cells in BM and PB. The percentage of NK cells among nuclear cells and lymphocyte was significantly decreased and led to increased ratio of T and NK cells in BM and PB post‐G‐CSF in vivo application. Relative expansion of CD56bri NK cells led to a decreased ratio of CD56dim and CD56bri NK subsets in BM and PB post‐G‐CSF in vivo application. The expression of CD62L, CD54, CD94, NKP30 and CXCR4 on NK cells was significantly increased in PB after G‐CSF treatment. G‐CSF treatment decreased the IFN‐γ‐secreting NK population (NK1) dramatically in BM and PB, but increased the IL‐13‐secreting NK (NK2), TGF‐β‐secreting NK (NK3) and IL‐10‐secreting NK (NKr) populations significantly in BM. Clinical data demonstrated that higher doses of NK1 infused into the allograft correlated with an increased incidence of chronic graft‐vs‐host disease post‐transplantation. Taken together, our results show that the in vivo application of G‐CSF can modulate NK subpopulations, leading to an increased ratio of T and NK cells and decreased ratio of CD56dim and CD56bri NK cells as well as decreased NK1 populations in both PB and BM.

allo-HSCT procedures consist mostly of bone marrow (BM) cells or granulocyte colony-stimulating factor (G-CSF)-primed peripheral blood stem cells (GPB) or G-CSF-primed bone marrow (GBM). Natural killer (NK) cells mediate the early, non -adaptive responses against viruses and intracellular bacteria and modulate the activity of effector cells within the adaptive and innate immune systems. 1 NK cells mediate these effects through the production of cytokines and the direct killing of transformed or infected cells. [2][3][4] The function of NK cells is modulated by the balance of expression of inhibitory killer immunoglobulinlike receptors (KIRs) and activating receptors on NK cells. Previous studies have shown that donor and recipient KIR ligand mismatch can initiate donor NK cell alloreactivity, leading to decreased leukaemia relapse and decreased GVHD incidence post-haploidentical transplantation with T cell depletion in vitro. 5 However, the predictive value of KIR ligand mismatch on clinical outcome has been inconsistent among different transplantation centres utilizing different protocols. 6,7 One of the most important reasons would be the in vivo application of G-CSF for donor stem cell preparation that decreased the cytotoxicity of NK cells. 8,9 However, previous studies have considered only the cytotoxic roles of NK cells under haploidentical transplantation while neglecting the immune regulatory effect of NK cells.
Although both GBM and GPB contain large numbers of mature donor T cells that could cause GVHD, 10 clinical data have demonstrated that the cumulative incidence of acute GVHD is acceptable with these cell sources. 11 Based on the cytokine secretion model, CD4 + T cells can be classified as type 1 (Th1), type 2 (Th2) and type 17 (Th17) subpopulations of T cells. [12][13][14][15][16][17] Similar to Th1 and Th2 cells, human NK cells cultured in the presence of IL-12 or IL-4 can differentiate into populations with distinct patterns of cytokine secretion. [18][19][20][21][22][23] Specifically, IFN-c-secreting NK cells (NK1) were shown to be important for infection control, 24,25 while IL13-secreting NK cells (NK2), which contribute to IgE production by B cells, participate in the regulation of allergic airway responses. [21][22][23] In addition, peripheral blood NK cells producing TGF-b (NK3) and IL-10 (NKr) have been shown to have a regulatory function in humans. [26][27][28][29] Our previous work demonstrated that the in vivo application of G-CSF induced T cell hyporesponsiveness. In particular, the levels of Th1 and Th17 cells were decreased, while those of Th2 cells were increased in BM and PB grafts using G-CSF. [13][14][15]30 However, the content and function of NK cells in BM before and after G-CSF in vivo application have not been analysed.
The aim of this study was to explore the effect of G-CSF on the NK1/NK2/NK3/NKr subpopulations, including CD56 bri and CD56 dim NK subsets, as well as the proliferation and cytotoxicity of NK cells in BM and PB. We also assessed the predictive roles of NK1, NK2, NK3 and NKr cells in allografts on clinical outcome post-allo-HSCT.
2 | ME TH ODS 2.1 | G-CSF treatment of healthy donors and sample collection Steady-state peripheral blood (NGPB) and bone marrow (NGBM) before G-CSF in vivo treatment, as well as GBM and GPB, were obtained from 15 allogeneic donors. This group of donors, 8 men and 7 women, provided informed consent and had a median age of 29 years ranging from 18 to 54 years. Donors received recombinant G-CSF (filgrastim; Kirin Co., Ltd., Tokyo, Japan) at a dosage of 5 lg/ kg/d for 5 consecutive days. GBM was collected on the 4th day of treatment by aspiration, and GPB was obtained on the 5th day by leukapheresis using a continuous-flow blood cell separator (Gambro BCT, Lakewood, CO, USA; or Baxter, Chicago, IL, USA). The reason for using this protocol was that patients in our institute receive transplants composed of GBM plus GPB, which are harvested on days 4 and 5, respectively. 31,32 Additional donors were collected GBM and GPB only. Therefore,

| Transplant procedure and definition
All patients received a myeloablative regimen, and conditioning was performed as previously described. 31,32,34,35 In HLA-matched sibling transplants, 32

| Statistical analysis
To test differences in NK cell expression of receptors or cytokine secretion between NGPB and GPB; GPB and GBM; and NGBM and GBM, a Wilcoxon signed-rank test or paired-sample t test was used. Associations between the dose and percentage of NK1, NK2, NK3 and NKr cells infused in GBM or GPB and GVHD were calculated using cumulative incidence curves to accommodate competing risks. Gray's test was used in the cumulative incidence analyses.

| Effect of G-CSF on NK cell expansion
The percentages of overall NK cells among nuclear cells and lymphoid cells were significantly decreased in BM and PB cells post-G-CSF in vivo application (P < .05, Figure 1A,B). The ratio of T cells and NK cells was significantly decreased in PB and BM after G-CSF treatment (P < .05, Figure 1C). The relative expansion of the CD56 bri NK subsets led to a decreased ratio of CD56 dim and CD56 bri NK cells in GBM and GPB compared to that in NGBM and NGPB, respectively (P < .05, Figure 1D).  increased compared to those on NK cells in GBM ( Figure 1E-G). The expression levels of CXCR4 on NK cells in GPB were only higher compared to those in NGPB ( Figure 1H). In contrast,the expression of CX3CR1 on NK cells in GPB was significantly decreased compared to those in NGPB ( Figure 1I). The MFI of CD62L, CD54, CD94 and CXCR4 on NK cells in GPB were also higher compared to those in NGPB (data not shown). The percentage of CD11a on NK cells was comparable among NGPB, GPB, NGBM and GBM, but the MFI of CD11a on NK cells in GPB had a trend to be higher compared to those in NGPB and GBM ( Figure 1J).

| G-CSF differentially affects NK cell subpopulations in BM compared to PB in vivo
Cytotoxicity and proliferation capacity of NK cells were evaluated in BM and PB before and after G-CSF in vivo application. No significant differences were found in the proliferation capacity of NK cells among NGPB, GPB, NGBM and GBM (Figure 2A-B). Because NKG2A + CD57 À , NKG2A + CD57 + as well as NKG2A À CD57 + NK subsets formed different development stages of CD56 dim NK cells, 36 we further analysed the proliferation of NKG2A + CD57 À , NKG2A + CD57 + as well as NKG2A À CD57 + NK cells to explore whether the different proliferation capacity of NK subsets could contribute to the decreased ratio of CD56 bri and CD56 dim in PB and BM after G-CSF in vivo treatment. But no differences were found among the above-mentioned NK subsets before and after G-CSF in vivo application in PB and BM. Among the NKG2A-NK subpopulation, we further compared the proliferation capacity of licensed single KIR + NK cells and unlicensed single KIR + NK cells, and no significant differences were found (data not shown). As shown in Figure 1H, consistent with previous reports, the cytotoxicity of NK cells against K562 target cells was significantly decreased in GPB compared to that in NGPB; however, no significant difference was found in the cytotoxicity of NK cells between NGBM and GBM ( Figure 2C). Meanwhile, the expression of IFN-gamma and CD107a against K562 cells of overall NK cells, CD56 bri NK as well as CD56 dim NK cells had a trend to be decreased without CD56 dim and CD56 bri NK subsets in NGPB,NGBM and GBM, but was higher in CD56 bri NK subsets compared to those in CD56 dim NK subsets among GPB (P = .017). Secretion of TGF-beta was comparable between CD56 dim and CD56 bri NK subsets in NGBM and GBM, but was higher in CD56 dim NK subsets compared to those in CD56 bri NK subsets among NGPB and GPB (P = .05 and .01, respectively). Secretion of IL-13 and IL-10 was higher in CD56 dim NK subsets compared to those in CD56 bri NK subsets among NGPB and NGBM (P = .008 and .01 for IL-13, P = .034 and .007 for IL-10, respectively), but was comparable between CD56 dim and CD56 bri NK subsets in GPB and GBM.  Figure 5B) and extensive cGVHD (45.41 AE 12.86% vs 11.46 AE 7.9%; P = .04; Figure 5C) than patients in the low NK1 group. The patient characteristics of the high and low NK1 groups (as shown in Table 1) were comparable, with the exception of the NK1 cell dose in the allografts.

| Correlation between NK1\NK2\NK3\NKr cells in allografts with chronic GVHD occurrence posttransplantation
No associations were found between the NK subpopulation numbers in allografts and the relapse, treatment-related mortality (TRM), overall survival (OS) and event-free survival (EFS) rates (data not shown).

| DISCUSSION
In this study, the following main findings were made: 1) in vivo application of G-CSF not only decreased the percentage of NK cells but also modulated NK subpopulations, leading to an increased ratio of

F I G U R E 3 Comparison of IFN-gamma and CD107a expression of NK cells against K562 cells among NGPB and GPB, NGBM and GBM. The expression of IFN-gamma and CD107a of NK cells (A and D), CD56 bri NK cells (B and E), CD56 dim NK cells (C and F) against K562 cells without stimulation or with IL-2 or IL-15 stimulation among NGPB and GPB, NGBM and GBM
A previous study demonstrated that myeloid progenitors were expanded, and lymphoid progenitors were decreased after the in vivo application of G-CSF. 8 Consistent with this previous report, 8 the percentages of NK cells among GBM and GPB were significantly decreased in our study; therefore, the ratio of T and NK cells in GBM or GPB was significantly increased compared to that in NGBM and NGPB, respectively. However, we found that the ratio of CD56 bri to CD56 dim subsets was significantly increased in GBM and GPB compared to that in NGBM and NGPB, respectively, which differed from this previous report. 8 Miller JS et al demonstrated that G-CSF in vivo treatment had no effect on the proportion of CD56 bri NK cells. 8 The reason for these disparate results could be because of to expand in vitro and in vivo in response to low (picomolar) doses of IL-2. These NK cells also express the c-kit receptor tyrosine kinase, whose ligand enhances IL-2-induced proliferation. 37    with the fact that Th1 cells can contribute to chronic GVHD progression, 40  Meanwhile, in accordance with a previous report, 8 In summary, we demonstrate that the in vivo application of G-CSF can modulate NK subpopulations, leading to a high ratio of CD56 bri to CD56 dim subsets and low levels of the NK1 population. A higher dose of NK1 cells infused in allografts correlated with an increased incidence of chronic GVHD after transplantation. Future work is aimed at investigating the molecular pathway underlying the effect of G-CSF on NK cells, which would be helpful for decreasing GVHD but maintaining the potential antitumour benefits of NK cells in the allograft setting.

ACKNOWLEDGMENTS
This study was supported by grants from the National Natural Science

CONF LICT OF I NTERESTS
The authors declare that they have no Conflict of interests.