Granulocyte colony‐stimulating factor impairs CD8⁺ T cell functionality by interfering with central activation elements

Summary

Besides mobilizing stem cells into the periphery, granulocyte colony‐stimulating factor (G‐CSF) has been shown to influence various types of innate and adaptive immune cells. For example, it impairs the effector function of cytotoxic T lymphocytes (CTLs). It is assumed that this effect is mediated indirectly by monocytes, regulatory T cells and immunomodulatory cytokines influenced by G‐CSF. In this study, isolated G‐CSF‐treated CD8⁺ T cells were stimulated antigen‐dependently with peptide–major histocompatibility complex (pMHC)‐coupled artificial antigen‐presenting cells (aAPCs) or stimulated antigen‐independently with anti‐CD3/CD28 stimulator beads. By measuring the changes in interferon (IFN)‐γ and granzyme B expression at the mRNA and protein level, we showed for the first time that G‐CSF has a direct effect on CD8⁺ CTLs, which was confirmed based on the reduced production of IFN‐γ and granzyme B by the cytotoxic T cell line TALL‐104 after G‐CSF treatment. By investigating further elements affected by G‐CSF in CTLs from stem cell donors and untreated controls, we found a decreased phosphorylation of extracellular‐regulated kinase (ERK)1/2, lymphocyte‐specific protein tyrosine kinase (Lck) and CD3ζ after G‐CSF treatment. Additionally, miRNA‐155 and activation marker expression levels were reduced. In summary, our results show that G‐CSF directly influences the effector function of cytotoxic CD8⁺ T cells and affects various elements of T cell activation.

Introduction

Granulocyte‐colony stimulating factor (G‐CSF) mobilizes haematopoietic stem cells from the bone marrow into the peripheral blood. In comparison to bone marrow transplantation, the use of peripheral blood stem cells (PBSCs) has proved equally successful in stem cell transplantation 1, 2, 3, and PBSCs are used in more than 80% of transplantations today 4. G‐CSF also influences various other cells of the innate and adaptive immune system. It was shown to induce a tolerant phenotype by increasing monocyte numbers 5, leading to high interleukin (IL)‐10 levels 6. G‐CSF also suppresses the induction of the CD28 responsive complex in CD4⁺ T cells 7, and affects the migratory and homing abilities of dendritic cells (DCs) by up‐regulation of CCR7 8. The ratio of DC1/DC2 cells decreases in response to G‐CSF, favouring IL‐3 receptor‐positive DC2 cells, which mainly induce T helper type 2 (Th2) responses 9. G‐CSF also activates regulatory T cells (T_regs), leading to increased secretion of IL‐10 and transforming growth factor (TGF)‐β, both of which contribute to a tolerant phenotype. Deregulation of genes involved in effector function (granzyme B, granulysin) and induction of graft‐versus‐host disease (GvHD) [mitogen‐activated protein kinase (MAPK14), chemokine C‐X‐C motif (CXCR1)] during and after G‐CSF treatment was found in cytotoxic T lymphocytes (CTLs) 10. We showed recently that antigen‐specific T cells isolated from G‐CSF‐treated stem cell donors and directed against viral epitopes from cytomegalovirus (CMV), Epstein–Barr virus (EBV) and adenoviruses (ADV) secreted significantly lower amounts of effector molecules such as interferon (IFN)‐γ and granzyme B than untreated controls 11. These findings were supported by the results obtained using in vitro G‐CSF‐treated antigen‐specific T cells from healthy thrombocyte donors.

Successful T cell stimulation and activation by the four essential signals [T cell receptor (TCR) stimulation, co‐stimulation, cytokines and chemokines] 12, 13 induces several intracellular processes, such as Ca²⁺ mobilization, phosphorylation of kinases and changes in the expression of regulatory microRNAs (miRNAs) 12, 14. Following either antigen‐specific or antigen‐independent TCR recognition and interaction with co‐stimulatory molecules, two main regulatory branches are activated, resulting in further changes in the cells.

First, signalling pathways are activated by phosphorylation of kinases, leading to a change in gene expression and in the activation state of the cells. The lymphocyte‐specific protein tyrosine kinase (Lck) is associated with the cytoplasmic domains of the TCR co‐receptors CD4 or CD8. Lck is brought into close proximity to its target, the CD3‐ζ chain immunoreceptor tyrosine‐based activation motif (ITAM). The Lck‐dependent phosphorylation of CD3‐ζ ITAMs allows the recruitment of zeta‐chain‐associated protein kinase 70 (ZAP70) and sequential phosphorylation of ZAP70 by Lck 15. Activation of ZAP70 induces more phosphorylation events and subsequent activation of multiple adaptor and signalling molecules, resulting in the activation of several signalling pathways responsible for the differentiation, proliferation and exhibition of effector functions 16. One of these, the extracellular‐regulated kinase (ERK1/2) pathway, is critical for different T cell functions, including proliferation, differentiation and cytokine production; in particular, it is involved in IFN‐γ signalling 17, 18.

Secondly, miRNAs are important for many processes such as adaptive immune responses, T cell development, survival, proliferation and activation 14 and therefore form an additional regulatory element involved in and crucial for T cell activation and function. Several specific miRNAs have been reported so far to be expressed differentially in naive and end‐stage differentiated T cells 19, and the miRNA expression profile of CD8⁺ T cells is changed immediately after viral infections 20. Two targets of microRNA (miR)‐155 are suppressors of cytokine signalling (SOCS1) and Src homology 2‐containing inositol phosphatase‐1 (SHIP1) which, like other genes, are involved in IFN signalling, promoting T cell proliferation, survival, activation and effector function 14, 21. Recent studies showed that G‐CSF treatment modulates the expression of miRNAs in haematopoietic stem cells for up to 1 year after treatment 22, 23.

An extracellular event following T cell activation is the up‐regulated cell surface expression of molecules such as CD25, CD38, CD69 or CD137. This is a crucial part of the activation process, as it allows the interaction with other cells, the uptake of cytokines and the reception of co‐stimulatory signals, which further results in different gene expression patterns and the induction of effector functions. The pathways in effector T cells altered by G‐CSF and leading to the impaired anti‐viral effector function are not known. It is assumed that T cell function is impaired indirectly by the effects of G‐CSF on DCs and CD4⁺ T cell properties. However, the effects of G‐CSF on the regulation of miRNA expression patterns in effector T cells have not been investigated.

Recently we showed that T cell functionality is impaired by G‐CSF administration 11. This study aimed to determine if this effect is mediated indirectly by monocytes, regulatory T cells and immunomodulatory cytokines influenced by G‐CSF or if G‐CSF treatment directly affect effector T cell functionality. The effects of G‐CSF on T cell functionality were investigated after antigen‐dependent and ‐independent stimulation. By using major histocompatibility complex (MHC)‐coated artificial antigen‐presenting cells (aAPCs), loaded with viral CMV‐derived peptides, we could stimulate CD8⁺ T cells in an antigen‐specific manner without the presence of other G‐CSF‐susceptible cell populations in the system. Antigen‐independent stimulation was performed using anti‐CD3/CD28 stimulator beads. In both approaches, we could show for the first time that G‐CSF has direct effects on anti‐viral CD8⁺ effector T cells, leading to reduced IFN‐γ and granzyme B levels. Furthermore, we showed that key elements of T cell activation such as (1) surface expression of the activation markers CD25, CD38, CD69, CD137 and human leucocyte antigen D‐related (HLA‐DR), (2) the activation of TCR‐dependent and ‐independent signalling pathways, as reflected by the phosphorylation of ERK1/2, Lck and the CD3‐ζ chain (CD247), and (3) the expression of microRNA‐155 are affected negatively by G‐CSF, which contributes to the impaired effector function.

Material and methods

G‐CSF mobilization and apheresis

G‐CSF‐induced mobilization of haematopoietic stem cells was performed by administering 10 µg/kg body weight per day of G‐CSF (filgrastim; Amgen, Thousand Oaks, CA, USA) for 4 consecutive days. CD34⁺ counts were determined and apheresis was scheduled accordingly using the Cobe^® Spectra version 4·7 system (Terumo BCT, Lakewood, CO, USA).

Cells and cell lines

The cytotoxic T cell line TALL‐104 was cultured in Iscove's modified Dulbecco's medium (ATCC, Washington, DC, USA) supplemented with 20% fetal bovine serum (FBS) (Biochrome, Berlin, Germany), 2·5 µg/ml human albumin, 0·5 µg/ml D‐mannitol and 100 U/ml human IL‐2 (all Sigma Aldrich, Munich, Germany). K562 served as targets cells and were cultured in RPMI‐1640 (Lonza, Verviers, Belgium) supplemented with 10% FBS.

Peripheral blood mononuclear cells (PBMCs) from healthy HLA‐A*02:01‐positive, CMV‐seropositive thrombocyte donors and G‐CSF mobilized haematopoietic stem cell donors were isolated by discontinuous density gradient centrifugation. CD4⁺ and CD8⁺ T cells from stem cell donors and CD4⁺ T cells from thrombocyte donor peripheral blood mononuclear cells (PBMCs) were isolated using fluorescence‐activated cell sorting (FACS) at the Cell Sorting Core Facility of Hannover Medical School. CD8⁺ T cells from thrombocyte‐donor PBMCs were isolated by FACS or magnetic cell sorting using the CD8⁺ T cell isolation kit (Miltenyi Biotec, Bergisch Gladbach, Germany), according to the manufacturer's instructions. Purity was routinely higher than 95%.

Antigen‐specific stimulation of T cells using aAPCs

To examine the direct effects of G‐CSF on effector T cells, aAPCs were generated by coupling HLA‐A*02:01 molecules [DimerX; Becton Dickinson (BD), Heidelberg, Germany] loaded with HLA*02:01‐restricted CMVpp65_495–503 peptide (NLVPMVATV, A02pp65_P; ProImmune, Oxford, UK) and anti‐CD28 monoclonal antibodies (mAbs) (BD) onto Dynabeads magnetic stimulator beads (Life Technologies, Carlsbad, CA, USA), as described previously 24. PBMCs and isolated CD8⁺ T cells were cultured in aAPC medium consisting of RPMI‐1640 supplemented with 5% heat‐inactivated human AB serum (c.c.pro, Neustadt, Germany), 1% sodium pyruvate (Sigma Aldrich), 0.4% modified Eagle's medium (MEM) vitamins and 1% non‐essential amino acids (both Life Technologies) and 50 U/ml IL‐2 with or without 10 ng/ml G‐CSF (both from PeproTech GmbH, Hamburg, Germany). After 7 days, the frequency of antigen‐specific A02pp65_P‐positive T cells was assessed by pentamer staining and effector molecule secretion was determined by IFN‐γ and granzyme B enzyme‐linked immunospot (ELISPOT).

Polyclonal T cell activation

For antigen‐independent stimulation, isolated CD8⁺ T cells were stimulated with human T activator CD3/CD28 Dynabeads (Life Technologies), according to the manufacturer's instructions in aAPC medium with or without 10 ng/ml G‐CSF and without the addition of IL‐2. On days 1 and 2, mRNA and miRNA expression levels of IFN‐γ, granzyme B, SHIP1, SOCS1, miR‐21 and miR‐155 were determined. PBMCs from thrombocyte or stem cell donors (n = 3) were cultured in RPMI‐1640, 10% heat‐inactivated human AB serum and 50 U/ml IL‐2, and were stimulated in anti‐CD3‐coated 96‐well plates [muromonab‐CD3 (OKT3); eBioscience, San Diego, CA, USA]. PBMCs from thrombocyte donors were stimulated with or without 10 ng/ml G‐CSF in vitro, and expression of CD25, CD38, CD57, CD69, CD137 and HLA‐DR was assessed by flow cytometric analysis after 1, 2, 3 and 7 days of stimulation.

Antibodies and flow cytometric analysis

All flow cytometric analyses were performed using the FACSCanto II system (BD Biosciences, Heidelberg, Germany) and BD FACSDiva Software version 6·1·2. At least 100 000 events were acquired in the live gate or at least 30 000 in the CD3⁺ gate. CMV‐specific T cell frequencies were assessed by pMHC multimer staining using R‐phycoerythrin (R‐PE)‐conjugated pentamer HLA‐A*02:01/CMVpp65_495–503 (A02pp65_M; ProImmune), according to the manufacturer's instructions. Additionally, cells were stained with the following mAbs: anti‐CD8 allophycocyanin (APC), anti‐CD3 peridinin chlorophyll (PerCP) and anti‐CD19 FITC (all BD). For phosphorylation analysis, cells were stained with anti‐pERK1/2‐Alexa Fluor 647 or immunoglobulin (Ig)G1 Alexa Fluor 647 isotype control, anti‐pLck PE or IgG1 PE isotype control, anti‐CD247 Alexa Fluor 647 or IgG2a Alexa Fluor 647 isotype control, anti‐CD3 FITC and anti‐CD8 PerCP‐Cy5.5 (all BD). Expression of activation markers was assessed by staining with anti‐CD8 PerCP (BD), anti‐CD38 PE‐Cy7, anti‐HLA‐DR APC‐Cy7 (all BD), anti‐CD25 APC‐H7, anti‐CD57 FITC, anti‐CD69 FITC, anti‐CD95L PE and anti‐CD137 PE‐Cy7 (all BioLegend, San Diego, CA, USA). Pentamer staining was performed for 10 min at room temperature; all other antibodies were stained at 4°C for 30 min.

Gene and microRNA expression analysis

Total RNA from CD8⁺ T cells isolated from thrombocyte donors (n = 5; untreated and in vitro G‐CSF‐treated) was isolated using the mirVana RNA Isolation Kit (Life Technologies), and cDNA was reverse‐transcribed using either the high‐capacity cDNA reverse transcription kit or the microRNA transcription kit (both Life Technologies), according to the manufacturer's instructions. Expression of granzyme B, IFN‐γ, SOCS1, SHIP1, miR‐21 and miR‐155 was quantified using inventoried mixes (Life Technologies). Expression of glyceraldehyde 3‐phosphate dehydrogenase (GAPDH) and miR‐191 served as internal controls.

For global gene expression analysis, CD8⁺ T cells isolated by FACS (purity > 98%) from four thrombocyte and four stem cell donors were pooled and subjected to the 026652AsQuintuplicatesOn4x180K microarray at the Research Core Unit Transcriptomics of Hannover Medical School. This microarray represents a refined version of the Whole Human Genome Oligo microarray 4x44K version 2 (AMADID 026652; Agilent Technologies, Santa Clara, CA, USA) and was developed at the Research Core Unit Transcriptomics of Hannover Medical School. Microarray design was defined at Agilent's eArray portal using a 4x180k design format for mRNA expression as template. All non‐control probes of AMADID 026652 were selected to be printed five times onto one 180K microarray (on‐chip quintuplicates). Control probes required for proper Feature Extraction software algorithms were determined and placed automatically by eArray using recommended default settings.

Detection of the effector molecules granzyme B and IFN‐γ by ELISPOT assay in response to target cell recognition

Granzyme B release of antigen‐specific T cells was detected by ELISPOT assay, as described previously 25, with some modifications. Briefly, autologous PBMCs loaded overnight with 10 µg/ml A*02:01‐restricted CMV pp65 peptide (A02pp65_P) were used as target cells. After 7 days of aAPC stimulation, T cells from four donors were incubated at 37°C with target cells at effector to target cell ratios (E : T) between 0·004 : 1 and 0·1 : 1, depending on the percentage of antigen‐specific T cells. After 4 h of incubation, plates were washed, and biotinylated anti‐granzyme B antibody (Mabtech, Stockholm, Sweden) was added. Granzyme B secretion was detected using streptavidin–alkaline phosphatase (Mabtech) and revealed by 5‐13 bromo‐4‐chloro‐3‐indolyl phosphate/nitroblue tetrazolium (BCIP/NBT liquid substrate; Sigma‐Aldrich). Spots were counted using an ImmunoScan Core Analyzer, and the results were analysed using ImmunoSpot Academic Software version 5·0 (both from Cellular Technology Ltd, Bonn, Germany). Means of triplicates were calculated as spots per 100 pentamer⁺ cells. The basal cytolytic activity of effector T cells against unloaded target cells was subtracted from the specific values. Further negative controls consisted of effector cells only, target cells only and medium only. In the case of TALL‐104 cells, K562 cells were used as target cells and the basal cytolytic activity subtracted from the specific lytic activity was from TALL‐104 cells only.

Virus‐specific IFN‐γ‐producing T lymphocytes were enumerated by IFN‐γ ELISPOT assay as described for the granzyme B ELISPOT assay, with the exception that target and effector cells were incubated for 16 h before addition of the biotinylated anti‐IFN‐γ antibody (Mabtech).

Phosphorylation analysis

PBMCs from stem cell donors treated with G‐CSF in vivo and from untreated thrombocyte donors (n = 5) were stimulated with 50 ng/ml phorbol myristate acetate (PMA) (Sigma‐Aldrich) for 15 min or with anti‐CD3/CD28 antibodies for 1 min and evaluated for expression of phosphorylated ERK1/2 (pERK1/2), CD247/CD3ζ (pCD247) and Lck (pLck). Detection of phosphorylation for each phosphoprotein was optimized previously. Stimulation with anti‐CD3/CD28 antibodies for 1 min was specified as most suitable for pLck and pCD247, while stimulation with PMA for 15 min was evaluated as the optimal stimulation condition for detection of pERK1/2. Prior to staining, cells were fixed with Fix Buffer I for 10 min at 37°C and permeabilized with Perm Buffer III (both BD) for 30 min on ice. Staining was performed as described in the flow cytometric analysis section for 30 min at 4°C. For the detection of pLck and pCD247, PBMCs (n = 3) were chilled on ice for 15 min and stimulated with anti‐CD3 and anti‐CD28 antibodies, followed by incubation with goat anti‐mouse Ig for 15 min on ice (all 1 µg/100µl; all BD). PBMCs were then stimulated at 37°C for 1 min and fixed, permeabilized and stained as described for pERK detection, with the exception that unbound goat anti‐mouse Ig was blocked prior to staining by the addition of normal mouse Ig for 15 min at room temperature. Finally, the percentage and median fluorescence intensity (MFI) of pLck, pCD247 and pERK in CD8⁺ T cells were determined.

Statistics

Statistical analyses were performed using paired or unpaired Student's t‐tests run on GraphPad Prism version 5·02 software (GraphPad Software, San Diego, CA, USA). Levels of significance were expressed as P‐values [*P < 0·05, **P < 0·01, ***P < 0·001, not significant (n.s.)].

Study approval

Written informed consent was obtained from all donors, and the study was approved by the Ethics Committee of Hannover Medical School.

Results

G‐CSF directly acts on CD8⁺ T cells

To determine whether G‐CSF has a direct influence on the effector functions of CD8⁺ T cells, PBMCs and isolated CD8⁺ T cells were stimulated for 7 days with peptide (A02pp65_P)‐loaded aAPCs treated with (+ G‐CSF) or without G‐CSF (– G‐CSF). On day 7, the cells were restimulated with CMV peptide‐loaded autologous PBMCs, and assessed for the frequency of antigen‐specific A02pp ${65}_{\mathrm{P}}^{+}$ cells as well as for IFN‐γ and granzyme B expression. The frequency of antigen‐specific A02pp ${65}_{\mathrm{P}}^{+}$ cells ranged from approximately 40 to 90%, but was comparable in cells treated with or without G‐CSF (mean with G‐CSF: 58·80%, without G‐CSF 61·14%). A representative experiment is shown in Fig. 1a. As determined by ELISPOT, G‐CSF treatment reduced IFN‐γ secretion in PBMCs to 59 and 71% (relative to untreated cells) and reduced that of CD8⁺ T cells to 94 and 75% at E : T ratios of 0·004 : 1 and 0·01 : 1, respectively (Fig. 1b,c). Granzyme B release in G‐CSF‐treated versus ‐untreated cells decreased to 68 and 65% in PBMCs and to 73 and 79% in CD8⁺ T cells at the same E : T ratios (Fig. 1b,d).

Figure 1.

Primary CD8⁺ T cells treated with granulocyte colony‐stimulating factor (G‐CSF) secrete fewer effector molecules. Peripheral blood mononuclear cells (PBMCs) and isolated CD8⁺ T cells (n = 4) were stimulated for 7 days with A02pp65_P‐loaded artificial antigen‐presenting cells (aAPCs) and treated with G‐CSF or left untreated. The cells were then restimulated with A02pp65_P‐loaded autologous PBMCs, and interferon (IFN)‐γ secretion and granzyme B release were assessed by enzyme‐linked immunospot (ELISPOT) assays. The data are means plus minus (±) standard error of the mean (SEM). (a) Representative dot‐blot showing the results for one donor. There was no significant difference in the expansion and percentage of antigen‐specific CD8⁺ T cells treated with (+ G‐CSF) or without G‐CSF (– G‐CSF). (b) Representative IFN‐γ and granzyme B ELISPOT raw data are shown for cells at effector to target (E : T) ratios of 0·004 : 1 (ratio 1) and 0·01 : 1 (ratio 2). (c) IFN‐γ secretion and (d) granzyme B release of expanded PBMCs and CD8⁺ T cells treated with or without G‐CSF.

To confirm the results obtained with primary CD8⁺ T cells, the effect of G‐CSF on the effector functions of the CD8⁺ cytotoxic T cell line TALL‐104 was analysed (Fig. 2). IFN‐γ and granzyme B secretion by TALL‐104 cells pretreated with or without G‐CSF for 3 days were determined by ELISPOT assays using K562 target cells. G‐CSF treatment decreased IFN‐γ and granzyme B secretion of pretreated TALL‐104 cells significantly to 65% (IFN‐γ) and 77% (granzyme B) relative to non‐pretreated cells. Furthermore, G‐CSF‐treated K562 cells served as a control to exclude effects of G‐CSF treatment on target cells. TALL‐104 cells incubated with pretreated K562 cells showed no significant difference in IFN‐γ (109%) or granzyme B secretion (121%) relative to cells without pretreatment.

Figure 2.

Granulocyte colony‐stimulating factor (G‐CSF) treatment affects effector function of TALL‐104 cells negatively. Interferon (IFN)‐γ secretion and granzyme B release by TALL‐104 cells pretreated with G‐CSF or left untreated for 3 days were determined by enzyme‐linked immunospot (ELISPOT) assays using K562 target cells. K562 cells pretreated with G‐CSF for 3 days served as a control to exclude effects of G‐CSF administration on target cells alone. The results for G‐CSF‐pretreated cells are expressed as % expression relative to non‐pretreated cells. Values represent means of triplicate experiments at an effector to target ratio of 0.02 : 1. Statistical significance was determined by two‐tailed paired Student's t‐test (n.s., non‐significant).

Cell surface activation marker expression is reduced after in vitro and in vivo G‐CSF treatment

PBMCs from stem cell donors treated with G‐CSF in vivo and from untreated thrombocyte donors were stimulated on anti‐CD3‐coated plates for 7 days. Subsequently, the cells from untreated donors were either left untreated or were treated with G‐CSF in vitro for 7 days. The percentage and median fluorescence intensity (MFI) of expression of activation markers CD25, CD38, CD57, CD69, CD137 and HLA‐DR are shown in Fig. S1. The sum of expression for CD25, CD38, CD57, CD69, CD137 and HLA‐DR on CD8⁺ T cells after 1 day of stimulation was 228·42% for untreated cells, 196·10% for in vitro G‐CSF‐treated cells and 130·40% for in vivo G‐CSF‐treated cells (Fig. 3). On day 2 of stimulation, the sum of expression was 287·39% for untreated CD8⁺ T cells compared to only 267·57% for in vitro‐treated cells and 200·45% for in vivo‐treated cells. The sum of expression was similar for untreated and in vitro‐treated cells on days 3 and 7 (290·21 versus 294·49% and 297·04 versus 306·61%), but lower for in vivo‐treated cells (176·74 and 173·11%).

Figure 3.

Granulocyte colony‐stimulating factor (G‐CSF) decreases the expression of activation markers in vitro and in vivo. Peripheral blood mononuclear cells (PBMCs) from stem cell donors treated with G‐CSF in vivo and from untreated or in vitro‐treated thrombocyte donors (n = 3) were stimulated on anti‐CD3‐coated plates for 7 days. Activation marker expression on CD8⁺ T cells was measured after 1, 2, 3 and 7 days. Sum of expression of the activation markers CD25, CD38, CD57, CD69, CD137 and human leucocyte antigen D‐related (HLA‐DR) on CD8⁺ T cells after 1, 2, 3 and 7 days of stimulation of PBMCs with (+) or without (–) G‐CSF in vitro and in vivo.

G‐CSF treatment reduces miR‐155 but not miR‐21 expression

To investigate if G‐CSF treatment alters the expression of miR‐155 and miR‐21 (both known to be regulated differentially during T cell activation), as well as several target and effector molecules, expression levels of miR‐155 and miR‐21, SHIP1, SOCS1, IFN‐γ and granzyme B were analysed after 1 and 2 days in isolated CD8⁺ T cells from untreated donors. CD8⁺ T cells were stimulated using anti‐CD3/CD28 stimulator beads and were treated with or without G‐CSF in vitro. G‐CSF reduced miR‐155 expression on CD8⁺ T cells on days 1 and 2 of stimulation, as reflected by a 48·06‐fold versus 71·33‐fold increase (+ G‐CSF versus – GCSF) on day 1 and a 70·91‐fold versus 93·43‐fold increase on day 2 (Fig. 4a). In the same experiments, the expression of SHIP1 and SOCS1 mRNA, target molecules of miR‐155, decreased on days 1 and 2 in T cells but no effect of G‐CSF on the two mRNAs was detected at these time‐points (Fig. 4b,c). The expression of miR‐21 increased slightly after stimulation but was not affected by G‐CSF treatment (Fig. 4d).

Figure 4.

Granulocyte colony‐stimulating factor (G‐CSF) treatment of isolated CD8⁺ T cells alters microRNA (miR)‐155, regulatory and effector molecule expression. Expression levels of miR‐155 (a), Src homology 2 domain‐containing inositol‐5‐phosphatase 1 (SHIP1) (b), suppressor of cytokine signalling 1 (SOCS1) (c), miR‐21 (d), interferon (IFN)‐γ (e) and granzyme B (f) were analysed after 1 and 2 days of stimulation and expansion of isolated CD8⁺ T cells from thrombocyte donors with anti‐CD3/CD28 beads treated with (+) or without (–) G‐CSF in vitro. Shown is the fold change in expression levels in treated compared to untreated cells. Bar graphs represent the mean ± standard error of the mean (SEM). Statistical significance was determined by two‐tailed paired Student's t‐test (n.s., non‐significant).

Consistent with the decreased expression of miR‐155, the expression of IFN‐γ increased 2·50‐fold in untreated T cells and only 1·65‐fold in G‐CSF‐treated cells on day 1, and increased 2·10‐fold in untreated cells versus 1·80‐fold in G‐CSF‐treated cells on day 2; the reduction on day 1 was significant (Fig. 4e). T cells treated with G‐CSF exhibited significantly lower granzyme B mRNA levels (Fig. 4f) than untreated T cells (induction of 4·61‐fold (untreated) versus 2·69‐fold (G‐CSF‐treated) on day 1 and 3·82‐fold versus 2·80‐fold on day 2, relative to non‐stimulated cells).

G‐CSF impairs ERK1/2 and CD3ζ/Lck signalling pathways

The phosphorylation of kinases and other proteins involved in T cell activation was investigated using in vivo and in vitro G‐CSF‐treated cells in order to detect differences in signalling pathways caused by G‐CSF and responsible for the decreased effector function of T cells. PBMCs from mobilized stem cell donors treated with G‐CSF in vivo and from thrombocyte donors (untreated) were stimulated with either PMA or anti‐CD3/CD28 antibodies and evaluated for phosphorylation of ERK1/2, CD247 (CD3ζ) and Lck. Overall, the highest level of phosphorylation was detected for pERK1/2 (Fig. 5). Lower frequencies determined for pLck and pCD247 resulted also in a lower difference in MFI values. After 15 min of stimulation with PMA, the percentage and MFI of pERK1/2 phosphorylation was significantly lower in CD8⁺ T cells from in vivo G‐CSF‐treated donors (76.90 versus 93·49%, MFI 1008·2 versus 2108·6, Fig. 5a). After 1 min of stimulation with anti‐CD3 and anti‐CD28 antibodies, pLck levels were significantly lower in CD8⁺ T cells from in vivo G‐CSF‐treated donors (10·46%) than in those who were not treated (43·09%, Fig. 5b). Accordingly, pCD247 levels were significantly lower in CD8⁺ T cells from in vivo G‐CSF‐treated (1·84%) than in those from untreated donors (21·89%) after stimulation with anti‐CD3/CD28 antibodies for 1 min (Fig. 5c).

Figure 5.

Phosphorylation of kinases and signalling molecules is reduced in CD8⁺ T cells from granulocyte colony‐stimulating factor (G‐CSF)‐treated stem cell donors. Peripheral blood mononuclear cells (PBMCs) from G‐CSF‐mobilized stem cell donors (in vivo G‐CSF‐treated, dark grey) and from untreated thrombocyte donors (untreated, black) were stimulated with phorbol myristate acetate (PMA) for 15 min or with anti‐CD3/CD28 antibodies for 1 min and evaluated for expression of phosphorylated (a) extracellular‐regulated kinase (ERK)1/2 (pERK1/2, n = 5), (b) lymphocyte‐specific protein tyrosine kinase (Lck) (pLck, n = 3) and (c) CD247 (CD3ζ, pCD247, n = 3). Left panel: representative two‐dimensional histograms of pERK1/2, pLck and pCD247 (isotope control, light grey). The percentage (%) and median fluorescence intensity (MFI) of pERK1/2, pLck and pCD247 on CD8⁺ T cells are shown in the left and right panels, respectively. Bar graphs represent the mean ± standard error of the mean (SEM). Statistical significance was determined by two‐tailed paired Student's t‐test (n.s., non‐significant).

Discussion

Stem cell donors are used generally as the source for clinical‐grade anti‐viral T cells for adoptive T cell transfer to prevent or treat viral infection or reactivation in immunocompromised stem cell transplant recipients. However, the recent finding that G‐CSF has a negative impact on T cell function raised the question of whether stem cell donors are suitable T cell donors. Some studies reported that monocytes, dendritic cells and also CD4⁺ T cells are influenced negatively by G‐CSF treatment. Previously, it was assumed that G‐CSF indirectly inhibits CD8⁺ effector T cell function by impairing the T cell interaction partner and increasing the secretion of immune modulatory cytokines. This study showed for the first time that G‐CSF exerts direct effects on cytotoxic effector CD8⁺ T cells, resulting in partially impaired T cell function.

G‐CSF influences CD8⁺ T cells by direct interaction

In order to avoid any effects of G‐CSF on the stimulatory signals and to prove that G‐CSF has a direct effect on antigen‐specific CTLs, we used a bead‐based method with aAPCs instead of conventional APCs (e.g. DCs) for the stimulation of T cells. The bead‐based artificial APCs were coated with anti‐CD28 mAbs, which provide co‐stimulatory signals, and with an HLA‐A*02:01 dimer, which can be loaded with HLA‐A*02:01‐restricted peptides 24. CD8⁺ T cells stimulated with A02pp65_P‐loaded aAPCs treated with or without G‐CSF expanded to an equal extent and reacted similarly to the signals provided by the artificial APCs, resulting in similar percentages of antigen‐specific CD8⁺ T cells. These results are in concordance with earlier observations by our group showing that the cells’ proliferation capacity and T cell subset composition is not altered by G‐CSF treatment 11. However, functional assays revealed that aAPC‐stimulated CD8⁺ T cells treated with G‐CSF secrete less IFN‐γ and release less granzyme B than those expanded in the absence of G‐CSF, when stimulated with A02pp65_P‐loaded target cells. In addition, the impairment of effector function by G‐CSF was confirmed further in antigen‐independent stimulation experiments by analysing IFN‐γ and granzyme B mRNA levels in isolated CD8⁺ T cells. The down‐regulation of granzymes and other effector molecules in T cells has been reported in previous studies 10, 26. G‐CSF is reported to decrease IFN‐γ and increase IL‐4 secretion in CD4⁺ T cells 27. This study also showed that the effect on CD4⁺ T cells was mediated directly and independent of G‐CSF treatment of APCs. In conformity with these findings, our own observations (Fig. S2) and several other studies showed that activated CD4⁺ and CD8⁺ T cells express the G‐CSF receptor 27, 28, 29, supporting the possibility that G‐CSF has direct effects on these cells in addition to the previously proposed indirect effects. Furthermore, we also assessed the effects of G‐CSF on the cytotoxic cell line TALL‐104. Upon stimulation with target cells, these cells are able to secrete IFN‐γ and granzyme B comparable to primary T cells 30. Pretreatment of TALL‐104 cells with G‐CSF led to a decrease in the secretion of IFN‐γ and release of granzyme B, whereas pretreatment of the target cells did not have any impact. Thus, the results reflect the effects seen in primary T cells and confirm that G‐CSF can impact directly upon cytotoxic T cells and is not required to go through the stimulating cells.

Taken together, our approach provides for the first time evidence of a direct effect of G‐CSF on cytotoxic CD8⁺ T cells. Nevertheless, this does not rule out that G‐CSF also acts indirectly through other cells, as our results showed that the effect is stronger when whole PBMCs are treated with G‐CSF. Thus, G‐CSF not only influenced CD8⁺ T cells, but also monocytes, CD4⁺ T cells and other cell types, all of which showed impaired phenotype and function after G‐CSF treatment. The synergistic effect of these impaired cell types amplifies the negative effect of G‐CSF on effector CD8⁺ T cell function.

G‐CSF affects the main elements of T cell activation

After T cells receive the four signals needed for efficient stimulation 12, 13 and are activated fully, the main elements of T cell responses are the up‐regulation of receptors on the cell surface, a change in miRNA expression profile, and the activation of signalling pathways, all of which lead to an altered gene expression pattern and the induction of effector functions. All three elements of CD8⁺ T cell activation were found to be affected by G‐CSF in the present study.

First, in 2003, Franzke et al. 28 showed that G‐CSF treatment up‐regulated mRNA expression of CD69 and CD53 but down‐regulated the expression of other co‐stimulatory molecules (CD5 and CD44) in CD4⁺ T cells. A microarray analysis by Buzzeo and colleagues 31 revealed that G‐CSF down‐regulated elements of T cell activation such as HLA class II molecules, Lck, FYN, CD3 δ‐ and ε‐chain at the mRNA level. Our own microarray analysis for isolated CD8⁺ T cells showed that G‐CSF induces down‐regulation of CD25 (1·2‐fold), CD57 (3·0‐fold), CD137 (2·0‐fold), CIITA (3·3‐fold) and other molecules (Fig. S2). On the protein level, we found that in vivo as well as in vitro G‐CSF treatment reduced the expression of the analysed activation markers on CD8⁺ T cells. However, these markers are essential for T cell activation and effector function. For example, reduced expression of the co‐stimulatory molecule CD137 results in impaired co‐stimulation. Up‐regulation of CD69 is a very early event during T cell activation, thus reduced CD69 expression indicates incomplete activation. Further, defects in expression of CD25 diminish the IL‐2 uptake of CTLs, which is essential for T cells effector function. Thus, the sum of down‐regulation of all these markers results in defective interaction with other cells, signal transmission and cytokine uptake and thereby contributes to impaired effector function.

A stronger down‐regulation of activation marker expression was observed after in vivo G‐CSF‐treatment, due perhaps to repeated G‐CSF injections over 5 days prior to the apheresis and the presence of various cell populations, the functionality of which is also influenced by G‐CSF. Recently, it was found that G‐CSF influences T cells in vitro and in vivo by various mechanisms: (a) it inhibits the secretion of type I cytokines on the single‐cell level as well as by reducing the population of cytokine‐secreting cells 32, 33, 34; (b) it induces the polarization of T cell responses towards Th2 differentiation while inhibiting Th1 proliferation 35, 36; (c) G‐CSF promotes regulatory T cells that produce suppressor cytokines IL‐10 and TGF‐β 36, 37, 38, 39; and (d) reduces the number of Th17 cells approximately threefold 10. However, little is known so far about its effects on antigen‐specific CD8⁺ T cells, one of the main defence mechanisms against viruses.

Secondly, we observed the regulation of miRNA expression in G‐CSF‐treated T cells. G‐CSF treatment has been found to be able to change miRNA expression in human haematopoietic cells. In 2014, Báez and colleagues showed that G‐CSF induces the differential expression of six miRNAs in CD34⁺ haematopoietic stem cells 22. There is also evidence that G‐CSF treatment influences miR‐155 expression in human 23 and rhesus macaque CD34⁺ haematopoietic stem cells 40.

The expression of miR‐155 in T cells is induced upon TCR stimulation and nuclear factor (NF)‐κB signalling. By repressing SHIP1, miR‐155 expression induces effector functions such as IFN‐γ secretion 14, 41, 42, 43. We showed that miR‐155 expression in G‐CSF‐treated and ‐untreated CD8⁺ T cells was induced after TCR stimulation, which is consistent with the model proposed by Liang et al. in 2015 [14]. Importantly, the induction of miR‐155 expression was lower in G‐CSF‐treated CD8⁺ T cells than in untreated cells. Although G‐CSF did not affect the expression of the miR‐155 target genes SHIP1 and SOCS1 in our experiments, miR‐155 might have other as‐yet unidentified target genes that are involved in the complex network of T cell effector functions. This is likely, as the decrease in miR‐155 expression occurred together with a decrease in effector molecule expression (IFN‐γ and granzyme B) in our experiments. Analysis of in vivo G‐CSF‐treated samples could not be performed for all investigations, because of limited material. Therefore, experiments using T cells from stem cell donors before and after mobilization should be performed to further strengthen in vitro findings.

Thirdly, the ERK1/2 pathway has been shown to be an important regulator of IFN‐γ signalling pathways in T cells 17, 18. IFN‐γ secretion is reduced strongly after disruption of signal transduction via ERK1/2. Recently, it was shown that G‐CSF induces ERK1/2 signalling in CD34⁺ haematopoietic stem cells 23 and leukaemic cells 44. In contrast, we found significantly fewer CD8⁺ T cells with phosphorylation of ERK1/2 in G‐CSF‐treated versus untreated donors. G‐CSF‐exposed cells also contained less phosphorylated ERK1/2. The defective activation of the ERK1/2 signalling pathway correlated with the results obtained from our functional assays, which showed reduced IFN‐γ production at the RNA and protein level.

CD3ζ and Lck molecules are important parts of the TCR signalling complex; they are associated with signal transmission from the TCR to downstream molecules. Down‐regulation of Lck in G‐CSF‐treated leucocytes has been shown at the RNA level 31. In conformity with this, our study showed that the percentage of phosphorylated Lck and CD3ζ decreased significantly in T cells from in vivo G‐CSF‐treated donors.

Conclusions

We show for the first time that G‐CSF treatment has a direct effect on T cell functionality. This impairment was not as strong as if the direct and indirect effects were synergistic, but was detectable at both the RNA and protein levels. G‐CSF affected negatively all elements of T cell activation analysed in this study; together, they contribute to defective IFN‐γ secretion and granzyme B release, resulting in impaired effector function. More potent inhibitory effects of G‐CSF on cytotoxic T cells were demonstrated in vivo than in vitro. Overall, the obtained results improve understanding of the impairment of CTLs and might lead to the development of a set of markers for monitoring CTL function in donors considered for T cell donation.

Disclosure

The authors declare no disclosures.

Author contributions

C. E. B. designed and performed experiments, performed data analysis and co‐wrote the manuscript. S. T. supported multimer staining in donor material, contributed with helpful discussion and helped to draft the manuscript. J. L. participated in T‐cell functional assays, contributed with helpful discussion and helped to draft the manuscript. M. O. contributed protocols for aAPC generation and critical discussions and helped to draft the manuscript. C. F. contributed helpful and critical discussions and helped to draft the manuscript. R. B., Head of Hannover Medical School's Institute for Transfusion Medicine, contributed helpful discussions and helped to draft the manuscript. B. E.‐V. conceived the study, participated in its design and coordination, and co‐wrote the manuscript.

Supporting information

Additional Supporting information may be found in the online version of this article at the publisher's web‐site:

Fig. S1. Granulocyte colony‐stimulating factor (G‐CSF) reduces surface activation marker expression on CD8⁺ T cells. Peripheral blood mononuclear cells (PBMCs) from stem cell donors treated with G‐CSF in vivo and from thrombocyte donors treated with G‐CSF in vitro or with no G‐CSF (untreated) were stimulated on anti‐CD3‐coated plates for 7 days. The percentage (%) and median fluorescence intensity (MFI) of cell surface expression of the activation markers CD25, CD38, CD57, CD69, CD137 and human leucocyte antigen D‐related (HLA‐DR) was determined on CD8⁺ T cells on days 1, 2, 3 and 7 of stimulation by flow cytometric analysis. Data represent the mean ± standard error of the mean (s.e.m.) of three independent experiments (black: no G‐CSF, dark grey: in vitro G‐CSF, light grey: in vivo G‐CSF). Statistical significance was determined by two‐tailed paired Student's t‐test.

Fig. S2. Selected results from global gene expression analysis in CD8⁺ T cells from stem cell donors. Microarray analysis was performed to compare isolated CD8⁺ T cells from stem cell donors and untreated controls (n = 4). Representative examples of the differential expression of effector molecules (a), activation makers (b) and the granulocyte colony‐stimulating factor (G‐CSF) receptor (G‐CSFR) (c) after in vivo G‐CSF treatment are shown.