PRDM1/Blimp-1 Controls Effector Cytokine Production in Human NK Cells

Abstract

NK cells are major effectors of the innate immune response through cytolysis and bridge to the adaptive immune response through cytokine release. The mediators of activation are well studied however little is known about the mechanisms which restrain activation. In this report, we demonstrate that the transcriptional repressor PRDM1 (also known as Blimp-1 or PRDI-BF1) is a critical negative regulator of NK function. Three distinct PRDM1 isoforms are selectively induced in the CD56^dim NK population in response to activation. PRDM1 coordinately suppresses release of IFNγ, TNFα and TNFβ through direct binding to multiple conserved regulatory regions. Ablation of PRDM1 expression leads to enhanced production of IFNγ and TNFα but does not alter cytotoxicity, while overexpression blocks cytokine production. Novel PRDM1 response elements are defined at both the IFNG and TNF loci. Collectively, these data demonstrate a key role for PRDM1 in the negative regulation of NK activation and position PRDM1 as a common regulator of both the adaptive and innate immune response.

Introduction

Natural killer cells play critical functions in both innate and adaptive immunity. Although these lymphocytes were initially identified by their ability to lyse leukemia cells in a non-MHC restricted manner, subsequent studies have highlighted their role in cytokine production. In response to activating stimuli NK cells proliferate, increase cytotoxicity and produce cytokines such as IFNγ, TNFα and GM-CSF (1). IL-2 up-regulates the expression of effector molecules and enhances natural cytotoxicity against a variety of targets. Furthermore, IL-2 and IL-15 both signal through the common γ_c receptor to control proliferation, with IL-15 being uniquely required for survival in vivo (2). IL-12 and IL-18 signal through distinct heterodimeric receptor complexes to elicit increases in IFNγ via several mechanisms, including increased transcription, message stability and nuclear retention (3-5). Synergistic increases in cytotoxicity and IFNγ production are observed in response to co-stimulation with IL-12 and IL-18 (6, 7).

Cytokine-mediated activation of NK cells proceeds through several well-characterized nuclear transcription factors, many of which are functionally conserved between T and NK lineages (8). STAT4 is induced in response to IL-12 and is required for optimal IFNγ production and increased cytotoxicity (9). IL-18 induces nuclear localization of NF-κB p50/p65 which, cooperatively with AP-1, increase IFNγ and cytotoxicity (10). Furthermore, NFAT induces transcription of GM-CSF and TNFα in NK cells (11). Conversely, relatively few negative regulators of activation-induced transcription have been identified in NK cells. ATF3 was recently shown to down regulate IFNγ levels and ATF^−/− mice exhibit increased resistance to MCMV infection (12). The transcription factor H2.0-like homeobox (HLX) negatively regulates IFNγ production, primarily through degradation of phosphorylated STAT4, not direct DNA-binding activity (13).

PR Domain containing 1, with Zinc Finger Domain 1 (PRDM1, also BLIMP1 or PRDI-BF1) is a transcriptional repressor encoded by the PRDM1 gene on chromosome 6q21. It was originally identified as a post-induction suppressor of IFNB in virally-infected osteosarcoma cells (14). Subsequent work revealed a pivotal role in the terminal differentiation of antibody-producing plasma cells (15). We and others have previously shown that PRDM1 exerts its repressive functions through recruitment of histone-modifying enzymes (HDAC2, G9a, PRMT5, and LSD1) and Groucho corepressors (16-18). Through silencing of direct (cMyc, CIITA, Pax5) and indirect targets, PRDM1 is a master regulator of terminal differentiation of B-lymphocytes, mediating cell cycle exit, repression of early B-cell factors and induction of immunoglobulin secretion (19, 20).

More recently, a role for PRDM1 in T-lymphocytes has emerged. PRDM1 is expressed in both CD4 and CD8 T-cell lineages and is critical for maintenance of homeostasis. Conditional knockout in T-lymphocytes leads to increased effector populations, resulting in severe colitis (21, 22). Upon activation, an auto-regulatory loop exists whereby IL-2 induces PRDM1 expression which in turn negatively regulates IL-2 transcription (23, 24). During CD4 polarization, PRDM1 is preferentially expressed in Th2 cells and reinforces commitment to this lineage through repression of Ifng, cfos and tbx21 (24, 25). Within the CD8 lineage, PRDM1 is expressed at higher levels in exhausted subsets and promotes acquisition of the effector phenotype through suppression of memory potential (26-28). Thus, in addition to well-characterized B-cell specific functions, PRDM1 is also a critical regulator of T lymphocytes. Here, we provide a functional description of PRDM1 in NK cells.

Materials & Methods

Cells and Cytokines

Primary human NK cells were isolated via negative using the EZSep kit (StemCell Technologies) according to manufacturer's instructions. Purity was verified by flow cytometry and routinely found to be 90 – 95% CD3⁻CD56⁺CD16⁺. Cells were maintained in RPMI1640 (Gibco), supplemented with 10% FBS and 1% penicillin-steptomycin. For siRNA experiments, cells were grown in Accell Delivery Media (Dharmacon), supplemented with 2% FBS. For stimulations, recombinant human cytokines were used at the following concentrations: IL-2 100U/mL (Peprotech), IL-12 10ng/mL (Peprotech), IL-18 100ng/mL (MBL), TNFγ (20–100ng/ml), IFNγ.

Mice

C57BL/6 mice (n=4) were immunized i.v. with either 250 μg poly-IC (Oncovir, Inc., Washington, DC) or PBS. Mice were sacrificed at 48 hours post-injection and single-cell suspensions were prepared from pooled splenocytes. Murine NK cells were isolated via negative selection using the Murine NK Enrichment Kit (Stemcell Technologies). Lysates were prepared from both purified NK cells and total splenocytes and analyzed by immunoblot.

Microarray hybridization and data analysis

Two micrograms of total RNA served as the mRNA source for microarray analysis. The poly(A) RNA was specifically converted to cDNA and then amplified and labeled with biotin as described (29). Hybridization with the biotin labeled RNA, staining, and scanning of the chips followed the prescribed procedure outlined in the Affymetrix technical manual. Scanned output files were visually inspected for hybridization artifacts and then analyzed using Affymetrix GeneChip Operating Software (GCOS). Heatmaps were generated using Heatmap Builder ν1.1 using signal intensity values of ≥ 3-fold differentially expressed transcripts. Transcripts with ≥ 3 fold increases/decreases are provided in Supplemental Table 1 and the complete data has been deposited at GEO (GE22919, http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE22919).

Western blotting

Whole cell lysates from 5 × 10⁵ cells were prepared in lysis buffer (50mM tris pH 7.2, 150mM NaCl, 1% NP-40, 1% Na-deoxycholate, 0.1% SDS, 2mM EDTA) on ice, sonicated and separated by 8% SDS-PAGE. Gels were transferred to PVDF, blocked 5% skim milk and probed using antibodies directed against PRDM1 (PRDI-BF1) (Cell Signalling) 1:1000, PARP1 (Cell Signaling) 1:1000 or β-actin (Sigma) 1:10,000. Secondary antibodies conjugated to HRP were used for detection: anti-rabbit (GE) 1:2000 and anti-mouse (GE) 1:10,000.

Real-time PCR

RNA was isolated using Qiagen RNeasy kit according to manufacturer's instructions. Eluted RNA was DNase-treated, converted to cDNA and 1/20^th of the cDNA reaction analyzed by real-time PCR in duplicate using the BioRad iCycler (40 cycles, with primer-specific annealing temperatures between 55°C and 60°C). For expression analysis, data was analyzed by ΔΔCt method, with normalization to either 18S or GAPDH. Primers were quality checked for single curve on melt curve and efficiencies between 90% and 110%. Sequences of primers (PRDM1a, PRDM1b, IFNG, TNF, LTA, DAP10, CIITAp3, CIITAp4, PAX5, IFNB1, SLAMF7, MBexon2, IFNG I, II, III, IV, TNF I, II, III, IV) and T_m values are provided in Supplemental Table 2.

⁵¹Chromium release assay

NK cells isolated by negative selection were cultured for 72hrs prior to the assay. Cells were incubated with 5,000 K562 cells at various effector : target ratios in triplicate in a total volume of 100μl for 4hrs. Supernatants were harvested and cpm determined using a Perkin Elmer 1470 automatic gamma counter. Percent cytotoxicity was calculated using the following formula: (experimental – spontaneous release)/(total – spontaneous release) × 100.

Chromatin immunoprecipitation

20 × 10⁶ purified NK cells were stimulated for 24 h with IL-2 (100U/mL), IL-12 (10ng/mL), and IL-18 (100ng/mL). Chromatin was prepared as previously described (30). 4.5 × 10⁶ cell equivalents were used for each immunoprecipitation reaction. Primary antibodies were used at 0.5μg in each 900μl reaction and incubated overnight. Antibodies used were: PRDM1 (PRDI-BF1) (Cell Signaling) and normal rabbit IgG (Upstate). Immune complexes were captured with protein A/G beads (Santa Cruz) and washed as described. Eluted DNA was column purified (Qiagen) after reversal of crosslinks (4 h 65°C), RNase treatment and proteinase K treatment. PCR was performed using 1.5μl of eluted DNA (~1/40^th) in duplicate. Primers designed against the second exon of Myoblobin B (Diagenode) were used as a negative control locus. Percent input was calculated by linearization of ΔCt (C_tIP – C_t1%input) for both specific and IgG samples.

ELISA

ELISA assays were performed in 100μl volumes in triplicate using commercial kits for IFNγ and TNFα according to manufacturer's instructions (E-biosciences). Plates were read at 450nm on a Molecular Devices Spectramax Plus plate reader.

Adenoviral constructs and transduction

Adenoviral constructs were created in the Ad5/F35 vector from previously described constructs (16). This replication-deficient adenovirus utilizes the ubiquitously-expressed CD46 molecule to mediate high level transduction efficiency in a variety of hematopoetic cells (31). Briefly, PRDM1a in the AdTrack vector was recombined with Ad5/F35 to generate a bi-cistronic viral vector containing PRDM1a and GFP. Purified stocks were obtained by infecting Ad293T cells for 48hrs and concentrating via CsCl banding according to standard protocols. Viral titers were calculated using the CellBio Labs QuickTiter Adenovirus Quantitation Kit. For transduction experiments, Jurkat cells were transduced using an MOI of 500 at a density of 1×10⁶/ml for 44 hours before stimulation with PMA (1μg/ml) and PHA (10ng/ml) for 4hours.

DNA constructs and luciferase assay

A fragment of the human IFNG promoter (−507 - +120) was PCR cloned from human genomic DNA into pcr2.1. The fragment was then subcloned into pGL3basic using the HindIII and SmaI restrictions sites to generate pGL3-IFNγ_WT. pGL3-IFNγ_mut was obtained via site-directed mutagenesis of the -254 site, changing residues AAAAGT to TCTAGA, which created a novel XbaI site (Mutagenex, Inc). pGL3-IFNγ_Δ was obtained by ligation of an XbaI-XbaI fragment obtained from pGL3-IFNγ_mut with the XbaI-NheI fragment of the pGL3-Basic vector. Transfections were performed using 20μg of total plasmid into 10⁷ cells by elctroporation at 250V, 1070μF in 300μl RPMI using a BioRad Gene Pulser II. Cells were cultured 36 h at 10⁶/ml. Cells were lysed in 500μl Passive Lysis Buffer and assayed using the Dual Luciferase kit per manufacturer's instructions (Promega).

Statistical Analyses

For statistical analyses, two-tailed paired t-tests were used. P-values less than 0.05 were considered significant. All calculations were performed in Microsoft Excel.

Results

Human NK cells alter expression of multiple effector molecules and transcription factors in response to cytokine stimulation

Natural killer cells are well known for their ability to up-regulate both the production of effector cytokines and cytotoxic potential in response to IL-2 and other cytokines such as IL-12, IL-15, and IL-18. However, the roles of sequence-specific, DNA-binding transcription factors in the modulation of NK activity are incompletely characterized. To directly address this we isolated NK cells from peripheral blood of healthy donors and performed global gene expression profiling on RNA isolated immediately or after 24 hours stimulation in the presence of IL-2, IL-12 and IL-18 (Fig. 1). Purity of the NK population was determined by flow cytometry to be greater than 96% and this was further confirmed by the absence of significant signals for transcripts associated with B-cells, T-cells, and monocytes in the microarray (Fig. 1b). Biological reproducibility was extremely close giving a R² value of .967 and .970 between donors in freshly isolated and stimulated samples, respectively (Sup. Fig. 1). In total, 541 genes were increased 3-fold or more in both donors, while 609 genes were decreased 3-fold or more following 24 hour stimulation (Fig. 1a and Sup. Table 1). As expected, effector cytokines (e.g. IFNG, TNF and CSF1) and numerous TNF family members were found to be transcriptionally up-regulated in response to stimulation (Fig. 1c). Decreased levels of TGFB1, which is constitutively expressed in NK cells, were also observed in response to stimulation, consistent with its role as a negative regulator of activation/proliferation.

Figure 1.

Human NK cells activate effector molecules and transcriptional regulators in response to cytokine stimulation. (a) Heat-map depicting top 75 genes found to be up- or down-regulated at least 3-fold in response to 24h stimulation with IL-2 (100U/mL), IL-12 (10ng/mL) and IL-18 (100ng/mL) relative to time 0h. Pixel density (highest values in each are pure black, lowest are white) represents average hybridization signal intensity from two donors after and before stimulation for increased and decreased genes, respectively. Fold changes are shown in parentheses. Presentation of expression data in this manner allows appreciation of both magnitude of expression and increases/decreases in response to stimulation. (b) Average signal intensities of transcripts associated with “non-NK” lineages are shown from the 0h samples to demonstrate low frequency of contaminating cells. (c) Heat-map depicting top 15 up and down-regulated DNA-binding transcription factors as in a. (d) Average signal intensities of selected secreted cytokines and tumor necrosis factor family members. Raw data has been deposited at GEO (accession # GE22919).

Of the top 150 transcripts found to be modulated, nearly 10% were sequence-specific DNA-binding transcription factors (Fig. 1d). Among these factors, the transcriptional repressor PRDM1 was increased in response to stimulation. PRDM1 has not been previously identified in NK cells, however, within the immune system PRDM1 plays crucial roles in both cell fate decisions and regulation of homeostasis. The role of PRDM1 in the terminal differentiation of mature B-cells into immunoglobin-secreting CD138⁺ plasma cells is well established. More recently, roles for PRDM1 in the maintenance of homeostasis and effector versus memory lineage commitment in T-lymphocytes have been reported (26-28). This suggests that PRDM1 may also have key functional roles in NK cells.

Cytokine stimulation induces multiple PRDM1 isoforms in human NK cells

In order to directly establish PRDM1 activation in human NK cells, cells were isolated by negative selection and stimulated with multiple combinations of IL-2, IL-12, and IL-18. Immunoblot analysis reveals that freshly isolated and unstimulated NK cells have barely detectable levels of PRDM1 protein (Fig. 2a). Stimulation with IL-2 or a combination of IL-12 and IL-18 results in increased levels of PRDM1 protein which is markedly enhanced by costimulation with all three cytokines. Cell lines derived from malignant NK tumors constitutively express PRDM1 and exhibit distinct isoform expression patterns (Fig. 2b). YT and NK92 cell lines exclusively express the larger α-isoform, while NKL most closely resemble the expression pattern observed in primary human NKs.

Figure 2.

Cytokine stimulation induces PRDM1 isoforms in human NK cells. (a) Western blot analysis of purified human NK cells stimulated for 24h with IL-2 (100U/ml), IL-12 (10ng/ml) and IL-18 (100ng/ml) or combination. Blot shown is representative of numerous experiments. (b) RT-PCR analysis of cDNA synthesized from freshly isolated or 24h stimulated NK cells. Expression values were calculated using the ΔΔC_t method with 18S as the control gene. Upon linearization of C_t values, average Time 0 PRDM1α was arbitrarily set to “1” to account for basal differences among donors and the consistently higher expression levels of PRDM1β. Error bars represent standard deviation from at least three biologically independent samples isolated from different donors at different times. Paired two-tailed t-tests were conducted, with (*) representing p < .05 p-values were 0.016 and 0.023 for PRDM1α and PRDM1β, respectively. (c). Immunoblot analysis of lysates prepared from human NK cell lines demonstrates differential expression of PRDM1 isoforms. (d) CD56^dim and CD56^bright subsets were obtained from IL-12 and IL-18 stimulated human NK cells by FACs sorting as previously described. RT-PCR analysis was conducted on cDNAs prepared from two donors as above. (e) Immunoblot analysis of lysates prepared from CD56^dim and CD56^bright subsets as in panel d.

Primary NK cells express three distinct molecular weight forms of PRDM1. The largest molecular weight band corresponds to the full-length PRDM1α isoform, however this protein is significantly underrepresented compared to the smaller PRDM1 proteins. The predominant and smallest form corresponds to the PRDM1β originally characterized in multiple myeloma tumor cells and shown to have partially reduced repressive ability in reporter assays (32). In addition, immunoblot analysis detected an intermediate-sized isoform in primary NK cells. This protein migrating at ~85kD likely represents the human homologue of a previously-described murine splice variant, prdm1Δexon7 (33). Indeed, RT-PCR analysis using primers spanning the analogous exon in humans (exon 6) confirms the presence this splice variant (Sup. Fig. 2). Consistent with our immunbloting experiments, analysis of the mRNA levels specific for the α and β isoforms indicate that cytokine stimulation up-regulates both isoforms and that PRDM1β mRNA is present at approximately 20 fold higher levels than PRDM1α (Fig. 2c).

Human NK cells can be divided into two subsets based on the surface density of CD56 and the presence of CD16. The CD56^brightCD16^dim/− subset represents ~10% of the human peripheral NK cell compartment and has been suggested to play a immuno-modulatory role based on the increased ability to produce cytokines; the CD56^dimCD16⁺ subset represents ~90% of human peripheral NK cells and is considered the primary cytotoxic subset. We questioned whether PRDM1 is restricted to specific subsets or if it is broadly expressed in NK cells. Human NK cells were isolated from healthy donors via negative selection followed by flow cytometric sorting into CD56^bright and CD56^dim subsets. The cells were then stimulated with IL-12 and IL-18 for 24hr before analysis. Immunoblot and qRT-PCR analysis reveals that PRDM1 is preferentially expressed in the CD56^dim subset in response to stimulation (Fig. 2d,e). Furthermore, PRDM1β is the predominant isoform present.

PRDM1 associates with NK activation

Given that PRDM1 was maximally induced in response to co-stimulation with IL-2, IL-12 and IL-18, we hypothesized that levels of PRDM1 correlated with degree of activation. We profiled mRNA levels of the effector cytokines IFNγ and TNFα in purified human NK cells after 24 hour stimulation with various combinations of cytokines. Quantitative RT-PCR analysis indicates that IFNγ and TNFα mRNA are synergistically increased upon stimulation with IL-2, IL-12 and IL-18 stimulation (Fig. 3a). Consistent with this, IL-2, IL-12 or IL-18 alone each minimally alter PRDM1 expression (Fig. 3b). Thus, PRDM1 levels correlate with effector cytokine transcription.

Figure 3.

PRDM1 expression associates with NK activation. (a) Immunoblot analysis of purified human NK cells stimulated for 24h with IL-2 (100U/ml), IL-12 (10ng/ml) and IL-18 (100ng/ml) alone or in combination. (b) RT-PCR analysis of RT-PCR analysis of cDNA synthesized from freshly isolated or 24h stimulated NK cells. Expression values were calculated using the ΔΔC_t method with 18S as the control gene. Upon linearization of C_t values, average Time 0 for each donor was arbitrarily set to “1”. Error bars represent standard deviation from at least three biologically independent samples isolated from different donors at different times. (c) Immunoblot analysis of lysates prepeared from purified NK cells treated with recombinant human TNFα (20ng/ml or 100ng/mL), IFNγ (10ng/ml or 50ng/ml) or IL-2 (100U/ml), IL-12 (10ng/ml) and IL-18 (100ng/ml) for 24h.

We next assessed whether effector cytokines induced PRDM1 through autocrine or paracrine feedback mechanisms. Treatment with either recombinant IFNγ or TNFα fails to induce PRDM1. (Fig. 3c). Similarly, stimulation with α-interferon is also insufficient to induce PRDM1, suggesting that induction results primarily from cytokine receptor-mediated signaling and likely requires multiple signaling events.

To determine if PRDM1 is induced in vivo, we activated murine NK cells by polyI:C injections into C57/BL6 mice. PolyI:C is known to activate NK cells indirectly through TLR3-mediated release of cytokines from dendritic cells and other accessory cells. NK cells were purified by negative selection from the spleens of naïve and polyI:C treated mice. PRDM1 is robustly induced specifically in NK cells upon 48 hour treatment (Fig. 4). PRDM1 expression was also detected in total splenocytes and the level did not change with polyI:C treatment. This is likely due to the presence of PRDM1 in T cells and differentiating plasma cells within the spleen. Interestingly, only the PRDM1α homologous isoform was detectable in murine NK cells. Although an inability of the antibody to cross react with other murine isoforms cannot be excluded, it is possible that the expression and/or function of PRDM1 isoforms in murine cells may be differentially regulated.

Figure 4.

PRDM1 is induced in vivo. C57/BL6 mice were treated with 100μg polyI:C via tail vein injection for 48hrs. NK cells from total splenocytes (right panel) or purified via negative selection (left panel) and analyzed by immunoblot analysis.

PRDM1 is not involved in perforin-mediated cytotoxicity

The ability to lyse target cells in an antigen-independent manner is a hallmark of NK cells. Cytotoxicity against heterologous target cells proceeds through the release of perforin and granzymes and is markedly increased upon treatment with IL-2, IL-12 or IL-18 (1, 34). To assess whether PRDM1 regulates cytotoxicity, NK cells were isolated from healthy donors and stimulated with either IL-2 and IL-12 or the combination of IL-2, IL-12 and IL-18 in the presence of either a non-targeting control or a PRDM1-specific siRNA for 72 hours. Cytotocixity against the K562 leukemia cell line was assessed in a 4 hour ⁵¹Chromium release assay. As expected, cytokine-activated NK cells exhibit significant cytotoxicity against K562 targets (Fig. 5a). Knockdown of PRDM1 expression did not alter cytolytic activity across several effector:target ratios. Distinct from our experiments that revealed increases in IFNG and PRDM1 mRNA levels with addition of IL-18 to IL-2 and IL-12, we observed no additive effect of IL-18 in cytotoxicity assays, further demonstrating that cytotoxicity and PRDM1 levels are not directly linked. Knockdown of PRDM1 protein was confirmed by immunoblotting to be highly efficient and had no effect on viability as assessed by trypan blue staining (Fig. 5b, data not shown). Furthermore, no activation-induced cell death was detected in stimulated NK cells as measured by PARP cleavage (Fig. 5c). Collectively, these results show that PRDM1 does not have a significant role in regulating perforin-mediated cytotoxicity in NK cells.

Figure 5.

Perforin-mediated cytotoxicity is not affected by PRDM1 knockdown. (a) 4 h ⁵¹Chromium release assays were performed in triplicate from two donors. Purified NK cells were incubated for 72 h in the presence of either a non-targeting (NT) or PRDM1-specific (KD) siRNA and stimulated with combinations of IL-2 (100U/ml) and IL-12 (10ng/ml), with or without IL-18 (100ng/ml). % ⁵¹Chromium release was calculated as described in Methods. Error bars represent s.d. of triplicate wells. (b) Western blot analysis to confirm knockdown in cells used in cytotoxicity assay. IL-2-only sample was not assayed in cytotoxicity assay and is only included to avoid cropping of gel. (c) Purified NK cells were incubated for 48 h or 72 h in the presence of either a non-targeting (NT) or PRDM1-specific (KD) siRNA. Cells were stimulated with combinations of IL-2 (100U/ml), IL-12 (10ng/ml) and IL-18 (100ng/ml) and analyzed by western blot for PRDM1 induction and PARP cleavage.

PRDM1 binds promoters in multiple target genes

As PRDM1 is well-documented as a DNA-binding transcriptional repressor we sought to characterize its function in NK cells by identifying DNA elements to which PRDM1 specifically binds. To accomplish this we performed chromatin immunoprecipitation (ChIP) experiments in primary NK cells isolated from healthy donors. Cells were stimulated for 24 hours with IL-2, IL-12 and IL-18 to induce PRDM1 expression prior to isolation of chromatin and immunoprecipitation with either anti-IgG or anti-PRDM1 antibodies. Initially, we assayed promoter regions of genes known to be regulated by PRDM1 (Fig. 6a). A well-documented target is CIITA, the master regulator of MHC class II expression which is transcriptionally silenced as mature B-cells differentiate into antibody-producing plasma cells. We observed PRDM1 binding at the IFNγ-inducible CIITApIV promoter, but not the lymphoid-specific CIITApIII promoter. Although human NK cells can increase surface MHC Class II expression in response to activation (35), they express low to undetectable levels of CIITA. Binding of PRDM1 to the CIITApIV promoter may function to reinforce this repressed state and prevent IFNγ-mediated autocrine induction. We did not detect binding at the promoters of PAX5 and IFNB, both of which have previously been shown to be directly regulated via binding of PRDM1 to promoters in B-cells and osteosarcoma cell lines, respectively. Furthermore, we did not detect binding at the promoter of SLAMF7, which encodes the NK activating receptor CRACC and contains a potential PRDM1 binding motif within its proximal promoter. This indicates that PRDM1 binds target gene promoters in a selective and cell type specific manner. As expected, no binding was detected using primers to the second exon of Myoglobin B, which was used a negative control.

Figure 6.

PRDM1 binds promoters in multiple target genes. (a) Chromatin immunoprecipitation experiments were performed with chromatin isolated purified NK cells stimulated for 24 h with IL-2 (100U/ml), IL-12 (10ng/ml) and IL-18 (100ng/ml). Induction of PRDM1 was confirmed by western blot. Percent input was calculated as described in Methods. Error bars represent s.d. from three biologically independent experiments performed at different times with chromatin preparations from different donors. Two-tailed t-tests between IgG and PRDM1 immunoprecipitated samples, * p-value = 0.035, ** p-value = 0.0009. (b,c) ChIP was performed across the IFNG (b) or TNF (c) locus with primers specific regions specified. Locations of primers and genes in each locus are included in diagrams below each graph. Note that scales are different for each locus.

In murine T-cells, Blimp-1 has recently been shown to bind to a distal, conserved regulatory site within the Ifng locus in in vitro polarized CD4+ T_h1 lymphocytes (25). As NK cells are potent producers of IFNγ during the early innate immune response and induction is correlated with PRDM1 expression, we sought to evaluate PRDM1 binding to the human IFNG locus in cytokine-stimulated primary NK. We assayed four locations across the IFNG locus, which are highly conserved between rodents and humans and have been demonstrated to regulate IFNG expression (36). We detected binding at the distal regulatory site, which has been demarcated as CNS −22 based on the genomic distances relative to the transcriptional start site in the mouse (Fig. 6b). This site is analogous to the PRDM1 binding site detected in T cells. Additionally, through bioinformatic analysis we identified two potential PRDM1 binding sites within the minimal promoter located 370 base pairs and 254 base pairs upstream of the IFNG transcriptional start site. Very robust PRDM1 binding was detected in this proximal promoter region which has not been identified in T cells. We did not detect binding at CNS −6 or CNS +18-20, neither of which bear potential PRDM1 binding sites. Collectively, these data demonstrate that PRDM1 binds both proximally and distally to multiple sites across the IFNG locus in human NK cells.

TNF is coordinately induced with IFNG upon cytokine stimulation of NK cells. Thus, we sought to determine if this locus was also bound by PRDM1. To this end, we assayed four distinct locations across the ~12kB TNF locus by chromatin immunoprecipitation. The TNF locus contains 3 genes, each of which contains four exons (Fig. 6c). LTA and TNF are separated by ~1kB and transcribed from the same strand while LTB is transcribed from the opposite strand and separated from TNF by ~3kB. We measured PRDM1 binding at the proximal promoter regions of LTB and TNF, the intergenic region between TNF and LTB, and a recently identified regulatory site located ~3.5kB upstream of LTA. This upstream enhancer element coordinately regulates the LTA and TNF genes but not the opposing LTB gene (37, 38). PRDM1 binding was clearly detected at this upstream regulatory site, but not at either proximal promoter or the intergenic region. Furthermore bioinformatic analysis detected a consensus PRDM1 recognition sequence in the bound enhancer element. These data indicate that PRDM1 specifically associates with defined regulatory sequences of the TNF locus.

Blockade of cytokine-induced PRDM1 expression increases effector cytokine production

Our experiments demonstrate that PRDM1 is coordinately induced with effector cytokines upon stimulation and occupies specific regulatory regions within the IFNG and TNF loci. To investigate the functional effects of PRDM1 in NK cells, we performed gene expression knockdown experiments using primary human NK cells. Cells were stimulated with IL-2 and IL-12 in the presence of either a non-targeting control or PRDM1-specific siRNA. mRNA was isolated after 48 hours and analyzed by real-time RT-qPCR. We found significantly higher mRNA levels of IFNG, TNF, and LTA when PRDM1 induction was abrogated via siRNA, yet no significant differences were found for DAP10 (Fig. 7a). We next sought to determine whether these increases corresponded with detectable changes in protein expression. Secreted IFNγ and TNFα were measured by ELISA and consistent increases in secreted protein were detectable in multiple donors in response to PRDM1 expression knockdown (Fig. 7b). Consistent with this silencing, PRDM1 expression remains elevated over a four day post-stimulation time course, while IFNG and TNF levels decline (Sup. Fig. 3). Together, these data indicate that PRDM1 negatively regulates production of these cytokines in response to NK cell activation.

Figure 7.

Blockade of cytokine-induced PRDM1 expression increases effector cytokine production. (a) Purified NK cells were incubated for 48 h in the presence of either a non-targeting (NT) or PRDM1-specific (KD) siRNA and stimulated with IL-2 (100U/ml) and IL-12 (10ng/ml). RT-PCR was conducted for several genes on cDNA synthesized from RNA isolated from 3 different donors at different times using 18S control gene. For each donor, NT was normalized to “1”. Data are presented as mean, with error bars representing standard error from 3 biologically independent experiments. (b) ELISA assay was conducted on supernatants harvested at 48 hours from samples above. Data represent mean of three independent wells. For each analyte, two-tailed paired t-tests demonstrated p-value < 0.05 among all three donors for NT vs. KD.

Overexpression of PRDM1 mediates repression of activation-induced expression of IFNG and TNF

Because we observed occupancy of PRDM1 at effector cytokine loci and increased production in the context of PRDM1 knockdown, we wanted to assess whether PRDM1 was capable of blocking activation-induced transcription of IFNG and TNF. The Jurkat T cell line was transduced with adenovirus expressing either GFP alone or GFP and PRDM1α prior to stimulation. After PMA/PHA stimulation, RT-qPCR analysis was performed to assess induction of IFNG and TNF. GFP-transduced cells showed significant up-regulation of IFNG and TNF upon stimulation, which was nearly abolished in PRDM1α-transduced cells (Fig. 8a). Thus, introduction of PRDM1 prior to stimulation was sufficient to block stimulation-induced transcription of IFNG and TNF, providing further support that PRDM1 is a negative regulator of effector cytokine production.

Figure 8.

Ectopic Expression of PRDM1 diminishes expression of target genes. (a) Jurkat cells were transduced for 40 h, and then stimulated with PMA (50 ng/ml) and PHA (1 ug/ml) for 4 h prior to RNA isolation. After correction using 18S, relative expression for IFNG and TNF were determined by assigning a value of 100% to the stimulated GFP-transduced treatment. Data are presented as mean of three independent experiments, with error bars representing s.d., * p-value .0002, ** p-value .0005. (b) Luciferase assays were conducted in Jurkat cells, 36 h post-transfection. Data are presented as the mean of four independent transfections +/− s.d., with each reporter transfected with empty vector set to 100%. Two-tailed paired t-tests showed p-value of .005 for wild-type, indicated by *.

PRDM1 mediates repression of IFNG via elements in the proximal promoter

To directly assess the functionality of PRDM1 binding to the newly identified elements within the proximal promoter of the IFNG gene, we cloned the region comprising −507 to +121 of the human IFNG gene and inserted it upstream of a luciferase reporter. This promoter region contains the two potential PRDM1 binding sites, located at −370 and −254. Co-transfection of PRDM1 was sufficient to repress luciferase activity driven by the wild type IFNG promoter by approximately 50% (Fig. 8b). To assess the relative contributions of the two potential PRDM1 binding motifs, a deletion of approximately 300 base pairs encompassing both of these sites (ΔIFNG-pGL3) was created. This deletion eliminated the ability of PRDM1 to repress transcription from the construct. This repression was similarly eliminated by point mutations at the −254 site (mutIFNG-pGL3) demonstrating that this location is critical for PRDM1 mediated repression. Collectively, these data demonstrate PRDM1 mediated repression of IFNG through binding to this previously uncharacterized motif.

Discussion

In this report we provide the functional description of PRDM1 in natural killer cells. We show that PRDM1 accumulates upon cytokine-mediated activation and acts as a negative regulator of activation, attenuating inflammatory cytokine production. Such a negative feedback loop has important implications in the context of inflammation and immune homeostasis.

Using global gene expression profiling we identified PRDM1 as a highly expressed transcription factor in stimulated NK cells. Our data complement previous studies (39-42) providing the first global gene expression profile of NK cells utilizing a physiologically-relevant combination of cytokines to achieve activation. During the early phases of the innate immune response IL-12 and IL-18 are produced by monocytes and epithelial cells, while IL-2 is primarily produced by activated T-cells, typically a later event. Another early monocyte-derived cytokine, IL-15, signals through the shared IL-2 receptor and can substitute for IL-2 to induce PRDM1 in the presence of IL-12 and IL-18 (unpublished data). Consistent with this, maximal induction of murine NK cells in response to Salmonella-infected macrophages has recently been shown to require: 1) IL-2 and/or IL-15, 2) macrophage-derived IL-12 and IL-18 and 3) direct NK-macrophage cell contact (6). Thus the exposure of freshly isolated NK cells to these cytokines ex vivo can mimic the milieu present in vivo during the early innate immune response and can provide insight into the physiological gene expression changes occurring.

The ability of NK cells to produce a variety of cytokines in response to stimulation provides a crucial mechanism of cross-regulation between the innate and adaptive arms of the immune system. In order to maintain homeostasis, regulatory control must be exerted not only upon the activation phase, but also during recovery phase to restrain the degree of activation. While activation events have been thoroughly investigated, molecular events regulating dampening or recovery remains poorly characterized. The studies in this report now show that PRDM1 is an important mediator of this phase. PRDM1 attenuates production of multiple effector cytokines in a coordinate manner via direct binding to specific DNA sequences in known regulatory regions. At the TNF locus, PRDM1 binds specifically to a highly conserved regulatory element ~3.5kB upstream of the transcriptional start site of LTA. This region exhibits DNaseI hypersensitivity and contains binding sites for inflammatory activators such as NFκB and NFAT (37). Furthermore, this site has been shown to be crucial for activation-induced looping (38). This three-dimensional looped chromatin structure mediates interaction between distally- and proximally-bound NFAT at the TNF promoter forming an enhanceosome resulting in transcriptional activation of TNF and (to a lesser extent) LTA in response to stimulation. Consistent with the notion that this region imparts specific, localized control at the TNF locus, knockdown of PRDM1 results in increases in both LTA and TNF, but not LTB which is located further downstream and encoded on the opposite DNA strand.

The IFNG locus is also subject to PRDM1-mediated regulation in NK cells. Studies in the T cell compartment identified an upstream regulatory region in murine CD4⁺ T-cells (CNS-22) with enhancer activity which contributed to Tbet-dependent Ifng expression in in vitro polarized Th1 cells (36). This site was later shown to be bound by the murine homologue Blimp1 (25). In our study, we have confirmed the functionality of this distal site in human NK cells. In addition to this distal element, we have identified a novel promoter-proximal element within the IFNG promoter to which PRDM1 binds. This novel proximal site is required for silencing of a reporter gene driven by the human IFNG promoter. Freshly isolated, resting NK cells differ substantially from T-cells with respect to the kinetics of IFNG mRNA induction. Indeed, NK cells induce IFNG rapidly upon stimulation without the requirement for chromatin modifications because the entire locus is poised in a hyperacetylated state under basal conditions (43). Thus, PRDM1 binding directly upstream of the transcriptional start site may facilitate potent silencing in NK cells without necessarily requiring long-range chromatin modifications

Human NK cells exist in at least two functionally divergent subsets based on the surface density of CD56 and CD16. Several groups have demonstrated that CD56^dimCD16⁺ NK cells are the major population in peripheral blood, while CD56^brightCD16^−/dim NK cells represent less than 10% of peripheral NK cells and primarily localize to lymph nodes (44). The CD56^bright population has been described as a “regulator” population. These cells express the high-affinity IL-2 receptor, are highly proliferative in response to stimuli and localize at inflammatory sites. Accumulating evidence suggests that CD56^bright cells may, in fact, be developmental precursors to the more mature CD56^dim “effector” population (45, 46). Conversely, the CD56^dim “effector” population exhibits increased natural cytotoxicity, but produces lower levels of inflammatory cytokines relative to the CD56^bright population. Consistent with its role as a negative regulator of cytokine production, we observe that PRDM1 is preferentially expressed in the CD56^dim population. Thus, PRDM1-mediated transcriptional repression likely contributes to the functional divergence observed in these populations.

NK cells exhibit a unique pattern of PRDM1 protein isoform expression. In addition to the full-length PRDM1α, NK cells express high levels of two smaller molecular weight species. Expression of the PRDM1β-isoform has previously only been documented in myeloma cell lines and samples from myeloma and T-cell lymphoma patients, and is always present at lower levels than the full-length protein (32, 47). NK cells represent the first non-transformed cell type with PRDM1β expression and furthermore it is consistently observed at higher levels than the larger PRDM1α. The β-isoform is transcribed from an alternate promoter and utilizes a distinct transcriptional start site present in the third intron of the full-length protein. The resulting protein has a disruption of the highly conserved PR-domain. We have previously shown that the β-isoform localizes to the nucleus and maintains capacity for DNA-binding, albeit, it's repressive activity is dampened relative to PRDM1α. All of the cytokine combinations tested activate transcription of both isoforms in NK cells, although there remains the potential that isoform-specific activation signals may exist. We also consistently observe an intermediate-sized isoform (PRDM1αΔ6) corresponding to a deletion of the amino acids encoded by exon 6. This isoform is generated via splicing to exclude exon 6 resulting deletion of the 2^nd zinc finger as well as disruption of a portion of the 1^st and 3^rd zinc finger. An analogous splice variant was recently described in naive CD19⁺ murine B-cells although it was present at very low levels (33). The first two zinc fingers of PRDM1 have a role in DNA binding and are required for recruitment of the histone methyltransferase G9a, but dispensable for interaction with HDAC2 (16, 48). Interestingly, we do not observe induction of apoptosis or evidence of cell cycle exit concomitant with PRDM1 induction in NK cells. Furthermore, IL-2-expanded LAK cells continue to proliferate despite acquisition of high level PRDM1 expression (unpublished data). Thus, the presence of these lower molecular weight isoforms which have altered or impaired activity may provide a cytoprotective effect, whereby NK cells are protected from the anti-proliferative and apoptotic effect of PRDM1 but retain the ability to repress selective target genes. Characterization of the contribution of each isoform is an important area of future investigation.

The PRDM1 allele is encoded on chromosome 6 within a region that is known be frequently associated with B, T and NK-derived malignancies. A recent report used high resolution comparative genomic hybridization arrays to precisely map regions deleted within NK lymphomas and identified a minimal common region spanning approximately 2Mb of the 6q21 deletion, which is present in roughly 40% of cases (49). Within this region, the authors identified three genes which were down-regulated in tumor specimens harboring the 6q21 deletion: ATG5, PRDM1 and AIM1. More recently, PRDM1 expression levels were shown to be highly variable and found to be independent of the 6q21 deletion in an independent study of six primary NK lymphoma patient samples (50). Our data clearly establishes a functional role for PRDM1 in normal NK cell function. Thus, while PRDM1 can act as a tumor suppressor in Diffuse Large B Cell lymphoma, its contribution to the transformed phenotype in NK-derived malignancies remains controversial.

Our studies add NK cells to the expanding list of immune lineages in which PRDM1 has a key functional role. To date PRDM1 has been described in plasma cells, multiple T-cell subsets, dendritic cells and myeloid cells. Although lineage-specific regulation of target genes exists, a significant commonality is that PRDM1 regulates final effector function. For instance, negative regulation of Il2 and tbx21 by Blimp1 promotes Th2 polarization in CD4⁺ T-cell differentiation, yet differential expression of these genes does not play a role in either B or NK effector function. Coordinate regulation of IFNG and TNF, on the other hand, is critical to both T and NK effector function. Characterization of overlapping and non-overlapping functions of PRDM1 across multiple lineages will be an important area of focus in future experiments. For example, Lanier and colleagues recently provided evidence for a “memory” phenotype in murine NK cells using an MCMV infection model wherein they demonstrated that antigen-specific Ly49H⁺ NK cells underwent a prolonged contraction period, but were subsequently activated to higher levels upon re-stimulation (51). PRDM1-mediated restriction of memory potential has recently been described CD8⁺ T-cells (26). Given that PRDM1 is induced in vivo during NK activation, it will be interesting to ascertain if PRDM1 has such a role in NK memory commitment.

In summary, our data suggest that PRDM1 plays a crucial role in the post-activation phenotype of NK cells by negatively regulating cytokine transcription in a coordinate manner, without compromising perforin-mediated cytotoxicity or inducing exit from cell-cycle. Such a mechanism may have important implications in innate immunity and tumor surveillance.