Human umbilical cord blood samples were obtained at birth after full-term delivery from the Department of Obstetrics and Gynecology of the University Hospital of Vienna, Austria. Cord blood mononuclear cells (CBMCs) were collected by Ficoll density gradient centrifugation (Lympholyte Cell Separation Media, human, Cedarlane^®, Burlington, ON, Canada). Stem cells were isolated from CBMCs according to manufacturer’s protocol, using a magnetic bead-based CD34⁺ isolation kit (CD34 MicroBead Kit, human; Miltenyi Biotec, Bergisch Gladbach, Germany). The purity of the stem cells was evaluated by flow cytometry (CD34⁺ reached at least 95% purity) after which the cells were cultured as previously described (28, 30). In short, stem cells at an initial density of 10⁵ cells/ml were seeded into 6-well plates (Corning Incorporated, Corning, NY, USA) for 10 days in basal expansion medium (GBGM©; Glycostem Therapeutics, Oss, The Netherlands) supplemented with stem cell factor (SCF), interleukin-7 (IL-7), thrombopoietin (TPO), and FMS-like tyrosine kinase 3 ligand (Flt3L); all factors at a concentration of 25 ng/ml (CellGro^®, CellGenix GmbH, Freiburg, Germany) and granulocyte colony-stimulating factor (G-CSF, 250 pg/ml; Stemcell Technologies, Vancouver, BC, Canada), granulocyte-macrophage colony-stimulating factor (GM-CSF, 10 pg/ml; Stemcell Technologies), and IL-6 (50 pg/ml; CellGenix GmbH). At day 10, TPO was replaced with IL-15 (20 ng/ml; CellGro^®, CellGenix GmbH), and at day 14, Flt3L with IL-2 (1,000 U/mL; Chiron Corporation, Emeryville, CA, USA) by refreshing half of the medium. After a total culture period of 35 days, a regular culture consisted of >95% CD56⁺CD3⁻ NK cells as evaluated by flow cytometry.

Total RNA was extracted from cells at different time points of the ex vivo culture and transcribed into cDNA using the GeneChip^® Whole Transcript Sense Target Labeling Kit (Affymetrix, High Wycombe, UK). The labeled cDNA was hybridized to GeneChip Human Gene 1.0ST Arrays, and the arrays were scanned and analyzed according to protocols of the manufacturer¹ as described in Ref. (31). Robust multiarray average signal extraction and normalization were performed, as detailed at the Bioconductor website² (32). The microarray data have been submitted to the GEO database under the accession number GSE95018.

Cultured cells (5 × 10⁵) were lysed in Trizol (QIAzol Lysis Reagent, Qiagen Biosciences, MD, USA) and stored at −80°C. Total RNA was extracted following ThermoFishers manual³, and 1 µg RNA was used for cDNA synthesis according to manufacturer’s protocol (RevertAid H Minus First Strand cDNA Synthesis Kit, ThermoFisher Scientific, MA, USA). Two hundred nanograms of transcribed cDNA were analyzed by real-time PCR using the KAPA SYBR FAST UNIVERSAL kit (Kapa Biosystems, Inc., Wilmington, MA, USA) and primer sequences as shown in Table S3 in Supplementary Material. As internal controls, primers for either hypoxanthine-guanine phosphoribosyltransferase or β-actin were used. Samples were measured and analyzed with QIAGEN’s real-time PCR cycler and corresponding software (Rotor-Gene Q, Qiagen, Hilden, Germany).

For protein-expression analysis, cell pellets (5 × 10⁵ cells/sample) were resuspended in 2× Laemmli buffer, boiled for 10 min at 95°C and subsequently separated by SDS-polyacrylamide gel electrophoresis (10% Bis-Tris gels, Acrylamide:Bis 37.5:1) and transferred to a nitrocellulose membrane (Amersham Protran Supported 0.45 µm NC, GE Healthcare Europe GmbH, Eindhoven, The Netherlands) by semi-dry electrophoretic blotting in Towbin buffer with 20% methanol (PerfectBlue, “Semi-Dry” Electroblotter Sedec, Peqlab, Southampton, UK). The membrane was blocked with 5% non-fat dry milk in TBS-T (0.2% Tween in TBS) followed by overnight incubation at 4°C with primary anti-ZNF683 antibodies (goat-anti-ZNF683 antibody C-12, 1:500, Santa Cruz Biotechnology, Heidelberg, Germany) or (mouse-anti-ZNF683 antibody, 1:500, Sigma-Aldrich, Saint Louis, MO, USA) and as control anti-GAPDH (mouse-anti-GAPDH, 1:10,000, Merck KGaA, Darmstadt, Germany). This was followed by 2 h incubation at room temperature (RT) with secondary antibodies, all 1:5,000 (donkey-anti-goat, Santa Cruz Biotechnology or goat-anti-mouse, Thermo Fisher Scientific) and corresponding washing steps with TBS-T.

To generate a lentiviral construct for knockdown studies, a sense oligo, 5′-TGGAAACACATGGGCTATGACATTTCAAGAGAATGTCATAGCCCATGTGTTTCTTTTTTC-3′, corresponding with the sequence “GAAACACATGGGCTATGACAT” to position 1,402 to 1,412 from the start ATG of ZNF683/HOBIT cDNA (NCBI Reference Sequence: NM_001114759.2) and a corresponding antisense oligo, 5′-TCGAGAAAAAAGAAACACATGGGCTATGACATTCTCTTGAAATGTCATAGCCCATGTGTTTCCA-3′ were synthesized (Integrated DNA Technologies Inc., Coralville, IA, USA) as complementary overlapping oligos, with a Xho1 overhang at the 5′end of the antisense oligo, as detailed.⁴ The complementary oligos were annealed and subsequently ligated into the LeGO-G/BSD lentiviral vector (LeGO-G/BSD was a gift from Boris Fehse, Addgene plasmid #27354) that had been digested with HpaI and XhoI (New England Biolabs, Ipswich, MA, USA). The ligated product was transformed into the E. coli strain Stbl3 (New England Biolabs, Ipswich, MA, USA), recombinant colonies were selected, and plasmid-DNA isolated and verified by restriction digestion and Sanger sequencing. As a control, oligos containing a scrambled shRNA were synthesized and cloned into the LeGO-G/BSD vector using an identical strategy. The control shRNA sequence was obtained from the “Open Biosystems pGIPZ shRNAmir library” situated at University College London.⁵ To confirm the functioning of the shRNA, shRNA expression plasmids were co-transfected into HEK293 cells in the absence or presence of a commercially available ZNF683/HOBIT expression plasmid (ORF of ZNF683, transcript variant 1, in pEnter, with C-terminal Flag and His tag; Vigene Biosciences, Rockville, MD, USA) employing the CaPO₄ method using 4 µg of total DNA/6-well. Six hours after, transfection medium was changed, and cells were harvested 96 h after transfection by either direct lysis in Trizol (Qiagen) for RNA isolation or 1× Laemmli sample buffer for protein-expression analysis.

Virus production and all experiments involving the generated recombinant viruses were performed in a separate room under biosafety level (BSL)-2 conditions following the instructions given by the corresponding directives of the European Union (Council Directive 90/679/EEC) and the Austrian government (GTG-BGBl. Nr. 510/1994 and 114/2012). The specific work with the replication-defective human lentiviruses was registered at the Austrian Ministry of Science, Research and Economy (BMWFW-5.011/009-WF/V/3b/2015).

To produce lentiviruses, HEK293T cells (ATCC # CRL-11268) were seeded at 10⁷ cells per 15 cm² tissue culture plate in 15 ml DMEM (DMEM/high glucose, GE Healthcare) supplemented with 10% FBS and glutamine (200 mM, Sigma-Aldrich). After 24 h, 5 ml DMEM was added, and the 60–70% dense cultures were transfected with the LeGO-G/BSD-ZNF683/HOBIT-shRNA expression plasmid plus three complementing plasmids providing the essential Gag, Pol, Rev, and Tat gene products missing in the replication-defective virus. To this end, the following 2 transfection solutions were prepared: (1) DMEM with 10% polyethyleneimine (PEI, Polysciences, Warrington, PA, USA) (1.5 ml/15 cm² plate) and (2) DMEM with 2.3 µg pCAG-KGP3R, 1.2 µg pCAG-RTR2, 2.5 µg pCAG-VSVg, and 24 µg vector of interest (in total 30 µg plasmids in 1.55 ml/15 cm² plate). After 15 min incubation at RT, solution 2 was sterile filtered and added to solution 1. The combined solutions were added dropwise to HEK293T cells at a density of 70%. After 12 h, the transfection mix was replaced with fresh medium (12 ml DMEM/15 cm² plate). The supernatants were harvested after 48 and 72 h post-transfection. To remove cell debris, supernatants were centrifuged (1,500 rpm, 5 min, 4°C), filtered (Puradisc FP30mm cellulose acetate syringe filter, 0.45 µm, sterile, GE Healthcare Life Sciences), and finally the virus particles were concentrated by ultracentrifugation for 90 min at 4°C at 25,000 rpm/76,000 × g in a SW-32 rotor equipped with 32 ml open-top thickwall polycarbonate tubes in a XL-70 ultracentrifuge (Beckman Coulter, Mystic, CT, USA). After ultracentrifugation, the supernatant was discarded leaving a 100 µl drop at the bottom of the tube. The virus particles were gently resuspended after keeping the tubes on ice for 1 h, aliquoted, and stored at −80°C until use.

After culturing freshly isolated cord blood CD34⁺ stem cells for 5 days in basal expansion medium supplemented with SCF, IL-7, TPO, and Flt3L as described under “Ex vivo differentiation of CD34⁺stem cells into NK cells,” the expanded cells were transduced with lentiviral particles. To this end, 24-well plates (multi-well plate for suspension culture, Greiner Bio-one GmbH, Frickenhausen, Germany) were coated with 8 µg retronectin/500 μl/well (recombinant human fibronectin fragment, Takara Bio Inc., Shiga, Japan) 24 h prior virus transduction and stored at 4°C until use. The retronectin solution was removed, 250 µl GBGM containing virus particles (40 Transduction Units/cell) were added per well, and the plate was centrifuged at 4,000 rpm/1,900 × g for 2 h at 4°C. Immediately after centrifugation, the 5 days cultivated CD34⁺ stem cells were added (5 × 10⁴ cells in 250 µl basal expansion medium/well). The following day, 500 µl of basal expansion medium was added. Two days after transduction the cells were centrifuged (1,500 rpm for 5 min), suspended in fresh basal medium, and transferred to new uncoated 24-well cell culture plates (Corning^® Costar^®, Sigma-Aldrich). Transduction efficiency, determined as GFP⁺ cells, was measured by flow cytometry 3 days after transduction, and cultures were continued as described in the first section (for ex vivo differentiation). Further cell expansion was calculated as follows. Cell numbers were obtained at the various differentiation stages and divided by the cell numbers for the corresponding GFP⁺ and GFP⁻ fractions measured 3 days after transduction, to correct for different transduction efficiencies.

Cell surface expression of NK and monocytic markers was monitored using anti-CD56-PeCy7, anti-CD14-PerCPCy5.5 (all BD Biosciences, San Jose, CA, USA), and anti-KIR-PE antibodies (R&D Systems, Vienna, Austria) on a FACS Canto II (BD Biosciences) and data were analyzed using both FACS DIVA software v6.0 (BD Biosciences) and Flowjo v10.0.8 (Tree Star, Yorba Linda, CA, USA). Details on antibodies are given in Table S4 in Supplementary Material.

Target K562 and effector NK cells were cocultured at a 1:1 ratio (15 × 10⁴ cells of each cell type) in 200 µl RPMI medium (RPMI 1640 medium, Life Technologies, Carlsbad, CA, USA) in an U-bottom 96-well plate (Greiner Cellstar^® 96-well plates, Sigma-Aldrich) in the presence of anti-CD107a-APC (BD Biosciences). Brefeldin A and Monensin (BD Golgiplug and Golgistop, BD Biosciences) were added after 1 h of culture. After an additional 5 h of culture, cells were collected, stained for surface CD56 (anti-CD56-PeCy7, BD Biosciences), subsequently prepared for intracellular staining with IFN-γ (anti-IFN-γ-PE, BD Biosciences) using a fixation/permeabilization solution kit (BD Biosciences), and finally measured on a FACS Canto II.

Statistical analysis was performed with Prism 6 software (GraphPad, San Diego, CA, USA) using Student’s t-test or a two-way ANOVA as indicated in Figure Legends. A p-value of 0.05 was considered as statistically significant.

Initially, we were interested to identify novel transcription factors potentially contributing to human NK-cell differentiation that have not been described in this function before. For this purpose, we employed a recently developed ex vivo differentiation system (29). In this system, cord blood CD34⁺ stem cells are initially expanded for 10 days, prior to addition of IL-15. Following further addition of IL-2 from day 14, the differentiating cells are cultured for a total period of 35–42 days, by which time a regular culture comprises over 95% NK cells. To determine the repertoire of transcription factors differentially expressed during NK-cell development, we first performed a transcriptomic profiling study comparing samples from different time points up to day 35 with cultures at day 10 (start of the NK-cell differentiation). The 20 most highly upregulated transcription factor genes detected are shown in Table S1 in Supplementary Material. When real-time RT-PCR was performed for the corresponding transcripts, ZNF683/HOBIT mRNA was found to be by far the most highly upregulated mRNA at day 35 in relative terms (about 9,000-fold; Figure 1A). Among the most differentially expressed genes are many that have previously been reported to be important in NK-cell development and/or maturation such as GATA3 (33), TOX (34), ID2 (10), and ETS1 (16) (Figure 1A). Based on its near absence in the stem cell cultures before addition of IL-15 and its high upregulation during differentiation, we decided to focus our further studies on ZNF683/HOBIT, especially as the role of this transcription factor in NK-cell development has not previously been reported.

To determine if ZNF683/HOBIT is also highly expressed in adult peripheral blood NK cells, we performed real-time RT-PCR on different peripheral blood leukocyte subsets of multiple donors. Indeed, compared to T and B lymphocytes as well as monocytes, ZNF683/HOBIT mRNA is predominantly expressed in NK cells (Figure 1B). CD19⁺ B lymphocytes display lower levels of ZNF683/HOBIT mRNA (about 7-fold less as compared to CD56^dim NK cells) and the levels in T lymphocytes and monocytes are even much lower or at detection limit. We were furthermore interested to compare the ZNF683/HOBIT mRNA levels between the more immature CD56^bright and the rather mature CD56^dim subpopulations of NK cells. The data show that CD56^bright cells express the highest levels of ZNF683/HOBIT mRNA, whereas CD56^dim cells display about 20% lower levels (Figure 1B).

To define the potential role(s) of ZNF683/HOBIT throughout NK-cell development, we evaluated the kinetics of accumulation of ZNF683/HOBIT mRNA during ex vivo differentiation. During this process, three different cell types can be distinguished based upon the monocytic and NK-cell markers, CD14 and CD56, respectively (28). One of the subpopulations arising from the amplifying CD34⁺ stem cells consists of monocytic CD14⁺ cells that accumulate until day 14 after which they gradually disappear (Figure 2A). The second population consists of CD56⁻CD14⁻ cells, which peak at day 18. These CD56⁻CD14⁻ cells are presumed to comprise at least in part the NK-cell progenitors as indicated by their rapid decrease from day 18 onward, along with the concomitant rapid generation of CD56⁺ NK cells (Figure 2A). The CD56⁺ NK cells become the predominant population after day 25 and from day 35 onward the culture consists of over 95% CD56⁺ NK cells (Figure 2A).

We obtained samples at different time points during the ex vivo cultures and separated CD14⁻ and CD14⁺ cells by flow cytometry. We detected that ZNF683/HOBIT mRNA is nearly absent after the initial expansion phase of 10 days. A tremendous accumulation of ZNF683/HOBIT mRNA is observed from day 24 onward in CD14⁻ cells comprising NKPs and CD56⁺ NK cells (Figure 2B). The magnitude of ZNF683/HOBIT mRNA accumulation actually blurs the onset of expression in the display of Figure 2B. When plotted at a higher magnification, it is apparent that ZNF683/HOBIT mRNA increased in CD14⁻ cells already until day 17 followed by a first exponential accumulation up to day 21 (Figure 2C). Although we observed low expression of ZNF683/HOBIT mRNA in the CD14⁺ monocytes at day 17, the following first accumulation is absent (Figure 2C) as well as the exponential increase later during differentiation (Figure 2B).

After observing that ZNF683/HOBIT mRNA is mainly present in NK cells and their progenitors, we set out to elucidate the potential role(s) of ZNF683/HOBIT during NK-cell development and maturation in more detail. For this purpose, we performed loss-of-function studies using a lentiviral-based shRNA-mediated knockdown of ZNF683/HOBIT. The appropriate functioning of the shRNA was confirmed by shRNA-mediated knockdown of ectopically expressed ZNF683/HOBIT in HEK293T cells. As shown in Figure 3A, expression of ZNF683/HOBIT mRNA and protein was strongly reduced (>90%) upon co-expression of the shRNA, thus confirming the proper functioning of the shRNA (Figure 3A).

We then continued to investigate the effects of ZNF683/HOBIT knockdown on ex vivo differentiation cultures. These cultures were transduced with lentiviruses expressing either ZNF683/HOBIT shRNA (shHOBIT) or a scrambled control shRNA (shControl). Cultures, mock-treated for the transduction procedure but without addition of viruses, were performed in parallel. Based on initial experiments that evaluated best transduction rates and survival in relation to days in culture, cells were transduced during the stem cell expansion phase 5 days after isolation. The presence of a GFP cassette driven by a spleen focus-forming virus (SFFV)-promoter enabled us to discriminate transduced GFP⁺ cells expressing the shRNA from non-transduced GFP⁻ cells within the same culture in parallel (35). Regularly, transduction rates between 30 and 50% were achieved. Since these varied to some extent, we normalized obtained expansion rates for the analyzed fractions to the different transduction rates measured 3 days after transduction (8 days after isolation).

Based on the fact that the expression of the NK-cell marker CD56 becomes significant only from day 24 onward, we decided to split the analysis in two parts. In the first part, we analyzed the NKP stage until day 21, in the second part the generation and maturation of CD56⁺ NK cells. At day 21, we observed a fourfold reduced expansion of the shHOBIT GFP⁺ cells compared to the shControl GFP⁺ cells (Figure 3C). Exemplary dot plots of the GFP staining are shown in Figure 3B. The GFP⁻ cells of the shHOBIT culture also seemed to be somewhat reduced in comparison to the scrambled control and regular mock cultures, but the difference did not reach statistical significance.

Since at day 21 the culture comprises mainly two different cell types, CD56⁻CD14⁻ progenitor cells and CD14⁺ monocytic cells (Figure 2A), we continued to analyze these populations separately. Exemplary dot plots of this analysis are displayed in Figure 4A. The data showed that the effect can be mainly traced to a fourfold reduced expansion of the CD56⁻CD14⁻ progenitors in the shHOBIT GFP⁺ fraction. Although smaller inhibitory effects were also visible for CD14⁺ cells of the shHOBIT GFP⁺ pool and the GFP⁻ cells of the shHOBIT culture, both did not reach statistical significance (Figure 4B). However, this may indicate that a strong inhibitory effect on a fraction of the culture indirectly affects the whole culture, including the GFP⁻ part.

The number of CD56⁺ NK cells was too low to be analyzed in detail at this time point of the culture (Figures 4A,B). Together, the data clearly showed that knockdown of ZNF683/HOBIT reduces the expansion of presumptive NKPs already at day 21 of the culture. This suggests a role of ZNF683/HOBIT in the early stages of human NK-cell development before the initiation of CD56 expression.

The second part of our analysis focused on the formation of CD56⁺ cells and their maturation from day 21 onward. At first sight, when we analyzed the percentages of CD56⁺ NK cells formed during later stages at day 35, both GFP⁺ populations (shHOBIT and shControl) seemed comparable to regular cultures (>95% CD56⁺ NK cells, Figure S1 in Supplementary Material). However, when the number of cells was taken into account, it became apparent that the generation of CD56⁺ NK cells from the GFP⁺ progenitors in the shHOBIT pool is substantially reduced compared to the GFP⁺ control shRNA fraction. Whereas the GFP⁺CD56⁺ NK cells in the control shRNA cultures showed a strong expansion between days 21 and 35, the GFP⁺CD56⁺ NK cells in the shHOBIT cultures displayed almost no increase beyond day 21 and remained low in number (around 20% of the shControl cultures). These cells did not enter the typical expansion phase between days 24 and 35 (Figures 5A,B). In contrast, the GFP⁻ cells in the shHOBIT cultures displayed an expansion phase between days 21 and 30, although this expansion seemed reduced in comparison to the shControl cultures. Again this may suggest an overall indirect effect of the strongly compromised GFP⁺ cells on the GFP⁻ part of the cultures.

Taken together, our data showed that knockdown of ZNF683/HOBIT strongly abrogates the generation of CD56⁺ NK cells. This suggests that ZNF683/HOBIT downmodulation strongly affects NK-cell progenitors shortly before the initiation of CD56 expression and prevents the progression toward CD56⁺ NK cells and/or the further proliferation of the few detected CD56⁺ NK cells.

We were further interested in the phenotype of the few detected CD56⁺ NK cells. To this end, we evaluated if these cells would further differentiate/mature via upregulation of NKG2A and KIR. The expression of these important NK-cell receptors was measured after 35 days of cultivation and compared between the GFP⁺ cells of the shHOBIT and the shControl cultures. We observed high levels of NKG2A in all of the GFP⁺ cells of the cultures independent of shHOBIT or control shRNA expression (Figure 6A, exemplary dot plots in Figure S2 in Supplementary Material). The GFP⁻ fractions also displayed high NKG2A levels similar to regular control cultures. Similarly, KIR expression was not significantly influenced by the presence or absence of shHOBIT or control shRNA (Figure 6A, exemplary dot plots in Figure S2 in Supplementary Material) and was comparable to the typical levels of regular cultures. In this regard, our group previously showed that ex vivo generated NK cells express lower levels of KIR than peripheral blood-NK cells (28). Together, these observations demonstrate that cells starting to express CD56 continue to develop normally in regard of NKG2A and KIR expression. Furthermore, it implies ZNF683/HOBIT does not influence the expression of NKG2A or KIR receptors on CD56⁺ NK cells.

Two main functions of NK cells are cytotoxicity, the direct killing of target cells, and IFN-γ production (1, 36). To evaluate whether the CD56⁺ NK cells differentiating under knockdown conditions for ZNF683/HOBIT would functionally mature, we tested degranulation capacity in the CD107a assay and IFN-γ production. Following coculture with the target cell-line K562, we detected similar levels of CD107a on the cell surface of all cells irrespective of the culture conditions (Figure 6B, exemplary dot plots in Figure S2 in Supplementary Material). In the same cultures, we also determined the proportion of cells with intracellular IFN-γ. Intriguingly, the transduction with the scrambled control shRNA already caused a substantial increase in IFN-γ producing cells (up to 50%). This was further enhanced upon expression of the ZNF683/HOBIT shRNA leading to 80% IFN-γ-producing cells (Figure 6C). In contrast, only a very small fraction of the GFP⁻ cells was competent to produce IFN-γ (Figure 6C). In summary, transduction with shRNA lentiviruses by itself leads to a higher proportion of NK cells with the capacity to produce IFN-γ and this is further increased by knockdown of ZNF683/HOBIT. The ZNF683/HOBIT shRNA effect is restricted to an increase in the number of IFN-γ-producing cells and not reflected in the level of IFN-γ per cell as measured by mean flourescence intensity (MFI). However, virus transduced cells displayed higher levels of IFN-γ per cell in comparison to non-transduced cells (Figure S3 in Supplementary Material).